- Email: [email protected]
The UVIS scintillation detector. A proposed method of nuclear recoil discrimination for dark matter searches
Physics Letters B 314 (1993) 430-435
North-Holland
PHYSICS LETTERS B
The UVIS scintillation detector. A proposed method of nuclear recoil discrimination for dark matter searches N.J.C. Spooner Astrophysics, NuclearPhysics Laboratory, KebleRoad, Oxford OXI 3RH, UK
and P.F. Smith Rutherford Appleton Laboratory, Chilton, Oxon. OXl l OQX, UK
Received 6 May 1993 Editor: L. Montanet
The proposed UVIS scintillation detector is based on the emission of both UV ( 300 nm ) and visible (420 nm) photons by lowTl Nal crystals at temperatures of ~ 100 K, in different proportions for electron and nuclear recoils. Monte Carlo simulations of typical experimental geometries suggest the possibility of a two orders of magnitude background rejection at an energy threshold of ~ l0 keV for Na recoils and ~ 30 keV for I recoils. Options based on wavelength filtering and/or photon timing are discussed. i
The detection of weakly interacting massive particles that may constitute the Galactic dark matter requires detectors sensitive to nuclear recoil events in the range l-100 keV at rates typically 0.0 l-1 events/ day/kg. Studies o f possible experiments have been in progress since 1985, initially based on ionization detectors and low temperature bolometric techniques [ 1,2]. Scintillator/photomultiplier systems seemed initially less attractive for this low energy application owing to the difficulty of achieving good light collection while at the same time avoiding a high background rate from radioactivity in the photomultipliers. However, scintillators offer the potential advantages of larger target masses and simplicity of construction. This has led to active consideration of dark matter detectors based on crystal scintillators, organic liquid scintillators, and liquid noble gas scintillators, using various arrangements of light guide and shielding to achieve good light collection while minimising background from photomultipliers. Experiments now planned include those of the U K Collab430
oration (NaI, liquid scintillators and Xe) [ 3,4 ], the Beijing/Rome/Saclay Collaboration (NaI and Xe) [ 5 ] and the Osaka group in Japan ( N a I ) [ 6 ]. Target masses in the range 20-100 kg are envisaged in order to search for the annual modulation effects which would help confirm the Galactic origin of any signal. NaI is of particular interest as a dark matter detector because o f its high light output/keV and availability as large mass crystals. Since the predominant isotopes of NaI are 23Na and 127I,both with nuclear spin, this target also provides sensitivity to W I M P candidates with spin-dependent interactions, while the high atomic number of the iodine provides good coherent enhancement of spin-independent interactions. All currently proposed clark matter detectors are limited in sensitivity by intrinsic radioactive backgrounds. Of these backgrounds the contribution from muon-induced neutrons can be eliminated by siting the experiments underground. Remaining background is then due to photons and beta decay arising from radioactivity in the surroundings and in the detector itself. The environmental activity external to Elsevier Science Publishers B.V.
Volume 314, number 3,4
PHYSICS LETTERS B
the detector can be attenuated by sufficiently high purity shielding, such as high-purity lead, copper or de-ionized water [7]. A lower limit to the counting rate then results from beta decay and photons due to natural or cosmogenic radioactive impurities in the target material. The residual activity internal to a detector makes it difficult to reach levels as low as 1 event/kg/day by simple counting techniques, and it is therefore necessary to develop techniques which will identify specifically the nuclear recoil events and reject the majority of electron recoil events arising from photons and beta decay. A well-known example of such a technique is the "hybrid" detector based on the different ratio of ionization and bolometric signals in a semiconducting target [8]. However, this involves operation at temperatures below 100 m K and is potentially costly to scale up to larger target masses. A possible alternative is to perform discrimination in scintillator targets. This has been achieved in certain scintillator materials via pulse shape, utilising the existence of "fast" and "slow" scintillation components, stimulated to different extents by nuclear and electron recoils [ 9 ] and so providing a basis for neutron/photon or alpha/photon discrimination [ 9,10 ]. Typical liquid and organic crystal scintillators used in this way, have time constants in the 1-200 ns region, with mean decay times differing by up to a factor ~ 2 for excitation by different types of particle. However, this rather small difference between the decay times means that efficient discrimination can only usually be achieved at well above energy threshold, where sufficient photons are available to form a continuous light profile. The technique has also been applied to alkali halide scintillators, including NaI(T1) [ 11 ]. However, in NaI(T1) the time constants involved are longer than is typical in organic scintillators and have fast and slow components differing by a smaller factor so that discrimination at low energies is further restricted. In this paper we propose an alternative discrimination technique for NaI (T1) called the "UVIS" detector system, based on differences in the UV and VISible parts of the spectrum of the scintillation light for different particles. This possibility arises because the mechanisms which relate the pulse shape to the rate of energy loss, dE/dr, in a scintillator for different particles can also determine the scintillation
23 September1993
emission spectrum. Thus the shape of the spectrum can be expected to be significantly different for nuclear and electron recoils. This provides the basis of an alternative possibility for achieving background discrimination in a dark matter experiment with the benefit of potentially lower threshold energy. The principle of discrimination between electrons, deuterons and alphas by wavelength filtering has been demonstrated previously using CsI(TI), but only at energies in the region of 1 MeV [ 12 ]. In assessing the prospects for using this technique in the keV region, we prefer to avoid consideration of CsI (T1) as a low background target material due to contamination by the long-lived radioactive isotope ~37Cs. The preferred alternative o f N a I (T1) also presents problems. In conventionally doped NaI ( 10 -3 T1) at room temperature the differences in spectral response are small because for all types of particle the emission is dominated by a single peak at 420 nm. Both the fast and slow scintillation decays occur at this wavelength. However, measurements in the 1960s by Van Sciver [ l 3 ] showed that by using NaI (T1) with much lower T1 doping (typically 10 - 6 TI) and operating at temperatures of ~ 100 K a second peak is present at 300 nm, well separated from the 420 nm peak. These 300 nm photons are associated with the intrinsic emission found in pure undoped NaI. Measurements were made with both incident MeV alphas and incident MeV gammas, and the ratio of the intensities of the 420 nm and 300 nm peaks was found to be strongly dependent on both panicle type and temperature. The largest difference between alpha and gamma emission spectra was found at temperatures in the region 100-120 K, the results having the simplified form illustrated in fig. 1. The integrated intensities in the two peaks were in the approximate ratio 1:1 for alphas and 4:1 for gammas. Assuming the intensities are approximately linear with energy, that the intensity ratios can be applied as true probabilities on a photon by photon counting basis, and that the response to alpha particles is broadly equivalent to the response to recoiling Na and I ions expected from WIMP interactions, then we can expect the intensity ratios to apply in the keV energy range appropriate for dark matter detection. An estimate can then be made of the degree of background rejection that might be possible by filtering and separately counting photons from the two peaks. 431
Volume 314, number 3,4
PHYSICS LETTERSB
PHOTONS
ALPHAS
NaI
A
300
420
nm
300
421)
300
420
nm
300
420
(b) Nal pure
nm
Fig. 1. Form of the emission spectra due to alpha particles and gamma ray photons interacting in (a) NaI( l0 -6 Tl) at ~ 100 K and (b) NaI(pure) at ~ 100 K, adapted from Van Sciver [ 13].
edge filter edge filler -x 100Ke n c l o s u r e /
(a)
(b)
300nm 42Ohm
300nm 420am
Fig. 2. Possible wavelength-discriminating Nal systems using: (a) 2 photomultipliers and (b) 4 photomultipliers. Details of light guides and shielding not shown.
We have carried out Monte Carlo simulations for typical systems consisting of a sodium iodide crystal with two or more windows linked to photomultipliers via light guides. Figs. 2a and 2b show schematic designs with 2 and 4 photomultipliers. Filters select either the UV or visible component for each photomultiplier. In practice one has the choice between absorption filters, which simply absorb that component of the light not transmitted, and edge interference filters, which transmit one component and reflect the other back into the system. Another option 432
23 September1993
would be to omit the filters and distinguish the UV and visible components by photon timing, using the large time constant difference discussed below. For purposes of the simulation this choice does not have to be m a d e - one can simply assume a value for the overall light collection efficiency. For computational purposes we take this to be 20%, multiplied by a photomultiplier efficiency of ~ 20% to give an overall photon detection efficiency of 4%, divided equally between the two wavelength channels. However, these illustrative assumptions introduce no loss of generality, since our results will be expressed in terms of the number of photons detected, and can thus be applied to systems of higher or lower detection efficiency. To simulate the discrimination process, visible and UV photons are produced at random, in the proportions 0.8/0.2 for electron recoil events and 0.5/0.5 for nuclear recoil events. Each has the specified detection probability in each channel and the fraction of UV counts is recorded as a function of the absolute number of photons detected. A discrimination level rd for this ratio is chosen (e.g. ra=0.6) below which the events are assumed to be nuclear recoils. However, statistical fluctuations will result in some fraction of the background events also falling below the discrimination level. Denoting the fraction of nuclear recoil events falling below ra byf~, and the fraction of misidentified background byfb, the ratio fb/fn provides a measure of the potential background-reducing capability of the technique. The background reduction factor improves rapidly with increasing number of photons, but can also be improved, for a given number of photons, by setting the discrimination level lower. This reduces the misidentified events but at the expense of losing more of the true events. Fig. 3 shows the results for a simple two-channel system with the discrimination level ro set at 0.6, 0.5, 0.4 and 0.3. Note that the reduction in collection efficiency f , is in each case compensated by the greater improvement in fb, but at the expense of a longer running time. The initial rise in both curves is due to the inclusion of the additional constraint that a coincidence between at least one photon in each photomultiplier is required to define an event. The loss of low energy events and discriminating power due to this constraint can in principle be improved by using two photomultipliers for each wavelength,
V o l u m e 314, n u m b e r 3,4
IO0[
'
'
traction of [ events
'
¢a)
PHYSICS LETTERS B
"
'
~
'
'
'
(b) __
'
'
'
'
'
'
"
'
•
•
'
(c)
alO-~
lO -4 1
0
.l
.
{
4
,
,
,
l
•
•
•
8
'
12
"~
•
•
'
16
. TM
20
photons Fig. 3. Results for the b a c k g r o u n d - r e d u c i n g c a p a b i l i t y of a s i m p l e 2-channel N a l s p e c t r u m d i s c r i m i n a t i n g system with the discrimi n a t i o n level ra set at ( a ) 0.6, ( b ) 0.5, (c) 0.4 a n d ( d ) 0.3.
as illustrated in fig. 2b, since an event can then also be registered by a coincidence between two photons of the same wavelength. As an alternative to the arrangement in fig. 2b one might consider a rectangular crystal with four windows. Even with the simple arrangement of fig. 2a, however, it can be seen that at least one order of magnitude reduction in background could be achieved at a threshold of only 2 photons in each channel. With the collection efficiency assumed, and a light emission in NaI (10 -6 T1) of > 40 photons/keV at the reduced temperature [ 13 ], this would correspond to an observed scintillation threshold < 3 keV. As stated above, the calculations assume the scintillation response for alphas is equivalent to the response for recoiling I and Na that would result from WlMP interactions. Combining the result with measurements of the actual scintillation response of recoiling I and Na atoms in NaI relative to alphas at low energy obtained from neutron scattering experiments [14], leads to a higher expected recoil threshold due to WIMPs (typically < I0 keV f o r N a and <30 keV for I). However, these measurements, together with measurements of the scintillation efficiency of NaI (T1) for various heavy ions, also suggest that there
23 S e p t e m b e r 1993
may be a larger difference between the UV/visible ratios for heavy ions relative to electrons than for alphas relative to electrons [ 14,15 ]. This would tend to improve the discrimination and hence lower the threshold again, so that the discrimination of nuclear recoils for typical dark matter targets (i.e. A>~ 19) may in practice be better than the above simulations based on alpha/electron data. There will be additional losses associated with the filters and their angle-dependent transmission characteristics (typical efficiency probably 40%-70% depending on the type of absorption or interference filters used) which would tend to increase the scintillation threshold. However, there is also scope for reducing this by improved light collection schemes. Fig. 4 shows results for a simulation of a 7 5 × 7 5 × 7 5 mm crystal in which electron and nuclear recoil events are shown on a scatter plot in terms of the numbers of 300 nm and 420 nm photons collected in the two channels of a typical experiment, calibrated with respect to electron equivalent energy (keVee). The discrimination ratio is shown as a line separating the two classes of events resulting from nuclear recoil and background electron recoils. Because of the expected shape of a dark matter nuclear recoil spectrum, there is a rather rapid decrease in sensitivity with increasing energy threshold, making it normally the case that the best dark matter lim-
50 energy in 300 nm channel 40 (keVee)
n e u t r o n s
30
• ?7. : " : .?.. - L. :
• ? iiii ?
0
i i ¢ ~ . "
0
10
•
,"
.
.
,
20
30
40
50
60
energy in 420 am channel (keV~) Fig. 4. Results o f a s i m u l a t i o n of electron a n d nuclear recoil disc r i m i n a t i o n in a 75 × 75 X 75 m m NaI crystal. 300 n m and 420 n m p h o t o n s collected in a typical two channels e x p e r i m e n t are shown, calibrated with respect to electron e q u i v a l e n t energy ( keVoe ).
433
Volume 314, number 3,4
PHYSICS LETTERS B
its are set at the detector energy threshold [ 2 ]. However, it can he seen from fig. 3 ( a n d the simulation o f fig. 4) that a rapid gain in background discrimination results as one moves away from threshold to higher energies and larger numbers o f collected photons. F u r t h e r m o r e the absolute background decreases with increasing energy. It is quite probable, therefore, that better limits will be set at energies somewhat above threshold. Because o f the several factors involved, it will require further analysis to establish the o p t i m u m operating point, and the net gain achievable, in specific experiments but a two orders o f magnitude i m p r o v e m e n t seems within reach with this technique. A further gain m a y he possible by measuring the time structure of the emitted photons. The time constants for NaI (T1), at various T1 doping levels and temperatures, have been reported by a n u m b e r o f authors though the data remains incomplete and somewhat confused [ 9,1 l, 13,15,16 ]. The UV photons are characteristic o f the crystal lattice and have a single d o m i n a n t constant varying only slowly with temperature being ~ 30 ns at 100 K. However, the 420 nm photons, enhanced by the T1 doping, involve several processes with t e m p e r a t u r e - d e p e n d e n t time constants ranging from the ns to the ms range. These appear in different proportions for different particle types, due to the differences in ionization density (in relation to the relatively large spacing o f the Tl a t o m s in the lattice) and to the different efficiencies for ionization and exciton mechanisms. In N a I ( 1 0 .3 T1) (usual doping level) at room temperature, the differences in the 420 nm emission band are fairly small: alphas give a fast rise followed by a 230 ns decay, while photons give a 60 ns rise, 150 ns plateau, and 230 ns decay [ 11 ]. However, suppression o f some processes at reduced temperature leads to longer time constants in the 420 nm band, with the p r e d o m i n a n t c o m p o n e n t s likely to extend into the las range [ 16 ], and with the possibility o f larger time constant differences within the 420 nm emission. This prospect o f increased differences between the various particle d e p e n d e n t time constants in the 420 nm b a n d may allow further d i s c r i m i n a t i o n between particle types by photon timing analysis that would be a d d i t i o n a l to the photon wavelength d i s c r i m i n a t i o n technique. In this way a p r o p o r t i o n o f those background events misidentified through the wavelength ratio test could 434
23 September 1993
be recovered through a likelihood test on the p h o t o n arrival times. F r o m this preliminary study we conclude that wavelength-discriminating techniques offer the prospect o f low energy nuclear recoil discrimination in a sodium iodide target, with the advantage c o m p a r e d with discrimination techniques at m K temperatures of achieving target masses > 100 kg using relatively inexpensive technology. N.J.C.S. wishes to acknowledge support for this work under the SERC advanced fellowship scheme.
References [ 1] J.R. Primak, D. Seckel and B. Sadoulet, Annu. Rev. Nucl. Part. Sci. 38 (1988) 751. [2 ] P.F. Smith and J.D. Lewin, Phys. Rep. 187 (1990) 203. [ 3 ] See, for example, T.J. Sumner, Nucl. Phys. B 22 ( 1991 ) 165; T.J. Sumner et al., Proc. 22nd Intern. Cosmic Ray Conf. 4 (1991) 722; P.F. Smith et al., UK particle physics proposal 270 stage 2 (1991). [4] G. Davies et al., to be published. [5 ] See, for example, P. Belli et al., Proc. XXVth Rencontre de Moriond, Particle astrophysics: the early universe and cosmic structures, eds. J.M. Alimi et al. (Editions Frontieres, Gif-sur-Yvette, 1990) p. 225; Nuovo Cimento 103A (1990) 767; C. Belli et al., Proc. XXVth Rencontre de Moriond, Massive neutrinos, tests of fundamental symmetries, eds. O. Fackler et al. (Editions Frontieres, Gif-sur-Yvette, 1991 ) p. 185; L. Mosca et al., Nucl. Phys. B (Proc. Suppl.) 28A (1992) 302. [ 6 ] K. Fushimi et al., Phys. Rev. C 47 ( 1993 ) R 425. [7] A. Allesandrello et al., Nucl. Instrum. Methods B 61 ( 1991 ) 106; R.L. Brodzinski et al., NucL Instrum. Methods A 292 (1990) 337; N.J.C. Spooner et al., Proc. 21st Intern. Cosmic Ray Conf. l0 (1990) 264; P. Barnes et al., Proc. 1990 Summer Study on High energy physics, research directions for the decade, ed. E.L. Berger (World Scientific, Singapore, 1992) p. 311. [8 ] N.J.C. Spooner et al., Phys. Lett. B 273 ( 1991 ) 333; T. Shuttet al., Phys. Rev. Lett. 69 (1992) 3425, 3531. [9] See, for example, J.B. Birks, Theory and practice of scintillation counting (Pergamon, Oxford, 1964); W.H. Tait, Radiation detection (Butterworths, London, 1980); G.F. Knoll, Radiation detection and measurement (Wiley, New York, 1979); F.D. Brooks, Nucl. Instrum. Methods 162 (1979) 477.
Volume 314, number 3,4
PHYSICS LETTERS B
[ 10] See, for example, I.B. Berlman and O.J. Steingraber, Nucl. Instrum. Methods 108 (1973) 587; D.B.C.B. Syme and G.I. Crawford, Nucl. Instrum. Methods 104 (1972) 245; T.G. Miller, Nucl. Instrum. Methods 63 (1968) 121. [ 11 ] R.B. Owen, Nucleonics 17 (1959) 92; P. Belli et al., Nucl. Instrum. Methods A 294 (1990) 391; G. Gerbier, Proc. XXVth Rencontre de Moriond, New and exotic phenomena '90, eds. O. Fackler et al. (Editions Frontieres, Gif-sur-Yvette, 1990) p. 469; P. Doll et al., Nucl. Instrum. Methods A 285 (1989) 464. [ 12] G. Hrehuss, Nucl. Instrum. Methods 8 (1960) 344. [13]W. Van Sciver, IRE Trans. Nucl. Sci. NS-3 (1956) 39; Nucleonics 14 (1956) 50; Phys. Rev. 120 (1960) 1193; Phys. Lett. 9 (1964) 97. [ 14] N.J.C. Spooner et al., Phys. Lett. B, to be published.
23 September 1993
[ 15 ] See, for example, S.K. Allison and H. Casson, Phys. Rev. 90 (1953) 880; E. Newman and F.E. Steigert, Phys. Rev. 118 (1960) 1575. [ 16] See, for example, V.I. Startsev, Z.B. Baturicheva and Yu.A. Tsirlin, Opt. Spectrosc. 8 (1960) 286; N.N. Vasil'eva and Z.L. Morgenshtern, Opt. Spectrosc. 12 (1962) 41; D.E. Persek et al., IEEE Trans. Nucl. Sci. 27, no. 1 (1980) 168; P. Schotanus, R. Kamermans and P. Dorenbos, IEEE Trans. Nucl. Sci. NS-37, no. 2 (1990) 177; J.C. Robertson and J.G. Lynch, Proc. Phys. Soc. 77 ( 1961 ) 751; J.S. Schweitzer and W. Ziehl, IEEE Trans. Nucl. Sci. NS30, no. 1 (1983) 380. [ 17 ] J. Bonanomi and J. Rossel, Helv. Phys. Acta 25 ( 1952 ) 725.
435
North-Holland
PHYSICS LETTERS B
The UVIS scintillation detector. A proposed method of nuclear recoil discrimination for dark matter searches N.J.C. Spooner Astrophysics, NuclearPhysics Laboratory, KebleRoad, Oxford OXI 3RH, UK
and P.F. Smith Rutherford Appleton Laboratory, Chilton, Oxon. OXl l OQX, UK
Received 6 May 1993 Editor: L. Montanet
The proposed UVIS scintillation detector is based on the emission of both UV ( 300 nm ) and visible (420 nm) photons by lowTl Nal crystals at temperatures of ~ 100 K, in different proportions for electron and nuclear recoils. Monte Carlo simulations of typical experimental geometries suggest the possibility of a two orders of magnitude background rejection at an energy threshold of ~ l0 keV for Na recoils and ~ 30 keV for I recoils. Options based on wavelength filtering and/or photon timing are discussed. i
The detection of weakly interacting massive particles that may constitute the Galactic dark matter requires detectors sensitive to nuclear recoil events in the range l-100 keV at rates typically 0.0 l-1 events/ day/kg. Studies o f possible experiments have been in progress since 1985, initially based on ionization detectors and low temperature bolometric techniques [ 1,2]. Scintillator/photomultiplier systems seemed initially less attractive for this low energy application owing to the difficulty of achieving good light collection while at the same time avoiding a high background rate from radioactivity in the photomultipliers. However, scintillators offer the potential advantages of larger target masses and simplicity of construction. This has led to active consideration of dark matter detectors based on crystal scintillators, organic liquid scintillators, and liquid noble gas scintillators, using various arrangements of light guide and shielding to achieve good light collection while minimising background from photomultipliers. Experiments now planned include those of the U K Collab430
oration (NaI, liquid scintillators and Xe) [ 3,4 ], the Beijing/Rome/Saclay Collaboration (NaI and Xe) [ 5 ] and the Osaka group in Japan ( N a I ) [ 6 ]. Target masses in the range 20-100 kg are envisaged in order to search for the annual modulation effects which would help confirm the Galactic origin of any signal. NaI is of particular interest as a dark matter detector because o f its high light output/keV and availability as large mass crystals. Since the predominant isotopes of NaI are 23Na and 127I,both with nuclear spin, this target also provides sensitivity to W I M P candidates with spin-dependent interactions, while the high atomic number of the iodine provides good coherent enhancement of spin-independent interactions. All currently proposed clark matter detectors are limited in sensitivity by intrinsic radioactive backgrounds. Of these backgrounds the contribution from muon-induced neutrons can be eliminated by siting the experiments underground. Remaining background is then due to photons and beta decay arising from radioactivity in the surroundings and in the detector itself. The environmental activity external to Elsevier Science Publishers B.V.
Volume 314, number 3,4
PHYSICS LETTERS B
the detector can be attenuated by sufficiently high purity shielding, such as high-purity lead, copper or de-ionized water [7]. A lower limit to the counting rate then results from beta decay and photons due to natural or cosmogenic radioactive impurities in the target material. The residual activity internal to a detector makes it difficult to reach levels as low as 1 event/kg/day by simple counting techniques, and it is therefore necessary to develop techniques which will identify specifically the nuclear recoil events and reject the majority of electron recoil events arising from photons and beta decay. A well-known example of such a technique is the "hybrid" detector based on the different ratio of ionization and bolometric signals in a semiconducting target [8]. However, this involves operation at temperatures below 100 m K and is potentially costly to scale up to larger target masses. A possible alternative is to perform discrimination in scintillator targets. This has been achieved in certain scintillator materials via pulse shape, utilising the existence of "fast" and "slow" scintillation components, stimulated to different extents by nuclear and electron recoils [ 9 ] and so providing a basis for neutron/photon or alpha/photon discrimination [ 9,10 ]. Typical liquid and organic crystal scintillators used in this way, have time constants in the 1-200 ns region, with mean decay times differing by up to a factor ~ 2 for excitation by different types of particle. However, this rather small difference between the decay times means that efficient discrimination can only usually be achieved at well above energy threshold, where sufficient photons are available to form a continuous light profile. The technique has also been applied to alkali halide scintillators, including NaI(T1) [ 11 ]. However, in NaI(T1) the time constants involved are longer than is typical in organic scintillators and have fast and slow components differing by a smaller factor so that discrimination at low energies is further restricted. In this paper we propose an alternative discrimination technique for NaI (T1) called the "UVIS" detector system, based on differences in the UV and VISible parts of the spectrum of the scintillation light for different particles. This possibility arises because the mechanisms which relate the pulse shape to the rate of energy loss, dE/dr, in a scintillator for different particles can also determine the scintillation
23 September1993
emission spectrum. Thus the shape of the spectrum can be expected to be significantly different for nuclear and electron recoils. This provides the basis of an alternative possibility for achieving background discrimination in a dark matter experiment with the benefit of potentially lower threshold energy. The principle of discrimination between electrons, deuterons and alphas by wavelength filtering has been demonstrated previously using CsI(TI), but only at energies in the region of 1 MeV [ 12 ]. In assessing the prospects for using this technique in the keV region, we prefer to avoid consideration of CsI (T1) as a low background target material due to contamination by the long-lived radioactive isotope ~37Cs. The preferred alternative o f N a I (T1) also presents problems. In conventionally doped NaI ( 10 -3 T1) at room temperature the differences in spectral response are small because for all types of particle the emission is dominated by a single peak at 420 nm. Both the fast and slow scintillation decays occur at this wavelength. However, measurements in the 1960s by Van Sciver [ l 3 ] showed that by using NaI (T1) with much lower T1 doping (typically 10 - 6 TI) and operating at temperatures of ~ 100 K a second peak is present at 300 nm, well separated from the 420 nm peak. These 300 nm photons are associated with the intrinsic emission found in pure undoped NaI. Measurements were made with both incident MeV alphas and incident MeV gammas, and the ratio of the intensities of the 420 nm and 300 nm peaks was found to be strongly dependent on both panicle type and temperature. The largest difference between alpha and gamma emission spectra was found at temperatures in the region 100-120 K, the results having the simplified form illustrated in fig. 1. The integrated intensities in the two peaks were in the approximate ratio 1:1 for alphas and 4:1 for gammas. Assuming the intensities are approximately linear with energy, that the intensity ratios can be applied as true probabilities on a photon by photon counting basis, and that the response to alpha particles is broadly equivalent to the response to recoiling Na and I ions expected from WIMP interactions, then we can expect the intensity ratios to apply in the keV energy range appropriate for dark matter detection. An estimate can then be made of the degree of background rejection that might be possible by filtering and separately counting photons from the two peaks. 431
Volume 314, number 3,4
PHYSICS LETTERSB
PHOTONS
ALPHAS
NaI
A
300
420
nm
300
421)
300
420
nm
300
420
(b) Nal pure
nm
Fig. 1. Form of the emission spectra due to alpha particles and gamma ray photons interacting in (a) NaI( l0 -6 Tl) at ~ 100 K and (b) NaI(pure) at ~ 100 K, adapted from Van Sciver [ 13].
edge filter edge filler -x 100Ke n c l o s u r e /
(a)
(b)
300nm 42Ohm
300nm 420am
Fig. 2. Possible wavelength-discriminating Nal systems using: (a) 2 photomultipliers and (b) 4 photomultipliers. Details of light guides and shielding not shown.
We have carried out Monte Carlo simulations for typical systems consisting of a sodium iodide crystal with two or more windows linked to photomultipliers via light guides. Figs. 2a and 2b show schematic designs with 2 and 4 photomultipliers. Filters select either the UV or visible component for each photomultiplier. In practice one has the choice between absorption filters, which simply absorb that component of the light not transmitted, and edge interference filters, which transmit one component and reflect the other back into the system. Another option 432
23 September1993
would be to omit the filters and distinguish the UV and visible components by photon timing, using the large time constant difference discussed below. For purposes of the simulation this choice does not have to be m a d e - one can simply assume a value for the overall light collection efficiency. For computational purposes we take this to be 20%, multiplied by a photomultiplier efficiency of ~ 20% to give an overall photon detection efficiency of 4%, divided equally between the two wavelength channels. However, these illustrative assumptions introduce no loss of generality, since our results will be expressed in terms of the number of photons detected, and can thus be applied to systems of higher or lower detection efficiency. To simulate the discrimination process, visible and UV photons are produced at random, in the proportions 0.8/0.2 for electron recoil events and 0.5/0.5 for nuclear recoil events. Each has the specified detection probability in each channel and the fraction of UV counts is recorded as a function of the absolute number of photons detected. A discrimination level rd for this ratio is chosen (e.g. ra=0.6) below which the events are assumed to be nuclear recoils. However, statistical fluctuations will result in some fraction of the background events also falling below the discrimination level. Denoting the fraction of nuclear recoil events falling below ra byf~, and the fraction of misidentified background byfb, the ratio fb/fn provides a measure of the potential background-reducing capability of the technique. The background reduction factor improves rapidly with increasing number of photons, but can also be improved, for a given number of photons, by setting the discrimination level lower. This reduces the misidentified events but at the expense of losing more of the true events. Fig. 3 shows the results for a simple two-channel system with the discrimination level ro set at 0.6, 0.5, 0.4 and 0.3. Note that the reduction in collection efficiency f , is in each case compensated by the greater improvement in fb, but at the expense of a longer running time. The initial rise in both curves is due to the inclusion of the additional constraint that a coincidence between at least one photon in each photomultiplier is required to define an event. The loss of low energy events and discriminating power due to this constraint can in principle be improved by using two photomultipliers for each wavelength,
V o l u m e 314, n u m b e r 3,4
IO0[
'
'
traction of [ events
'
¢a)
PHYSICS LETTERS B
"
'
~
'
'
'
(b) __
'
'
'
'
'
'
"
'
•
•
'
(c)
alO-~
lO -4 1
0
.l
.
{
4
,
,
,
l
•
•
•
8
'
12
"~
•
•
'
16
. TM
20
photons Fig. 3. Results for the b a c k g r o u n d - r e d u c i n g c a p a b i l i t y of a s i m p l e 2-channel N a l s p e c t r u m d i s c r i m i n a t i n g system with the discrimi n a t i o n level ra set at ( a ) 0.6, ( b ) 0.5, (c) 0.4 a n d ( d ) 0.3.
as illustrated in fig. 2b, since an event can then also be registered by a coincidence between two photons of the same wavelength. As an alternative to the arrangement in fig. 2b one might consider a rectangular crystal with four windows. Even with the simple arrangement of fig. 2a, however, it can be seen that at least one order of magnitude reduction in background could be achieved at a threshold of only 2 photons in each channel. With the collection efficiency assumed, and a light emission in NaI (10 -6 T1) of > 40 photons/keV at the reduced temperature [ 13 ], this would correspond to an observed scintillation threshold < 3 keV. As stated above, the calculations assume the scintillation response for alphas is equivalent to the response for recoiling I and Na that would result from WlMP interactions. Combining the result with measurements of the actual scintillation response of recoiling I and Na atoms in NaI relative to alphas at low energy obtained from neutron scattering experiments [14], leads to a higher expected recoil threshold due to WIMPs (typically < I0 keV f o r N a and <30 keV for I). However, these measurements, together with measurements of the scintillation efficiency of NaI (T1) for various heavy ions, also suggest that there
23 S e p t e m b e r 1993
may be a larger difference between the UV/visible ratios for heavy ions relative to electrons than for alphas relative to electrons [ 14,15 ]. This would tend to improve the discrimination and hence lower the threshold again, so that the discrimination of nuclear recoils for typical dark matter targets (i.e. A>~ 19) may in practice be better than the above simulations based on alpha/electron data. There will be additional losses associated with the filters and their angle-dependent transmission characteristics (typical efficiency probably 40%-70% depending on the type of absorption or interference filters used) which would tend to increase the scintillation threshold. However, there is also scope for reducing this by improved light collection schemes. Fig. 4 shows results for a simulation of a 7 5 × 7 5 × 7 5 mm crystal in which electron and nuclear recoil events are shown on a scatter plot in terms of the numbers of 300 nm and 420 nm photons collected in the two channels of a typical experiment, calibrated with respect to electron equivalent energy (keVee). The discrimination ratio is shown as a line separating the two classes of events resulting from nuclear recoil and background electron recoils. Because of the expected shape of a dark matter nuclear recoil spectrum, there is a rather rapid decrease in sensitivity with increasing energy threshold, making it normally the case that the best dark matter lim-
50 energy in 300 nm channel 40 (keVee)
n e u t r o n s
30
• ?7. : " : .?.. - L. :
• ? iiii ?
0
i i ¢ ~ . "
0
10
•
,"
.
.
,
20
30
40
50
60
energy in 420 am channel (keV~) Fig. 4. Results o f a s i m u l a t i o n of electron a n d nuclear recoil disc r i m i n a t i o n in a 75 × 75 X 75 m m NaI crystal. 300 n m and 420 n m p h o t o n s collected in a typical two channels e x p e r i m e n t are shown, calibrated with respect to electron e q u i v a l e n t energy ( keVoe ).
433
Volume 314, number 3,4
PHYSICS LETTERS B
its are set at the detector energy threshold [ 2 ]. However, it can he seen from fig. 3 ( a n d the simulation o f fig. 4) that a rapid gain in background discrimination results as one moves away from threshold to higher energies and larger numbers o f collected photons. F u r t h e r m o r e the absolute background decreases with increasing energy. It is quite probable, therefore, that better limits will be set at energies somewhat above threshold. Because o f the several factors involved, it will require further analysis to establish the o p t i m u m operating point, and the net gain achievable, in specific experiments but a two orders o f magnitude i m p r o v e m e n t seems within reach with this technique. A further gain m a y he possible by measuring the time structure of the emitted photons. The time constants for NaI (T1), at various T1 doping levels and temperatures, have been reported by a n u m b e r o f authors though the data remains incomplete and somewhat confused [ 9,1 l, 13,15,16 ]. The UV photons are characteristic o f the crystal lattice and have a single d o m i n a n t constant varying only slowly with temperature being ~ 30 ns at 100 K. However, the 420 nm photons, enhanced by the T1 doping, involve several processes with t e m p e r a t u r e - d e p e n d e n t time constants ranging from the ns to the ms range. These appear in different proportions for different particle types, due to the differences in ionization density (in relation to the relatively large spacing o f the Tl a t o m s in the lattice) and to the different efficiencies for ionization and exciton mechanisms. In N a I ( 1 0 .3 T1) (usual doping level) at room temperature, the differences in the 420 nm emission band are fairly small: alphas give a fast rise followed by a 230 ns decay, while photons give a 60 ns rise, 150 ns plateau, and 230 ns decay [ 11 ]. However, suppression o f some processes at reduced temperature leads to longer time constants in the 420 nm band, with the p r e d o m i n a n t c o m p o n e n t s likely to extend into the las range [ 16 ], and with the possibility o f larger time constant differences within the 420 nm emission. This prospect o f increased differences between the various particle d e p e n d e n t time constants in the 420 nm b a n d may allow further d i s c r i m i n a t i o n between particle types by photon timing analysis that would be a d d i t i o n a l to the photon wavelength d i s c r i m i n a t i o n technique. In this way a p r o p o r t i o n o f those background events misidentified through the wavelength ratio test could 434
23 September 1993
be recovered through a likelihood test on the p h o t o n arrival times. F r o m this preliminary study we conclude that wavelength-discriminating techniques offer the prospect o f low energy nuclear recoil discrimination in a sodium iodide target, with the advantage c o m p a r e d with discrimination techniques at m K temperatures of achieving target masses > 100 kg using relatively inexpensive technology. N.J.C.S. wishes to acknowledge support for this work under the SERC advanced fellowship scheme.
References [ 1] J.R. Primak, D. Seckel and B. Sadoulet, Annu. Rev. Nucl. Part. Sci. 38 (1988) 751. [2 ] P.F. Smith and J.D. Lewin, Phys. Rep. 187 (1990) 203. [ 3 ] See, for example, T.J. Sumner, Nucl. Phys. B 22 ( 1991 ) 165; T.J. Sumner et al., Proc. 22nd Intern. Cosmic Ray Conf. 4 (1991) 722; P.F. Smith et al., UK particle physics proposal 270 stage 2 (1991). [4] G. Davies et al., to be published. [5 ] See, for example, P. Belli et al., Proc. XXVth Rencontre de Moriond, Particle astrophysics: the early universe and cosmic structures, eds. J.M. Alimi et al. (Editions Frontieres, Gif-sur-Yvette, 1990) p. 225; Nuovo Cimento 103A (1990) 767; C. Belli et al., Proc. XXVth Rencontre de Moriond, Massive neutrinos, tests of fundamental symmetries, eds. O. Fackler et al. (Editions Frontieres, Gif-sur-Yvette, 1991 ) p. 185; L. Mosca et al., Nucl. Phys. B (Proc. Suppl.) 28A (1992) 302. [ 6 ] K. Fushimi et al., Phys. Rev. C 47 ( 1993 ) R 425. [7] A. Allesandrello et al., Nucl. Instrum. Methods B 61 ( 1991 ) 106; R.L. Brodzinski et al., NucL Instrum. Methods A 292 (1990) 337; N.J.C. Spooner et al., Proc. 21st Intern. Cosmic Ray Conf. l0 (1990) 264; P. Barnes et al., Proc. 1990 Summer Study on High energy physics, research directions for the decade, ed. E.L. Berger (World Scientific, Singapore, 1992) p. 311. [8 ] N.J.C. Spooner et al., Phys. Lett. B 273 ( 1991 ) 333; T. Shuttet al., Phys. Rev. Lett. 69 (1992) 3425, 3531. [9] See, for example, J.B. Birks, Theory and practice of scintillation counting (Pergamon, Oxford, 1964); W.H. Tait, Radiation detection (Butterworths, London, 1980); G.F. Knoll, Radiation detection and measurement (Wiley, New York, 1979); F.D. Brooks, Nucl. Instrum. Methods 162 (1979) 477.
Volume 314, number 3,4
PHYSICS LETTERS B
[ 10] See, for example, I.B. Berlman and O.J. Steingraber, Nucl. Instrum. Methods 108 (1973) 587; D.B.C.B. Syme and G.I. Crawford, Nucl. Instrum. Methods 104 (1972) 245; T.G. Miller, Nucl. Instrum. Methods 63 (1968) 121. [ 11 ] R.B. Owen, Nucleonics 17 (1959) 92; P. Belli et al., Nucl. Instrum. Methods A 294 (1990) 391; G. Gerbier, Proc. XXVth Rencontre de Moriond, New and exotic phenomena '90, eds. O. Fackler et al. (Editions Frontieres, Gif-sur-Yvette, 1990) p. 469; P. Doll et al., Nucl. Instrum. Methods A 285 (1989) 464. [ 12] G. Hrehuss, Nucl. Instrum. Methods 8 (1960) 344. [13]W. Van Sciver, IRE Trans. Nucl. Sci. NS-3 (1956) 39; Nucleonics 14 (1956) 50; Phys. Rev. 120 (1960) 1193; Phys. Lett. 9 (1964) 97. [ 14] N.J.C. Spooner et al., Phys. Lett. B, to be published.
23 September 1993
[ 15 ] See, for example, S.K. Allison and H. Casson, Phys. Rev. 90 (1953) 880; E. Newman and F.E. Steigert, Phys. Rev. 118 (1960) 1575. [ 16] See, for example, V.I. Startsev, Z.B. Baturicheva and Yu.A. Tsirlin, Opt. Spectrosc. 8 (1960) 286; N.N. Vasil'eva and Z.L. Morgenshtern, Opt. Spectrosc. 12 (1962) 41; D.E. Persek et al., IEEE Trans. Nucl. Sci. 27, no. 1 (1980) 168; P. Schotanus, R. Kamermans and P. Dorenbos, IEEE Trans. Nucl. Sci. NS-37, no. 2 (1990) 177; J.C. Robertson and J.G. Lynch, Proc. Phys. Soc. 77 ( 1961 ) 751; J.S. Schweitzer and W. Ziehl, IEEE Trans. Nucl. Sci. NS30, no. 1 (1983) 380. [ 17 ] J. Bonanomi and J. Rossel, Helv. Phys. Acta 25 ( 1952 ) 725.
435