Simultaneous Measurement of Ionization and Scintillation from Nuclear Recoils in Liquid Xenon as Target for a Dark Matter Experiment

0.32 MB·English

by  E. Aprile

Simultaneous Measurement of Ionization and Scintillation from Nuclear Recoils in
Liquid Xenon as Target for a Dark Matter Experiment
E.Aprile1,∗ C.E.Dahl4, L.DeViveiros3, R.Gaitskell3, K.L.Giboni1,
J.Kwong4, P.Majewski1, K.Ni1, T.Shutt2,† and M.Yamashita1
1 Department of Physics, Columbia University, New York, NY 10027, USA
2Department of Physics, Case Western Reserve University, Cleveland, OH 44106, USA
3Department of Physics, Brown University, Providence, RI 02912, USA and
4Department of Physics, Princeton University, Princeton, NJ 08540, USA
(Dated: February 4, 2008)
Wereportthefirstmeasurementsoftheabsoluteionizationyieldofnuclearrecoilsinliquidxenon,
asafunction ofenergy andelectric-field. Independentexperimentswerecarried out with twodual-
phase time projection chamber prototypes, developed for the XENON Dark Matter project. We
6 find that the charge yield increases with decreasing recoil energy, and exhibits only a weak field
0 dependence. Theseresultsareafirstdemonstration ofthecapability ofdualphasexenondetectors
0 to discriminate between electron and nuclear recoils, a key requirement for a sensitive dark matter
2 search at recoil energies down to 20 keV.

n
PACSnumbers: 95.30.CqElementaryparticleprocessesinAstrophysics,95.35.+dDarkMatter,25.40.Dn
a
Elasticneutronscattering,29.40.McScintillatorDetectors
J

4
2 Introduction: Current evidence indicates that one The active LXe is defined by two electrodes, a cathode

quarter of the mass-energy density of the universe is and a gate grid, at a distance of 1.9 cm (Columbia) and
1
composed of cold, non-baryonic dark matter, which has 1.0 cm (Case). An event in the active LXe produces
v
thus far been observed only through its gravitational both ionization electrons and scintillation photons. If
2
5 interactions with normal matter. Its precise nature is trapping by impurities is negligible, the electrons which
5 undetermined, but weakly interacting massive particles escapeinitialrecombinationdriftfreelyunderanapplied
1 (WIMPs) are an attractive candidate which may be de- electric field of a few kV/cm up to the liquid-gas phase
0 tectable via rare elastic scattering interactions deposit- interface where they are extracted and accelerated by
6 ing a few tens of keV in target nuclei. For a review the high (∼10 kV/cm) field applied across the gas gap.
0
of the motivation for WIMPs, and current experimental These acceleratedelectrons produce proportionalscintil-
/
h WIMPsearches,see[1]. Thebestlimits,fromCDMS,are lation photons before being collected on the anode. The
p <0.06 events/kg/day (in Ge) [2]. Improving the search fieldsarecreatedbysettingpotentialsonfourelectrodes:
-
o sensitivitywillrequirebothlargerdetectorsandlowerra- cathode, gate grid, anode, and top grid. The electrodes
r dioactive backgrounds. An attractive method to achieve are stretched wire grids, except for a solid stainless steel
t
s this usesliquidnoblegas-baseddetectors,whichpromise cathode in the Case detector and a mesh top electrode
a to be readily scalable to the multi-ton scale. For the in the Columbia detector. The wire grids are made with
:
v XENONexperiment[3],wearedevelopingalargevolume 120µmdiameterBeCuwiresona2mmpitch,exceptfor
i liquid xenon (LXe) dual-phase time-projection-chamber 40 µm gate grid in the Case detector, and are soldered
X
(TPC) that features simultaneous measurementof recoil to SS (Columbia) and Cirlex (Case) frames.
r
ionization and scintillation to determine the energy and
a Photomultiplier tubes (PMTs) detect both the direct
3-D localization on an event-by-event basis. Nuclear re-
scintillation light (S1) produced at the event site in the
coils from WIMPs (and neutrons) have denser tracks,
liquid and the proportional scintillation light (S2) pro-
andthus havebeen assumedto havegreaterelectron-ion
duced in the gas. The time difference between the two
recombinationthanelectronrecoils,providingabasisfor
signals, determines the depth of the event. The 5 cm
discrimination against radioactive background gammas
diameter, 4 cm long metal channel PMTs (Hamamatsu
and betas. However, ionization from these nuclear re-
ModelR9288)weredevelopedtobedirectlycoupledwith
coilshasnotbeenmeasureduntilnow. Inthisreport,we
LXe, and have a quantum efficiency of ∼15% at the 178
describe the firstmeasurementsofthe ionizationyieldof
nm wavelength of the Xe scintillation [4]. PTFE is used
low energy nuclear recoils in LXe and confirm the base-
as a VUV reflector enclosing the active LXe volume in
line discrimination capability of this technique.
both detectors. The Columbia detector uses two PMTs,
one in gas and one in liquid. Because of total internal
Independent measurements were carried out with two
reflection at the liquid surface (index of refraction 1.65
separate dual-phase detectors at Case and Columbia.
at 187 K [5]), the light collection is significantly better
for the PMT in the liquid. The total S1 light collection
efficiency(fractionofphotonsincidentonthephotocath-
∗Electronicaddress: [email protected] odeofaPMT)isestimatedfromMonteCarlotobemore
†Electronicaddress: [email protected] than 50%, uniform across the volume to within 5%. An
2
internal blue light LED is used to calibrate and monitor after, nuclear recoil energies calculated in this way will
the gain of the two PMTs. The Case detector uses a be denoted ’keVr’, and electron recoil energies based on
single PMT in the gas, with an S1 light collection effi- the linear S1 scale will be denoted ’keVee’. In the Case
ciency of ∼16%. The PMT gain is monitored using the and Columbia detectors the S1 calibration gives, respec-
S1 light from 5.3 MeV alphas from a source described tively, 1.5 photoelectrons (pe)/keVee and 5 pe/keVee, at
below. The operation of a dual phase LXe detector is zeroelectric field, which correspondto 0.28pe/keVr and
described in more detail in [6]. 1 pe/keVr for 55 keVr nuclear recoils.
Both detectors use a liquid nitrogen-cooledcold finger TheS2signalmeasurestheionizationfromeachevent.
cryostat with temperature control to maintain a stable Thereisnopublisheddataofchargeyieldvsfieldforlow
liquid temperature. The detectors were operated with energy gammas in LXe, so this measurement was per-
57
vapor pressures between 2.0 and 2.8 atm, with better formed with Co 122 keV gammas, in both the Case
than 1% stability. The Case detector also uses a triple- [15] and Columbia [16] detectors operated in single (liq-
parallelplatecapacitorsystemtoalignthe liquidsurface uid) phase. The results are shown in Fig. 4. The S2
with the grids and monitor the level and stability of the calibrationis a function of gas pressure and electric field
liquid surface, which was[7] stable to within 20 µm over across the gap [17], with the added complication of pos-
two months. The LXe in both detectors was purified us- sible electron multiplication near the anode wires. Typ-
ing a high temperature getter. The Columbia detector ical values were 19 pe/e− with 4.6 kV/cm (Case) and
usedthe purificationsystemwithcontinuousgascircula- 8.4 pe/e− with 2 kV/cm (Columbia).
tion through the getter, developed for the first XENON Figures1 and2showsthe detectors’responsesto neu-
prototype [8, 9]. The Case detector used a similar recir- tron and low energy Compton scattering events. The
culating system only at the start of its two-month run. logarithmoftheratioS2/S1isplottedasafunctionofnu-
External gamma ray sources, including 57Co, 133Ba, clear recoil energy (keVr). In both detectors, the elastic
and 137Cs, were used for calibration of both detectors. nuclearrecoilbandis clearlyseparatedfromthe electron
The Case detector also had an internal 210Po source de- recoil events. The events identified around 40 keVee are
posited on the center of the cathode plate, providing from gamma rays produced by neutron inelastic scatter-
5.3 MeV alpha particles and 100 keV 206Pb nuclear re- ing on 129Xe. Additional gamma ray events are emitted
coils. Events <0.5 mm above the cathode were tagged following inelastic scattering of neutrons on 131Xe and
aspossiblealphasand206Pbrecoils. Forneutrondata,a 19FcontainedinPTFE.The detectorresponsesto 137Cs
5 Ci AmBe source (Columbia) and 25 µCi 252Cf source (Columbia) and 133Ba (Case) show only the gamma ray
(Case) were used, with lead shielding to attenuate gam- band.
mas from the sources. Nuclear Recoil Ionization Yield: The ionization
The PMT signals were digitized with multiple ADCs yield is defined as the number of observed electrons per
sampling at 5 MHz-1 GHz. The effective trigger thresh- unit recoil energy (e−/keVr). The nuclear recoil ioniza-
oldforthe Casedetectorwas4.5electronsforthe S2sig- tion yield as a function of energy is shown in Fig. 3,
nal, and 50% single photoelectron (spe) acceptance for for several drift fields. The uncertainty on the yield is
S1. For the Columbiadetector the triggerwas either the dominatedby the systematic errorfromthe57Co S2 cal-
coincidence of the S1 signals from the two PMTs (at a ibration.
fewspelevel)ortheS1signalfromthePMTintheliquid Fig. 4 summarizes the relative light and charge yields
(at a ∼6 spe level). as a function of drift field for different particles in LXe:
Calibration and Raw Data: Both detectors were 122 keV gamma rays from [15, 16], 56 keV Xe nuclear
calibratedprimarilywith57Co122keVgammas,forboth recoils from [10] and this paper, and 5.5 & 5.3 MeV al-
the S1 and S2 signals. The S1 signal servesas an energy phas from [18] and the Case detector. The relative light
axis for nuclear recoils, given by Er = Ee/Leff ·Se/Sr yield S(E)/S0 is the light yield relative to that at zero
where Er is the nuclear recoilenergy, Ee is a linear elec- fieldS0. TherelativechargeyieldQ(E)/Q0 isthecharge
tronrecoilenergyscale basedonthe 57Co S1 peak, L collectedrelativetothatatinfinite field(i.e., withnore-
eff
is the effective Lindhard factor relating the scintillation combination) Q0. For nuclear recoils, Q0 = Er ·L/We,
yieldsofnuclearandelectronrecoilsatzerofield,andS where E is the recoil energy, L is the supression pre-
x r
is the loss of scintillation light due to the recombination dicted by Lindhard theory [19], and W = 15.6 eV [20]
e
suppression by the electric field for species x, equal to isthe averageenergyrequiredtoproduceanelectron-ion
S(E)/S0 in Fig. 4. Recent measurements at Columbia pair in LXe. For gammas and alpha particles, we divide
give Sr for 56.5 keV recoils and Leff at energies up to the known energy by We to obtain Q0.
56.5 keV [10]. Other experiments [11, 12, 13, 14] have Discussion: The important characteristics of the
measured Leff at higher energies. The Columbia and measured ionization yield of nuclear recoils in LXe are
other data, except [13], were parameterized by Leff = its value relative to that of other particles (Fig. 4), its
0.0984E 0.169. For simplicity, we assume S has no en- energy dependence (Fig. 3), and its field dependence
r r
ergy dependence since the Columbia measured value is (Fig. 4). The yield is a function of the Lindhard fac-
close to unity, as seen in Fig. 4. Also shown in Fig. 4 tor and the amount of recombination in the track. The
aremeasurementsofS asafunctionofdriftfield. Here- Lindhard factor has no field dependence, and does not
e
3
Vr] 6
e
k5.5
e/
eld [ 5 Columbia Case
n Yi4.5 0.27 kV/cm 0.10 kV/cm
o 2.00 kV/cm 2.03 kV/cm
ati 4
z
ni
o3.5
I
3
2.5
Sys.+Stat. Error (Columbia)
2
1.5 Sys.+Stat. Error (Case)
1
20 30 40 50 60 70 80 90 100 110 120
Energy [keVr]
FIG.3: Energydependenceofnuclearrecoilionization yield
at different drift fields.
1 alpha, light
0.9 NR, light
Q00.8
E)/ 0.7 ER, charge
Q( 0.6
S(E)/S or 0000...345 ER, light
0.2 NR, charge
0.1 alpha, charge
0
FanIGd.1317:CsCgoalummmbaiasdoeutreccetso(rbroetstpoomn)s,eatto2AkmVB/ecmnedurtirfotnfi(etldop.) E)/Q(4.5 kV/cm)0.150 N0R.5 E1R 1.5 2alpha2.5 3 3.5 4 4.5 5
Q( 00 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Drift Field (kV/cm)
FIG.4: Top: Field dependenceofscintillation andionization
yield in liquid xenon for 122 keV electron recoils (ER), 56
keVr nuclear recoils (NR) and alphas. Bottom: Ionization
yields scaled by their 4.5 kV/cm values.
affect the relative charge yield (Fig. 4). Lindhard does
predict a slight decrease in charge yield with decreasing
energy,theoppositeofwhatwesee. Thereforeweexpect
recombination and its dependence on track density and
shape to explain the above phenomena.
Recombination is primarily a function of electric field
and ionization density, with stronger recombination at
low fields and in denser tracks. Ionization density along
atrackcorrespondsroughlytoelectronicstoppingpower,
plotted in Fig. 5 for alphas, electrons, and Xe nuclei in
LXe, as given by ASTAR, ESTAR, and SRIM [21], re-
spectively . Also shown is the total energy lost to elec-
tronic excitation per path-length for Xe nuclei, which
differs from the electronic stopping power in that it in-
cludes energy lost via electronic stopping of secondary
recoils [22].
Atlowerenergies,thestoppingpowerforXenucleide-
creases, indicating lower recombination and higher ion-
FIG. 2: Case detector response to 252Cf neutron (top) and izationyield,asobserved. Thestoppingpower,including
133Ba gamma sources (bottom) at 1.0 kV/cm drift field. daughter recoils for Xe nuclei at 56 keV is higher than
that of alphas at 5.5 MeV. This is in conflict with what
we observe in the relative ionization yields, indicating
4
that only modest fields are needed for the background-
103 discriminating measurement of nuclear recoils.
n)
o
cr
mi102 Background Discrimination: The different values
V/ of S2/S1 between nuclear recoils and electron recoils
e
k
Xe (101 makeliquidxenonanexcellenttargettodiscriminatebe-
L tween these two types of interaction. Acceptance win-
n
x i dows for nuclear recoils were defined from the neutron
E/d100 ele. rec., ESTAR elastic recoil band as shown in Fig. 1 and 2. The re-
d alphas, ASTAR
nuc. rec., SRIM jection of background electron recoils was measured by
10-1 nuc. rec., Hitachi irradiating both detectors with pure gamma ray sources
100 101 102 103 104 (133Ba or 137Cs), which produce low energy electron re-
Energy (keV)
coils from Compton scatters, and recording the number
FIG. 5: Predicted electronic stopping power, dE/dx, for dif- of electron recoils outside of 50% acceptance nuclear re-
ferent particles in LXe. The circles indicate the particle en- coil windows. The Columbia detector, with higher S1
ergies discussed in thetext. light collection, achieved an electron recoil rejection effi-
ciency of 98.5% independent of energy down to 20 keVr
at2 kV/cmdrift field. This efficiency appearsto be lim-
that the higher ionization yield of a low energy Xe re- ited by charge loss at the edge of the detector where the
coil compared to an alpha is due to the different track field was non-uniform (no field shaping rings were used
geometriesforthesetwoparticles. Recombinationforal- in this prototype). The rejection efficiency in the Case
phasis enhancedbecauseoftheir cylindricaltrackshape detectorimprovedwithincreasingenergy(becauseofthe
with a very dense core and a ”penumbra” of delta rays. non-optimal light collection) and was better than 99.5%
SRIM simulations show that nuclear recoil tracks have above80keVr,presumablyduetoabetterfieldgeometry
many branches, with most energy lost by secondary re- with closer electrodes and better S2 resolution. The re-
coils. Each of these secondary branches presumably end jection efficiency improves with higher drift field (in the
inaverysparsetracksince the stoppingpowerfalls with Casedetector,e.g.,from91%at0.2kV/cmto97%at4.6
energy. This may result in a low-density halo of ioniza- kV/cm for 35 keVr). With an optimized field configura-
tion around the track, which does not recombine even tion, and with the XY position sensitivity of a LXeTPC
at very low fields. If this lack of recombination extends as proposed for the XENON experiment [3], the prob-
to zero field, it would explain the difference between the lem of edge events should be negligible. The dual phase
measured scintillation yield for nuclear recoils and the xenontechniquethusappearsverypromisingforalarge-
Lindhard prediction. The importance of the track ge- scale detector with powerful electron recoil background
ometriesis alsoseenin the striking contrastbetweenthe discriminationfora sensitiveWIMP darkmatter search.
field dependence of nuclear recoils and those of electrons
and alphas (bottom graph of Fig. 4), though electrons The authors wouldlike to thank A. Hitachifor insight
have a much lower, and alphas a comparable stopping and discussion on particle interaction in LXe. TS would
power to nuclear recoils. Our new finding of the weak like to thank K. McDonald and C. Lu for advice and
field dependence for nuclear recoils in LXe has the im- support. This work was funded by NSF grants PHY-03-
portant practical consequence for dark matter detection 02646 and PHY-04-00596.
[1] R.J.Gaitskell, Ann.Rev.Nucl.Part.Sci.54,315(2004). (2000).
[2] CDMS Collaboration, D. S. Akerib et al., (2005), [13] R. Bernabei et al., Phys. Lett. B389, 757 (1996).
astro-ph/0509259. [14] V. Chepel et al., (2005), physics/0512136.
[3] E. Aprile et al., NSFProposal 0400596 (Sep 2003), [15] T. Shuttet al., To be published.
NewAstronomy Reviews, 49 (2005)289 [16] E. Aprile et al., To be published .
[4] J. Jortner et al., J. Chem. Phys.42, 4250 (1965). [17] A. I. Bolozdynya, Nucl. Instrum. Meth. A422, 314
[5] V. N. Solovov et al., Nucl. Instrum. Meth. A516, 462 (1999).
(2004). [18] E. Aprile, R.Mukherjee, and M. Suzuki, Nucl. Instrum.
[6] E.Aprileet al., IEEETrans.Nucl.Sci.51, 1986 (2004). Meth. A307, 119 (1991).
[7] T. Shuttet al., To be published. [19] J. Lindhard et al., Mat. Fys.Medd.Dan. Vid.Selsk. 33
[8] E.Aprileetal.,Nucl.Phys.Proc.Suppl.138,156(2005). (1963).
[9] XENON Collaboration, E. Aprile, (2005), [20] T. Takahashi et al., Phys.Rev. A12, 1771 (1975).
astro-ph/0502279. [21] http://physics.nist.gov/PhysRefData/Star/Text/contents.html
[10] E. Aprile et al., Phys. Rev.D72, 072006 (2005). http://www.srim.org/
[11] D.Akimov et al., Phys. Lett.B524, 245 (2002). [22] A. Hitachi, To besubmitted to Phys. Rev.
[12] F. Arneodo et al., Nucl. Instrum. Meth. A449, 147