Krypton and radon background in the PandaX-I dark matter experiment

0.77 MB·

by  Shaoli Li

PreprinttypesetinJINSTstyle-HYPERVERSION
Krypton and radon background in the PandaX-I dark
matter experiment
7
1
0
2
Shaoli Lia, Xun Chena, Xiangyi Cuia, Changbo Fua, Xiangdong Jia,b,c,d, Qing Lina∗,
n
a Jianglai Liua†, Xiang Liua,b, Andi Tanc, Xuming Wanga, Mengjiao Xiaoc,b, Pengwei
J
Xiea
5
2 aINPACandDepartmentofPhysicsandAstronomy,ShanghaiJiaoTongUniversity,

ShanghaiKeyLaboratoryforParticlePhysicsandCosmology,Shanghai,200240,China
]
M bCenterofHighEnergyPhysics,PekingUniversity,Beijing,100080,China
cDepartmentofPhysics,UniversityofMaryland,CollegePark,MD,20742,USA
I
h. dT.D.LeeInstitute,Shanghai,200240,China
p
-
o
r
t
s ABSTRACT:Wediscussanin-situevaluationofthe85Kr,222Rn,and220RnbackgroundinPandaX-
a
[ I, a 120-kg liquid xenon dark matter direct detection experiment. Combining with a simulation,

1 their contributions to the low energy electron-recoil background in the dark matter search region
v areobtained.
7
0
3 KEYWORDS: Krypton;Radon;Background;DarkMatter;Xenon;PandaX.
7
0
.
1
0
7
1
:
v
i
X
r
a
∗NowatDepartmentofPhysics,ColumbiaUniversity
†Correspondingauthor,[email protected].
Contents
1. Introduction 1
2. Kryptonbackground 2
3. 222Rn 4
4. 220Rn 8
5. Singleα studies 11
6. Summary 12
7. Acknowledgement 13
1. Introduction
Radioactive rare gases, most notably 85Kr, 222Rn and 220Rn, are important background in xenon-
baseddarkmatterexperiments. Theycannotberemovedbyconventionalhotgettersviachemical
reactionsandhavehighmobilitytodiffusethroughoutthetargetvolume,whichdegradethepower
of fiducialization. The air contains unstable 85Kr (10.8 year half-life, β-decay) whose activity
is ∼1 Bq/m3 produced by the fissions of uranium and plutonium in our nuclear age [1]. If air
is leaked into the detector, 85Kr would produce β-decay background in the detector. Radon gas
is produced by the decay of the long-lived 238U and 232Th, which could be introduced into the
detectoreitherexternallybyanairleakorinternallybysurfaceemanationfromdetectormaterials.
Directmeasurementofsuchbackgroundduringdatatakingiscritical.
PandaX-I[2]wasadarkmattersearchexperimentoperatedattheChinaJinPingunderground
laboratory(CJPL)[3,4,5]usingadual-phasetimeprojectionchamber(TPC)witha120-kgactive
liquidxenontarget. TheapparatushasbeendescribedindetailinRef.[2],soonlythekeyaspects
relevant to this paper are presented here. A particle interacting in liquid xenon (LXe) produces
prompt scintillation photons (S1). Some ionized electrons will drift under a drift field defined by
the 60 cm diameter cathode grid (−15 kV) and the gate grid (−5 kV) located at the bottom and
top of the sensitive liquid region, separated by 15 cm. The electrons will be extracted into the
gaseous xenon by a stronger field across the liquid surface between the gate grid and the anode
mesh (ground), separated by 8 mm, and produce proportional electroluminescense photons (S2).
The S1 and S2 are collected by the photomultiplier (PMT) arrays located at the top and bottom
of the TPC. When in combination with the time separation between S1 and S2, they allow three-
dimensional reconstruction of the event vertex. The ratio between the S2 and S1 for the nuclear
recoil(NR)signalproducedbydarkmatterscatteringwithxenonnucleusissignificantlylowerthan
–1–
that for electron recoil (ER) background produced by the γs or βs, enabling effective background
rejection. α events, characteristic for radon decays, can be identified by their discrete high value
ofenergydepositionbutanevenlessS2-to-S1ratioincomparisontotheNRevents.
The results for dark matter search from PandaX-I have been published in Refs. [6, 7]. In
thispaper,wepresentadetailedin-situevaluationofthekryptonandradonbackgroundinthefull
exposure(54×80.1kg-day)darkmattersearchdatainPandaX-I.SameasRef.[7],wehaveapplied
thewidelyuseddelayedcoincidencetotagthesebackgroundinthedata, asinRefs.[8,9,10,11,
12]. Inthispaper,weupdatedthecorrespondingresultsfromRef.[7]duetothefollowingupdates.
First,theβ andγ energycutsfor85KrwerechangedslightlyincomparisontoRef.[7],andinthis
paperweusedall80-daydatatosearchfor85Krdelay-coincidence(Ref.[7]onlyusedtheblinded
63-day data). Second, we have updated the signal finding efficiency from the Monte Carlo (MC)
simulation, which has a more realistic treatment for close-by delay-coincidence signals. Third,
we have made accidental background subtraction in all delayed coincidence analysis. The rest
of this paper is organized as follows: Secs. 2, 3 and 4 are dedicated to 85Kr, 222Rn, and 220Rn
background,respectively,usingcharacteristicdelayedcoincidencealongtheirdecaychains;Sec.5
isanalternativeradonstudybasedonsingleαs;andSec.6containsthesummary.
2. Kryptonbackground
85Kr β-decays with 99.566% of the probability directly into stable 85Rb with a Q-value of 687
keV, and 0.434% probability into the meta-stable 85mRb with the maximum β energy of 173 keV.
85mRbthende-exciteswithahalf-lifeof1.015 µs,emittingaγ rayof514keV.The85Krlevelcan
thereforebeestimatedusingtheβ-γ delayedcoincidencesignature.
Asearchontheβ andγ coincidenceswasperformedineachevent. AsdescribedinRef.[13],
foreachPMTchannel,thewaveformdataina200µswindowwererecorded. Toavoidambiguity
in the S1 and S2 pairing, we selected events in which both the β and γ made a single scatter. For
thedelayedcoincidenceeventsfrom85Kr,about60%ofthemproducedtwoclean(S1,S2)signals.
TherestcontainedtwoS1sbutonlyasingleS2asthetimeseparationbetweenthetwoS2signals
couldalsobelessthanthewidthofatypicalS2signal(about2µs).
The selection cuts were applied as follows. The energy values (estimated by S1) of the β
and γ were required to be within 20 to 200 keV, and 300 to 700 keV, respectively, and the time
separation between the two S1s was required to be within 0.3 and 3 µs. An example coincidence
waveformisshowninFig.1. ForeventswithtwoS1sandS2s,wealsorequiredthatbothβ andγ
wereconsistentwithERevents(withinalooseERcutobtainedfromtheγ sourcecalibrationdata,
cf. Fig.4(a)). Notethattomaximizethestatistics,allcandidateeventswithinthe120kgsensitive
targetvolumewereselected.
The cut efficiency was obtained using the PandaX-I MC simulation package [14], which was
developedbasedonGeant4[15](v9.6)withrecommendedphysicslistsforlowenergyphysicsand
included the full detector geometry of PandaX-I. The delay time cut required that both the β and
γ haddepositedenergyinsidethedetector,andthetimeseparationbetweenthemwasbetween0.3
to 3 µs. The efficiency was estimated to be 56.2%. The energy cut efficiencies on β and γ were
88.8%and72.9%,respectively,thereforetheoverallcutefficiencywas36.4%.
–2–
0
S1
b
S2
-500 S1 b
E g
P
e/
d
u -1000
plit
m
A
-1500
S2
g
-2000
9000 10000 11000 12000 13000 14000
Sample/10 ns
Figure1. Typicalwaveformofaβ-γ delayedcoincidencesignalfromkryptondecayinwhichtwoS2sare
overlapped.
Afterapplyingthecutsabove,16eventsfromthedatasurvivedintheentire120-kgsensitive
volume. The delay time and the square distance (∆L2 =∆x2+∆y2+∆z2) between the β and γ
areshowninFigs.2(a)and2(b),bothexhibitingcorrelationdespitelimitedstatistics. Theposition
distribution of the βs is shown in Fig. 2(c). A number of candidate events happened close to the
gas/liquid xenon interface. This could be due to the fact that the boiling point of krypton (∼120
K) is lower than that of the xenon (∼165 K).To be conservative, we assumed that krypton was
distributed uniformly in the detector. The event selection above contained accidental background
arising from the random coincidence of single β-like and γ-like events which satisfied the selec-
tion cuts above. Based on the dark matter data, we estimated the β-like and γ-like event rates to
be0.24Hzand0.4Hzrespectively. Sotherandomcoincidenceeventratewithinthesamecoinci-
dencewindow(0.3−3µs)is2.3×10−2evts/dayor1.8eventsin80.1-dayexposure,withnegligible
statisticaluncertainty.
Basedonthedelayedcoincidencecandidatesandaccidentalbackground,thekryptonlevelin
xenoncanbeestimatedby
N −N
data acc
N =N /f, N = (2.1)
Kr 85 85
ε ·BR·(T/τ)
cut
where N and N are the numbers of 85Kr and total krypton atoms, respectively, related by the
85 Kr
85Krabundance f (about2×10−11, seeRef.[1]). N andN arethenumberofrawcandidates
data acc
andaccidentaleventsinthedata,respectively,ε istotalcutefficiency,BRisthebranchingratio
cut
which is 0.434%, T =80.1-day is the live-time of the dark matter run, and τ is the mean lifetime
of 85Kr (15.52 years). Using Eq. 2.1, number of 85Kr atoms is estimated to be (6.4±1.8)×105,
leadingtoamolarconcentrationof58±16partpertrillion(ppt)kryptonatomsinxenon.
TheMCsimulationwasusedtotranslatethe85Krdecayratetothebackgroundrateobserved
inthedarkmattersearchregionbetween0.5−5keV,leadingtoa2.0±0.6mDRU(mDRU=10−3
evt/day/kg/keV) background. No apparent time dependence was observed in the β-γ coincidence
candidates, nor did the low energy background in the data exhibit the dependence. Therefore, the
krypton was likely introduced during the detector filling period when the detector was underpres-
surized,notbyaleakdevelopedduringtherun.
–3–
5 2
4.5 1.8
4 1.6
3.5 1.4
3 1.2
s s
unt2.5 unt 1
o o
C 2 C0.8
1.5 0.6
1 0.4
0.5 0.2
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 20 40 60 80 100120 140 160180 200
Delay time (m s) D L2 (cm2)
(a) Delaytime (b) ∆L2
0
-2
-4
-6
m)
z (c -8
-10
-12
-14
0 100 200 300 400 500 600 700 800 900
r2 (cm2)
(c) Position
Figure 2. Distributions in a) Delay time, b) ∆L2, and c) z vs. r2 for 85Kr β-γ candidates. Reconstructed
position based on the S1 signal is used in b) and c) if there is only one S2 signal tagged in the delayed
coincidence. Inc),theβ positionsareshown.
3. 222Rn
222Rn is a decay progeny of 238U, which is either external airborne or internal due to surface em-
anation of radioimpurity in the internal detector components. 222Rn has a half-life of 3.82 day,
with its decay chain illustrated in Fig. 3. Along the chain, number of β-decays could contribute
to the background in the low energy signal region. Given that the PandaX-I running period was
muchlongerthanthedecayhalf-lifeof222Rnprogeniesallthewayto210Pb,wehaveassumedthat
secular equilibrium was achieved upstream of 210Pb, based on which we can estimate the 222Rn
level.
Totag222Rnevents,weusedthe214Bi-214Poβ-α delayedcoincidenceevents,inwhich214Bi
emitsβ witha maximumenergyof 3.272MeV,followed bytheα decayof214Powith ahalf-life
of164.3µsandanα energyof7.69MeV.
As mentioned earlier, the PandaX-I data acquisition window was 200 µs [13]. The delayed
coincidence signal could either be recorded in one or two adjacent events, which will be referred
–4–
Rn-222 decay chain
222Rn
86
3.82 d
28114308P do 28114644P uso Ebma=x3.27 MeV 28143.81P moin Ea=5.49 MeV
210Bi 214Bi a
83 83
206Pb Ea=5.30 5.01 d 210Pb Ea=7.69 19.9 min 214Pb Ea=6.00 b
82 Me 82 Me 82 Me
stable V 22.2 y V 26.8 min V
Figure3. 222Rndecaychain. Theredarrowindicatestheβ-α delayedcoincidenceusedinthisanalysis.
to as the “BiPo1E" and “BiPo2E" hereafter. The β energy (scaled from S1) and α energy (see
later)cutsof0.1−3.5MeVand6.0−9.5MeVwereapplied,respectively. Thetimeseparationcut
betweenthetwoS1swassettobe20−100µsforBiPo1Eor300−1000µsforBiPo2E.Inaddition,
β signalshouldbeconsistentwithanERsignal(Fig.4(a)).
With these selection cuts, we first searched for BiPo2E from the data. Fig. 4 shows the dis-
tributions of β and α in the plane of log (S2/S1) vs. S1. Two clusters in the α distribution can
10
be clearly observed in Fig. 4(b), labeled as A and B in the figure. The delay time and distance
8 2
2.6 30
7
2.4 25
6 1.5
1) 2.2 5 1) 20
S S
(S2/log101.82 34 (S2/log10 1 A 1105
2 0.5
1.6 B
5
1
1.4
0 0 0
0 1000 2000 3000 4000 5000 6000 7000 8000 20000 30000 40000 50000
S1 (PE) S1 (PE)
(a) 214Bi (b) 214Po
Figure4. Thedistributionoflog (S2/S1)vs. S1of214Bi(a)and214Po(b)ofBiPo2Eevents. Redcurves
10
in(a)aretheERcutdefinedbytheERcalibration.
betweenβ andα inclustersAandBareshowninFig.5,witheventpairsinbothclustersexhibit
propertimingandspatialcorrelations, thereforecorrespondingtogenuineβ-α coincidences. The
position distribution of events in clusters A and B are shown in Fig. 5(c). Events in A are mostly
locatedclosetothecathode,whileeventsinBareuniformlydistributedintheentiredetector. This
phenomenonwasfirstdiscussedinRef.[16]foraliquidxenonTPCandexplainedby214Bi+ ions
driftingtowardandgettingattachedontothecathode,amodelwhichweshallrefertoasthe“ion-
drift model”. Since the electric field near the cathode wires was much stronger than the average
drift field and due to the short range of αs, cluster A is displaced from cluster B in Fig. 4(b). To
reconstruct both clusters to the 214Po α energy, we derived an effective α energy reconstruction
–5–
100 deltaT/1.e3 {deltaT/1.e3<1000&&qS1[alpha]>2.68e4-3*2.14e3&&qS1[alpha]<2.68e4+3*2.14e3&&qS2[s2max]>4.65e5-3*7.97e4&&qS2[s2max]<4.65e5+3*7.97e4&&deltaT/1.e3>=300} EMRnMetarSine s h1 51 2165095..891 (xTopNN_pre[s2max_pre]-xTopNN[s2max])^2/100.+(yTopNN_pre[s2max_pre]-yTopNN[s2max])^2/100.+(((tS2_pre[s2max_pre]-tS1_pre[beta])-(tS2[s2max]-tS1[alpha]))/1.e3*0.17)^2 {deltaT/1.e3<1000&&qS1[alpha]>2.68e4-3*2.14e3&&qS1[alpha]<2.68e4+3*2.14e3&&qS2[s2max]>4.65e5-3*7.97e4&&qS2[s2max]<4.65e5+3*7.97e4&&deltaT/1.e3>=300}EMREMRnnMMeettaarrSSiinnee ss hh11 1212 22....717155080855858511
Cluster A, t 242.3 – 7.5 m s
Cluster A
Cluster B, t 240.1 – 14.1 m s
80 Cluster B
102
60
Counts40 Counts10
20
1
0
200 300 400 500 600 700 800 900 1000 0 1 2 3 4 5 6 7 8 9 10
Delay time (m s) D L2 (cm2)
(a) Delaytime (b) ∆L2
(tS1[alpha]-tS2[s2max])/1000.*0.17:pow(xTopNN[alpha]/10.,2)+pow(yTopNN[alpha]/10.,2) {deltaT/1.e3<1000&&qS1[alpha]>2.68e4-3*2.14e3&&qS1[alpha]<2.68e4+3*2.14e3&&qS2[s2max]>4.65e5-3*7.97e4&&qS2[s2max]<4.65e5+3*7.97e4&&deltaT/1.e3>=300}
0
-2
-4
-6
m)
z (c -8
-10
-12
-14
0 100 200 300 400 500 600 700 800 900
r2 (cm2)
(c) αposition
Figure5. Distributionsof214Bi-214Podelayedcoincidenceeventsa)indelaytime, b)in∆L2, andc)inα
positionforclustersA(bluedot)andB(redopensquare),clusterAdistributedatthetopandbottommainly
whileclusterBdistributeduniformlyinbetween.
functionas
S1(PE) S2(PE)
E (keV)= + (3.1)
α
7.19PE/keV 116.88PE/keV
Thereconstructedβ andα energydistributionsareshowninFig.6. ForBiPo1Eevent,werequired
thattherewasatleastoneS2signalandtheenergycutswerethesamewiththeabove. Similaras
thoseinBiPo2E,α candidateswereobservedtoformtwoclustersasinFig.4(b)withsimilarpo-
sitiondistributions. NotimedependencewasobservedfromBiPo1EandBiPo2Erates. Giventhat
theundergroundradonlevelvariedfromtenstofewhundredBq/m3 duringtherun,weconcluded
thatthe222Rnbackgroundwasnotduetoanexternalairleak,butratherduetotheinternalsurface
emanation.
Combining all the BiPo2E and BiPo1E candidate events, distributions of delay time in the
54kgfiducialvolume(FV)isshowninFig.7andisfittedwithasingleexponentialfunction. The
datainthefirsttwobinsappearhighercomparedtothefit,implyingaslightcorrelatedbackground
for small delay time but contributing only ∼3% to the total delayed coincidence rate above the
–6–
Graph
10 10 18
9.5 9.5 16
9 9 14
8.5 8.5
12
eV) 8 8 10
M7.5 7.5
( 8
Ea 7 7
6
6.5 6.5
6 6 4
5.5 5.5 2
5 5 0
0 100 200 0 0.5 1 1.5 2 2.5 3 3.5 4
150
100
50
0
0 0.5 1 1.5 2 2.5 3 3.5 4
E (MeV)
b
Figure6. Combinedenergyof214Poα vsβ energydistribution.
h2
200
180 N 144.6 – 4.9
0
160
t 250.3 – 21.0
140
120
s
nt
u100
o
C
80
60
40
20
0
0 100 200 300 400 500 600 700 800 900 1000
Delay time D t (m s)
Figure7. CombinedcandidateeventsdelaytimedistributionsintheFV,wherethegapbetween100to300
µsisduetothedelaytimecutsforBiPo1EandBiPo2E.Theuncertaintyofthefittedlifetimeincludesboth
statisticalandsystematicuncertainties.
expected exponential curve. The fitted decay time constant agreed well with the expectation. The
totalradonratecanalsobeestimatedbyintegratingthefitfunction.
The222RndecayratebasedonBiPo1E,BiPo2E,orthecombinedfitmethodsaresummarized
inTable1. ThesignalselectionefficiencieswereestimatedbytheMCsimulation. Thedelaytime
–7–
Table1. 222RnlevelcalculatedfromBiPo1E,BiPo2EandcombinedfitinFV.
Method BiPo1E BiPo2E Combinedfit
Delaytimecutacceptance 26.3% 26.4% 100.0%
β energycut 98.2% 98.2% 98.2%
α energycut 100.0% 100.0% 100.0%
Branchingratio 99.98% 99.98% 99.98%
0.66mBq 0.79mBq 0.68±0.13mBq
222Rnlevel
12.3 µBq/kg 14.7µBq/kg 12.5±2.4 µBq/kg
cutacceptancesforBiPo1EandBiPo2Ewere26.3%and26.4%,whereasthatforthecombinedfit
is100%duetotheintegrationrangefromzerotoinfinity. Theaccidentalbackgroundwasestimated
similarly as in Sec. 2 and confirmed to be negligible. The mean 222Rn level was obtained using
the combined fit, and the uncertainty was estimated based on the largest difference among three
methods. This result is at a similar level as in the XENON100 and LUX experiments [8, 9, 10,
11, 12], and such an internal background will pose challenge to next generation of liquid xenon
experiment.
As mentioned earlier, β-decays in the 222Rn chain contribute to the low energy background.
To properly estimate the off-equilibrium β-decay contribution downstream of 210Pb, in the MC
simulation, 222Rn events were assumed to be produced uniformly in position in the liquid xenon
(sowerealltheprogenies)andinadurationsameastheentireperiodoftheexperiment,andwere
letdecayalltheway. Eventsthatpassedallselectioncutsandfellintothedarkmatterdatataking
periodwerecountedasbackground. Basedonthe222RnlevelintheFV,themeanERbackground
contributiontothePandaX-Iexperimentforeachβ-decayprogenyissummarizedinTable2.
Table2. Backgroundcontributionfrom222Rn.
Isotope Background(mDRU)
214Pb 0.17
214Bi 0.002
210Pb 0.11
210Bi 0.03
218Po 0.002
Total 0.32±0.06
4. 220Rn
220Rn is the decay progeny of 232Th with a decay half-life of 55 s. The decay chain of 220Rn is
illustrated in Fig. 8. There are two delayed coincidences which can be used as clean tags, 212Bi-
212Poβ-α and220Rn-216Poα-α delayedcoincidences. 212Bihasa64.06%probabilitytoβ-decay,
withamaximumenergyofβ of2.25MeV.Thedaughter212Podecayswithahalf-lifeof0.3µsinto
a stable element 208Pb emitting an α particle of 8.78 MeV. The half-life of 216Po is 0.145 second
–8–
Rn-220 decay chain
220Rn
86
55 s
281402.3P usoEbma=x2.25 MeV 281406.1P45 os Ea=6.29 MeV
212Bi
28028Pb Ea=8.78 MeV 8631 min 28122Pb Ea=6.78 MeV a
stable 28301.18 mTinl Ea=6.06 MeV 10.6 h b
Figure8. 220Rndecaychain. Theredarrowsindicatetheβ-α andα-α delayedcoincidencesusedinthis
analysis.
andtheα energyis6.29MeVand6.78MeVfor220Rnand216Po,respectively. Thetwoα decays
wererecordedindifferentevents.
Similar to the selection of BiPo1E events in 222Rn, the event selection energy cuts for 212Bi-
212Poweresetintherangefrom100keVto3MeVforβsand>3MeVforαs,andthedelaytime
cut was set between 0.3 to 1.0 µs. Similar to 214Po in Fig. 4(b), distribution of selected 212Po α
events also has two separated clusters, corresponding to the cathode and bulk αs. The α energy
wasreconstructedusingEq.3.1, displayedinFig.9. Duetotheshortdecayof212Po, themeanα
energyislargerthantheexpected8.78MeVduetosomepileupofS2sfromtheβ andα.
50
40
30
s
nt
u
o
C
20
10
0
0 2 4 6 8 10 12 14 16
Energy (MeV)
Figure9. Distributionofthereconstructedenergyforthe212Poαs.
Toselectthe220Rn-216Podelayedcoincidence,theα energycutswerebothfrom5MeVto12
MeV,andthedelaytimecutisfrom0.05to1s. Werequiredthateachwaveformshouldhaveonly
oneS2signal. Thedistributionsofenergy,positionandtimingdifferencebetweentheselectedtwo
αs are shown in Figs. 10 and 11. The delay time distribution agrees with the expectation and
–9–