Suppression of alpha-induced lateral surface events in the COBRA experiment using CdZnTe detectors with an instrumented guard-ring electrode

0.55 MB·

by  J. -H. Arling

PreparedforsubmissiontoJINST
Suppression of alpha-induced lateral surface events in
the COBRA experiment using CdZnTe detectors with
an instrumented guard-ring electrode
7
1
0
2

n
a
J The COBRA collaboration

5
J.-H.Arlinga,b M.Gerhardta C.Gößlinga D.Gehrec R.Klingenberga K.Kröningera
2
C.Nitscha T.Quantea K.Rohatschc J.Tebrüggea,1 R.Temminghoffa R.Theinerta

t] S.Zatschlerc K.Zuberc
e
d aTechnischeUniversitätDortmund,LehrstuhlfürExperimentellePhysikIV,Otto-Hahn-Str.4a,44221Dort-
-
s mund
n
bnowatDeutschesElektronen-SynchrotronDESY,Hamburg
i
.
s cTechnischeUniversitätDresden,InstitutfürKern-undTeilchenphysik,ZellescherWeg19,01069Dresden
c
i
s E-mail: [email protected]
y
h
p Abstract: The COBRA collaboration searches for neutrinoless double beta-decay (0νββ-decay)
[
using CdZnTesemiconductor detectors withacoplanar-grid readout andasurrounding guard-ring

1
structure. The operation of the COBRA demonstrator at the Gran Sasso underground laboratory
v
2 (LNGS)indicates thatalpha-induced lateralsurface eventsarethedominantsourceofbackground
3
events. Byinstrumentingtheguard-ringelectrodeitispossibletosuppressthistypeofbackground.
4
7 In laboratory measurements this method achieved a suppression factor of alpha-induced lateral
0 surfaceeventsof5300+2660,whileretaining(85.3±0.1)%ofgammaeventsoccurringintheentire
. −1380
1
detector volume. This suppression is superior to the pulse-shape analysis methods used so far in
0
7 COBRAbythreeordersofmagnitude.
1
:
v Keywords: CZT,CdZnTe,semiconductordetector,coplanar-grid,guardring,backgroundsuppres-
i
X sion,surface events
r
a
1Correspondingauthor
Contents
1 Introduction 1
2 GuardringinstrumentationofCdZnTedetectors 2
3 Charge-clouddynamicsofalphaparticles 3
4 Predictionsfromanelectric-field calculation 4
5 Experimentalsetup 6
6 Measurementsandresults 7
7 Summaryandoutlook 9
1 Introduction
TheCOBRA(CdZnTe0NeutrinoDoubleBetaResearchApparatus)collaboration[1]searchesfor
neutrinoless double beta-decay (0νββ-decay) [2]. This decay is forbidden in the Standard Model
ofparticlephysics. Thequestforlepton-number violationisamainmotivationtosearchfor0νββ-
decay. Furthermore, the detection of this process could give information about several general
properties of neutrinos like the neutrino mass scale and mass hierarchy. The decay has not been
measured yet, limits on the half-life are of the order of 1025yr, depending on the nuclide under
study[3–5]. Specialexperimentaltechniquesarerequiredtomeasuresuchraredecays. Onecrucial
issue is the reduction of background events which can mimic the searched-for decay, coining the
term’low-background experiment’.
TheCOBRAcollaboration operates ademonstrator setupattheGranSassounderground lab-
oratory (Italy),technical detailscanbefoundinRef.[6]. Thedemonstrator consists of64CdZnTe
coplanar-grid (CPG) semiconductor detectors with the volume of 1cm3 each. CdZnTe contains
nine nuclides that canundergo double beta-decays: 64Zn,70Zn,106Cd,108Cd,114Cd,116Cd,120Te,
128Te and 130Te. Due to its high Q-value of 2814keV [7], 116Cd is the most promising candidate
fortheCOBRAexperiment. Nosignificantexcessovertheestimatedbackgroundwasfoundwhich
resultsinhalf-lifelimitsofabout1021yrfortheground-state toground-state 0νβ−β− transitions of
70Zn, 114Cd, 116Cd, 128Te and 130Te as documented in Ref. [8]. One conclusion drawn from the
operation of the demonstrator is that the main background component stems from alpha-induced
lateral surface events. In the current scheme, lateral surface events are identified by analyzing the
recorded pulseshapesinthedetector; detailsofthismethodcanbefoundinRef.[9].
All detectors used by COBRAfeature a guard ring, which is a boundary electrode surround-
ing the CPG anodes. It is a common method to improve the detector performance, as it leads to
a better balanced weighting potential [10] and to a reduction of leakage currents. In the current
–1–
configuration, the guard ring is not instrumented and left on a floating potential. Setting a de-
fined potential on a guard-ring structure to suppress surface events was studied in Ref. [11] using
Germanium detectors. First measurements have shown that this novel method can be used for the
COBRA detectors as well without deteriorating the detector performance significantly [12]. This
paper discusses the instrumentation of guard rings for large 6cm3 detectors used for the upgrade
oftheCOBRAdemonstrator.
2 Guardring instrumentation ofCdZnTe detectors
InCdZnTe,theproduct oflifetimeandmobilityforelectrons andholesisdifferentbythreeorders
of magnitude, introducing a strong interaction-depth dependence of the detectors response. To
compensate for this effect, the so-called coplanar-grid technology [13] is used in COBRA. It is a
single-polaritycharge-carriersensingmethod,whereinthiscaseonlytheelectronsignalisreadout.
Inthistechnique,twoanodesarecomb-shapedinterleavedandaresetonaslightlydifferentelectric
potential, called grid bias (GB). One anode is set to ground potential and referred to as collecting
anode(CA)asitcollectsthegeneratedelectrons. Theotherissetonanegativepotentialoftypically
−40Vandisreferredtoasnon-collectinganode(NCA).Abiasvoltage(BV)oftypically−1200V
is applied to the cathode. The detectors discussed in this publication are (20 × 20 × 15)mm3 in
sizewithavolumeof6cm3 andamassof36g. Theirelectrodes areconfigured according toaso-
called coplanar quad-grid (CPqG), i.e. four individual CPG structures, rotated against each other
◦
by 90 . The whole CPqG structure is surrounded by a common guard ring (GR) which is left on
a floating potential in a default configuration. Such adetector, but without aguard ring electrode,
was characterized extensively in Ref. [14]. Details of the electrode design and its dimensions are
depicted inFigure1.
Figure1. Scheme of the electrodeconfigurationof a
CPqG detector. The four sectors comprise individ-
ualCPGs,eachinstrumentingacollectinganode(CA,
red) and a non-collecting anode (NCA, green). The
surrounding guard ring (GR) is shown in blue. All
distancesare given in mm. The purpleboxindicates
thepositionofthedetailedviewshowninFigure2.
For the instrumentation of the guard ring a defined potential is set to it, in this case ground
potential, which is the same potential as the CA. The guard ring can thus collect charges drifting
to it. This is particularly important for charge clouds originating from interactions in the vicinity
of the lateral detector surfaces. The anodes and the guard ring are connected to charge sensitive
preamplifiers (Cremat CR1101). These convert the induced charge signals coming from interac-
tions in the detector volume into voltage signals. The signals are afterwards amplified in linear
1http://www.cremat.com/CR-110.pdf
–2–
amplifiers and digitized in flash analog-to-digital converters (FADC SIS33002) with a sampling
rate of 100MHz. The data contain the full pulse-shape information of the measured channels, so
itispossible toapplypulse-shape analysis toolsintheoff-lineanalysis [9]. Moredetailsaboutthe
data-acquisition canbefoundinRef.[6].
ThedepositedenergyinaCPGdetectorcanbereconstructedbycalculatingtheamplitudeAof
the difference signal between the CA and NCAsignal amplitudes including aso-called weighting
factorw,
E ∝ A −w·A . (2.1)
CA NCA
Theweighting factorcompensates foreffectsofelectron trapping [13]andcanbedetermined dur-
ingthecalibrationprocess. TheenergydepositedineachCPqGsectorcanbereconstructedindivid-
uallyusingEquation2.1. Thesignaloftheguardringisnotconsideredintheenergyreconstruction
inthisstudy.
3 Charge-cloud dynamics ofalpha particles
The penetration depth of alpha particles from radioactive decays in solids like CdZnTe is very
short,about20µmforanalphaparticlewithakineticenergyof5MeV. Therefore,eventsinduced
bysourcesofalpharadiationoutsidethedetectorvolumedeposittheirenergyinthevicinityofthe
surfaces and are referred to as lateral surface events. Due to the deposited energy, electron-hole
pairs are created at the interaction point. The effect of holes is ignored here because their drift
towards the cathode is much slower. The charge cloud of the generated electrons expands while
drifting through the detector towards the electrodes due to two main reasons: mutual repulsion
of the electrons as well as thermal diffusion. The maximal values for these two quantities can be
calculated by considering near-cathode events in the detector. In this case, the created electron
clouddriftsthroughthewholedetector volume.
Thespreadσ ofthechargecloudduetodiffusion[15]canbecalculatedasafunctionofthe
diff
driftlength xas
2k Tx
σ (x) = B , (3.1)
diff r eE
where k is the Boltzmann constant, T the absolute temperature, e the electric elementary charge
B
and E the electric field strength. At room temperature and with an applied BV of −1200V the
resulting spreadduetothermaldiffusionforamaximaldriftlengthof15mmis
σmax = σ (15mm)≈ 100µm. (3.2)
diff diff
The expansion R due to mutual repulsion is defined as the largest diameter of the charge cloud
rep
[16]. Itcanbecalculated asfunction ofthedriftlengthto
3eNx
R (x) = 3 , (3.3)
rep r4πǫ ǫ E
0 r
with the number of created charge carriers N and the permittivity of free space and the relative
permittivity, ǫ and ǫ = 10.9[17],respectively. Theenergy ofthealpha particles from the241Am
0 r
2http://www.struck.de/sis3300.htm
–3–
source isabout 5.5MeV,whichisalso typical foralphaparticles from natural decay chains. Inan
interaction of this energy, about 106 charge carriers are produced. Therefore, the resulting spread
duetorepulsion isatmost
Rmax = R (15mm) ≈ 420µm. (3.4)
rep rep
Thequadraticsumofthesetwoeffectscanbecalculatedtoestimateanupperlimitonthemaximal
spread L ofachargecloud[18,19]assumingamaximaldriftlengthof15mmfornear-cathode
max
events,
2 2
L = σmax + Rmax ≈ 430µm. (3.5)
max r diff rep
(cid:16) (cid:17) (cid:16) (cid:17)
Anaspect nottaken into account inthis estimation isthat thecharge cloud iscreated directly next
tothedetectoredgeandthereforecannot expandequally inalldirections.
The effect of the charge cloud expansion of alpha-induced lateral surface events on the elec-
trodes isshowninFigure2. Theinitial interaction hasadistance ofabout20µmfromthesurface.
Theexpansion ofthe charge cloud isshownafter adrift length of 1mm,3mm, 6mm, 10mmand
15mm. Evenfor the largest drift length of 15mm, the outermost CPGanode (CA)isnot affected
by the charge cloud. Hence, a clear separation of alpha-induced lateral surface events and events
occurring intheinnerdetector volumeshouldbepossible byinstrumenting theguardring.
Figure2. Detailedtopviewoftheanodeside
asmarkedinFigure1. Thedifferenttypesof
magentaindicatetheexpansionofthecharge
cloud after a drift length of 1mm, 3mm,
6mm,10mmand15mm. Thedarkdothigh-
lightstheinitialchargecloud.
4 Predictions from anelectric-field calculation
Tostudytheeffectofdifferentguard ringpotentials, e.g.floatingpotential inthedefault setting or
ground potential forenabling charge collection, anelectric-field simulation isperformed usingthe
simulation toolCOMSOLMultiphysics inversion5.2[20].
ThecurrentCPqGdetectordesignisimplementedintermsofdimensionsandelectrodedesign,
thetypical material-specific values forCdZnTearetaken from Ref.[21]. Theapplied biasvoltage
isassumed tobe−1200Vandatypical value of−40Vischosen fortheGB.Configurations with
different guard ring potential are simulated. In addition, the configuration of the CPqG anodes is
varied: ifaCAoranNCAbiasisappliedtotheoutermostanode,theconfigurationisreferredtoas
CA and NCA mode, respectively. The simulated electric field-line distribution for the CA
out out out
mode with the guard ring on ground potential is shown in Figure3. Here, field lines start on the
cathode in equidistant steps of 25µm and they end on the different electrodes (GR, CA and NCA
–4–
Figure3. Electric field-line distribution shown
inacrosssectionclosetothedetectoredge.The
guard ring is set on groundpotential. The cho-
sen GBandBV are−40V and−1200V. Field
linesendingontheGRarecoloredinpink,those
endingontheCPqGanodesinblue.
as highlighted). Field lines ending on the guard ring are colored in pink in contrast to field lines
ending oneitheroftheCPqGanodes, whichareshowninblue.
The innermost field line ending on the guard ring starts at a distance of 625µm from the
detector edge. Themaximaldistance depends onthedepth ofthestarting position, andaslopefor
the first 2mm below the anode side is observed before the field lines are almost parallel. Energy
deposited between the detector surface and the innermost field line ending on the guard ring will
not be detected by the CA. This will lead to a decrease in the overall detection efficiency, but it
will in particular suppress a large fraction of alpha-induced lateral surface events, according to
Equation3.5. The area influenced by the instrumented guard ring is visualized in Figure4. This
Figure4. Contour plot of charge collec-
tion of the guard ring and the CPqG an-
odes in CA mode as seen from the top.
out
Shown are the projected start positions of
field linesfinally endingon the guardring
(pink) or on either of the CPqG anodes
(blue). The chosen GB is −40V and BV
is−1200V.
contourplotdepictsthestartpositionsofelectricfieldlinesanddifferentiatesbetweenendpositions
on the guard ring electrode or the CPqG anodes. The dips in the area influenced by the CPqG at
themiddleofthesidesarisefromthetransitionsbetweenthedifferentCPGsectors. Atthebottom,
thedipislargerbecauseoftheguardringcontactpad. Theratiooftheareainfluencedbytheguard
ring relative to the full area is a measure of reduction of the fiducial detector volume, ǫ . The
fid
results fortheCA mode,asdepicted inFigure4,andfortheNCA modeare
out out
ǫCAout = 87.7% and ǫNCAout = 86.0%, (4.1)
fid fid
respectively. The uncertainties in these calculations are negligible as the mesh for the numerical
calculation isfine(25µm),andsmallvariations ofthebiasesdoalsonotshowanotableeffect.
–5–
Only the CA mode is evaluated in the following because of the expected larger efficiency than
out
in NCA mode. In addition, other beneficial effects like less charge-sharing between the sectors
out
favorthismode.
5 Experimental setup
The effect of instrumenting the guard ring in terms of a suppression of alpha interactions from
the lateral surfaces is studied in measurements. A prototype CPqG detector with guard ring, pro-
duced by Redlen Technologies3, is used. During characterization of this test detector, the energy
resolution in terms of full-width at half maximum (FWHM) for an incident energy of 662keV
of the different CPG sectors is measured. Two of the four sectors have only poor energy resolu-
tionsof5.0%and7.0%. Anothersectorissufferingofleakagecurrentsdegradingitsperformance
immensely, while not influencing the other sectors. Finally, the last remaining sector provides a
satisfactory energy resolution of3.0%. Thissectoristhereforechosen asthesectorfortesting.
A241Amsourceemitting alphaparticles withanenergy of5.5MeVisusedtoprovide asam-
ple of alpha-induced surface interactions. The source is pointed directly on the sector under test.
Furthermore,thedetectorcanbeirradiatedwithgammaradiationfroma232Thsourcewithvarious
gamma lines, the highest at an energy of 2614keV. In this case the whole detector is irradiated
almosthomogeneously.
The detector is mounted on a movable stage, so that it is possible to irradiate the detector at
first with gamma radiation and afterwards with the alpha source without switching off the biases
when exchanging the sources. Hence, the recorded gamma energy spectrum can be used for a
common calibration for each configuration. The measurement setup as well as the tested sectors
aredepicted inFigure5.
Figure 5. Top view of the measurementsetup for evaluating the effect of the guard-ringinstrumentation.
Thedetectorismountedonamovablestageallowingtoshiftitbetweenthetwodifferentradioactivesources
(α:241Amsource,γ:232Thsource).Thealphasourceispointedtothesectorundertest,whereasthegamma
sourceirradiatesthewholedetector. Theyellowbrickssymbolizeleadshielding.
To account for background, a dedicated measurement of the background spectrum is per-
formed. The detector is kept in the same position as for an alpha measurement, but the 241Am
source is removed. Only background events originating from the natural laboratory background,
cosmicmuonsorscattered photons fromthe232Thgammasourcearerecorded.
3http://redlen.ca/
–6–
Toevaluatetheeffectoftheinstrumentedguardring,onemeasurementisperformedwiththeguard
ringongroundpotential andonewiththeguardringkeptonafloatingpotential. Theappliedvolt-
agesforthismeasurementcampaignareaGBof−40VandaBVof−1200V.
6 Measurements and results
The energy spectra of the sector under test irradiated with 232Th gamma radiation are shown in
Figure6. Thereconstructedenergyspectrumforthecaseofafloatingguardringisdepictedinblue
and for thecase ofaninstrumented guard ring inred. Thesharp drop around 280keV stemsfrom
the threshold for the energy reconstruction needed for event-by-event triggering. The spectrum
shows the typical features of a 232Th source. The total event rate in the case of an instrumented
guard ringisreduced compared tothefloatingcaseduetothereduced efficiency. Thisisexpected
fromtheresultsoftheelectric-field calculation insection 4.
V)]) floating GR
e
k
2 instrumented GR
⋅(s10−1
1/( COBRA
e [
at
unt r10−2
o
c
10−3
Figure6. 232Th energy spectra measured with the
10−4 sector under test. Shown are the cases of floating
0 500 1000 1500 2000 2500 3000 (blue)andinstrumentedguardring(red).
energy [keV]
Forthealphameasurement,theresultingenergyspectraforbothguardringconfigurationsare
showninFigure7. Theunusualspectralshapeofthealphaspectrumforthefloatingguardringcan
beexplainedbytheattenuationduetoairbetweenthesourceanddetector,andthedetectorsurface
encapsulation, especially as the source is not collimated and effects of the solid angle occur. This
results in a smeared distribution shifted towards lower energies. The spectrum of alpha-induced
lateral surface events in the instrumented guard ring configuration is clearly reduced compared to
thefloatingpotential case.
V)]) floating GR
e
k
2 instrumented GR
⋅(s10−1
1/( COBRA
e [
at
unt r10−2
o
c
10−3
Figure7. 241Am energy spectra measured with the
10−4 sector under test. Shown are the cases of floating
0 500 1000 1500 2000 2500 3000 (blue)andinstrumentedguardring(red).
energy [keV]
–7–
The spectra resulting from dedicated measurements of the laboratory background are shown
in Figure8. One can see similar spectra for both configurations of the guard ring again. This
is expected because most background events stem from gamma radiation and muons. A large
fractionoftheeventsinthespectrumfromthealphameasurementwithaninstrumentedguardring
(red curve in Figure7) are caused by laboratory background, which needs to be considered when
estimating thesuppression ofalpha-induced events.
V)]) floating GR
e
k
2 instrumented GR
⋅(s10−1
1/( COBRA
e [
at
unt r10−2
o
c
10−3
Figure8. Backgroundenergyspectrameasuredwith
10−4 thesectorundertest. Shownarethecasesoffloating
0 500 1000 1500 2000 2500 3000 (blue)andinstrumentedguardring(red).
energy [keV]
This is evaluated quantitatively in Table1 for all six measurements by comparing the count rates,
integrated between 280keV and 3000keV. The quoted uncertainties are due to the statistical
Poissonerror.
Table1.Resultsofmeasurementsforsectorundertest. Givenaretheintegralcountratesr(betweentrigger
thresholdandchosenhighenergycut-off)ineachmeasurementandforbothguardringconfigurations.
alphameasurement background measurement gammameasurement
r [1/s] 208.6±0.2 0.493±0.008 42.33±0.05
floatingGR
r [1/s] 0.498±0.008 0.465±0.008 36.15±0.04
instrumentedGR
The resulting reduction of the fiducial volume for gamma radiation in the case of an instru-
mentedguardringisestimatedas
rγ −rbkg
ǫ = instrumentedGR instrumentedGR. (6.1)
γ
rγ −rbkg
floatingGR floatingGR
UsingthenumbersinTable1,thisyields
ǫ = (85.3±0.1)%. (6.2)
γ
This value is comparable with the prediction of ǫ = 87.7% from the calculation of the electric
fid
field(Equation4.1). Thedifferencesarisefromeffectsofarealdetectorthatarenotincorporatedin
theelectric-fieldcalculation,e.g.inhomogeneitiesinthedetectorbulk,uncertaintiesintheelectrode
placements ortheenergy resolution.
–8–
Thesuppression factorforalpha-induced lateralsurfaceeventsSF canbedefinedas
α
rα −rbkg
floatingGR floatingGR
SF = . (6.3)
α
rα −rbkg
instrumentedGR instrumentedGR
Astheratesintheenumeratorandthedenominator differbyaboutfourordersofmagnitude,com-
mon Gaussian uncertainty propagation cannot be used forthe calculation oftheratio. Instead, the
suppressionfactorisestimatedbasedonatoyMonte-Carlostudy: 107pairsofrandomnumbersare
drawnfromtwoGaussiandistributions, characterized bytheenumerator andthedenominator, and
the resulting frequency distribution of the ratios of these numbers is interpreted as the probability
density of SF . Because this asymmetric distribution features a strong non-Gaussian tail towards
α
larger values, the suppression factor is estimated by the mode of the distribution. The smallest
interval containing 68.3%ofallentriesisinterpreted astheuncertainty interval. Thisyields
SF = 5300+2660. (6.4)
α −1380
Thisisalargeimprovementcomparedtothepulse-shapeanalysiscurrentlyusedtosuppresslateral
surface eventsinCOBRA[8,9]: therelative detection efficiency forgammaradiation isabout the
same, but the suppression factor for alpha-induced lateral surface events is nearly three orders of
magnitude larger.
7 Summary and outlook
TheCOBRAcollaboration searches forneutrinoless double beta-decay and hence background re-
duction is a crucial issue. Alpha-induced lateral surface events are the main background source
when operating the COBRA demonstrator. In this paper, dedicated laboratory measurements and
calculations of electric fields are used to demonstrate that instrumenting the guard ring of CPqG
detectors leadstoasuppression factor foralpha-induced lateralsurface events of5300+2660,while
−1380
thereductionoffiducialvolumeforgammaeventsoccurringthroughouttheentiredetectorvolume
is(85.3±0.1)%.
ThisconceptwillbefollowedintheupgradeoftheCOBRAextendeddemonstrator (XDEM).Un-
der optimal conditions and the assumption that the background is dominated by alpha events, the
overallbackground rateisexpected tobeloweredbytwoordersofmagnitude.
Thesignalsoftheguardringaremeasured,butnotincludedintheanalysispresentedhere. Amodi-
fiedeventreconstructionincludingtheguard-ringsignalcaninprinciplebeperformed. Preliminary
studies indicate thatusingthisadditional information isfeasible andcanimprovethedetector per-
formance (e.g. the detection efficiency) and the veto capabilities for lateral surface events even
further.
Acknowledgments
Wethank the LNGSforthecontinuous support ofthe COBRAexperiment. COBRAissupported
by the German Research Foundation DFG (ZU 123/15-1 / GO 1133/3-1). Furthermore, we thank
COMSOL for support with the COMSOL Multiphysics simulation and O. Schulz for technical
support withtheDAQ.
–9–