Calibration of the STAR Forward Time Projection Chamber with Krypton-83m

13 Pages·0.17 MB·English

by  V. Eckardt

Calibration of the STAR Forward Time Projection
Chamber with Krypton-83m
V. Eckardt, T. Eggert, H. Fessler, H. Hu¨mmler, G. Lo Curto,
M. Oldenburg, N. Schmitz, A. Schu¨ttauf, J. Seyboth, P. Seyboth,
M. Vidal
1
0
0 Max-Planck-Institut fu¨rPhysik,Fo¨hringer Ring6,80805Mu¨nchen, Germany
2

n
a
J
Abstract
1
3
The principles of the calibration of a time projection chamber with radioactive Krypton-

2 83 are explained. The calculation of gain correction factors and the methods of obtaining
v a precise energy calibration are illustrated. The properties and advantages of 83mKr are
3
1 summarized and compared to other radioactive calibration sources. It was shown that the
0 Krypton calibration for the STAR FTPC is feasible and recommendable although the pad
1
geometryoftheFTPCcausesaconsiderabledeteriorationofthemeasuredspectrumdueto
0
1 onlypartially detected chargeclusters.
0
/
x
e
-
l The STAR experiment (Solenoidal Tracker at RHIC) is one of four experiments
c
u operating at the Relativistic Heavy Ion Collider (RHIC) in Brookhaven. STAR
n searches for hadronic signatures of the quark gluon plasma formation and inves-
:
v tigates the behavior of strongly interacting matter at high energy density in colli-
i
X sionsofheavynuclei[1,2,3].Whilethetrackinginthecentralregionisprovidedby
r a large TPC, the two Forward Time Projection Chambers (FTPCs) are required to
a
giveposition,charge, and momentuminformationof particle tracks in theforward
rapidityregionsoftheSTARexperiment[4,5].
The calibration method using Krypton-83m has been developed by the ALEPH
collaboration[6]andisbeingusedbytheDELPHIexperiment[7].Itisalsoapplied
very successfullybytheNA49 experimentwheretheachievedmeasured accuracy
(i.e. therelativeerrorofthemean value∆E/E)isbetterthan 0.5% [8].
1 The Calibration
A calibrationofa TimeProjection Chamber(TPC)is performedfortwo mainpur-
poses: The gain correction, which is needed to compensate variationsin the front-
end electronics (FEE) and in the gas gain, and the energy calibration, which is
performedinordertodeterminetheamountofelectricchargethatismeasuredata
knownenergy depositioninthechamber.
1.1 GainCorrection
Thegainofthepre-amplifierelectronicsvariesfromchiptochip(seealsosection4)
and within a chip from channel to channel. Moreover, variations in gas gain may
occurduetoinhomogenetiesoftheelectricfieldandgeometricalnon-uniformities.
A straightforward and effortless relative gain correction for the pre-amplifier elec-
tronics can be perfomed by a so-called pulser calibration. This is done by inject-
ing calibrated charge pulses into the anode or gating grid wires of the multi-wire
proportional chambers (readout chambers). Each pulse induces a charge on each
cathodepad i, resultingin acharge signalq which is proportionaltothechannel’s
i
response. Multiplicative calibration constants c are then obtained by dividing the
i
averagecharge perpad
N
1
q¯= Xqi (1)
N
i=1
(N = numberofpads)by thecharge inducedon an individualpad:
q¯
c = (2)
i
q
i
Eachmeasuredchargeperpadqmeas (i.e.integratedADCchannels)thengetsmul-
i
tipliedbythiscalibrationconstantinorderto obtainthecalibratedcharge:
cal meas
q = c q (3)
i i i
However,pulsingtheanodewiresinducesasignalonallpadssimultaneously,pro-
ducingalargecurrentloadoneachFEEchannelwhichmayresultinamodification
of the electronics response. Thus, the pulser calibration can only be considered as
afirst step inthecalibrationprocedure.
1.2 EnergyCalibration
It is not possible to perform an absolute energy calibration by a pulser calibration
only.Therefore, an alternativecalibrationmethodusingaradioactivesourcehasto
beconsidered.Althoughparticleidentificationvia dE/dxmeasurementisdifficult
2
with the STAR FTPC [4, Sect.6.6], an energy calibration is useful to optimize the
accuracy of the position measurement. It can easily be done with a radioactive
sourcesuchas 83mKrwhichisinjected intothechambergas.
Anelectronemittedduringthedecayprocessof83mKr(seesection3)producessec-
ondaryelectronsbyionizationinthegas,wherethenumberofproducedelectronsis
proportionaltotheinitialdecay energy(i.e. thekineticenergyoftheprimaryelec-
tron). Since the decay energies of suitable sources are relatively low (E ≪ m c2),
e
the primary electron is stopped very quickly, and the secondary electrons emerge
from virtually one point. These secondary electrons are then drifted by an electric
driftfieldtothereadoutchamberswheretheyareamplifiedattheanodewires,and
thechargesignalfrom thegas amplificationis detected atthecathodepads.
Thedecayenergiesoftheradioactivesourceareknown,andthusthecorresponding
energiescanbeassignedtothepeaksofthemeasuredchargespectra.Startingfrom
adistinctreferencepeak,theenergythatcorrespondstoameasuredchargeqisthen
calculated by
Eref
E(q) = q (4)
qref
whereEref istheenergyandqref isthemeasuredchargeofthereferencepeak.How-
ever, because the highest decay energy of 83mKr is relativelyhigh compared to the
energiesofafewkeVdepositedbyaminimumionizingparticle,theanodevoltage
U has to be reduced from the value suitable for minimum ionization. The results
from the energy calibration must then be extrapolated to higher gain voltages. As
thedependence oftheamplificationon theanode voltagehas been measured, such
an extrapolationisaccurate.
Further applications of the Krypton calibration are a cross-check of the pulsercal-
ibration and an investigationofgas flow effects in thechamber by studyingthelo-
cation of Krypton decays at thebeginningof theinjection.Moreover, a calibration
is needed to check the detector linearity over a wide energy range, and to monitor
thelongtermdetectorstability.
2 The STAR ForwardTPC
The high track density in the forward rapidity regions of the STAR experiment
covered by the FTPCs (pseudorapidity range of 2.5 < |η| < 4) requires a special
design for the tracking detectors in these regions. In contrast to conventional time
projection chambers, the two Forward TPCs of the STAR detector use radial drift
fields, and the curved readout chambers are part of the outer cylinder walls. The
compactness of the FTPCs (60cm diameter and 120cm length) with only 22 cm
3
Fig. 1. Two electron clusters produced
z byKrypton decays andarowofcathode
pads. Asthepadrows are not adjacent to
r each other, for some of the clusters the
chargecanonlybepartiallydetected[9].
drift lengthandtheuseofacool gasprovidesaverygood positionresolution(150
µm) and two-track separation (1 mm) [9,10]. In order to minimize the number of
readout channels without compromising the detector performance the surface is
covered only partially by cathode pads. There are 10 padrows per FTPC, with 960
padseach.Thelengthofonepadis20mm(inz direction,wherez pointsalongthe
beam pipe), and the pad pitch is 1.9mm (in azimuthal direction φ). The distance
betweenthepadrowsisintermittently65mmand85mm.Forfurtherdetailsonthe
STARFTPC, see[4,5].
As the padrows are not adjacent to each other, a problem arises when localized
chargeclustersfromradioactivedecaysaretobedetected.Asillustratedinfigure1,
only a fraction of the cluster charge can be detected by the pads when the cluster
is located at the border of a padrow. The question whether a calibration with a
radioactivesourcecanneverthelessbedonetriggeredasystematicinvestigationon
thepracticabilityofacalibrationwithradioactiveKrypton.
Fig. 2. Decay scheme of
83 83
Rb→ Kr. The ground state
83
of Rb decays predominantly
(76%) to the isomeric excited
state 83mKr at E = 41.6keV.
This most important level of
83
Kr for calibration is popu-
lated through the 571keV and
562keV intermediate levels. It
decays entirely to the 9.4keV
state, which then decays to the
83
Krground state.
4
3 Properties ofKrypton-83
83Krisastableisotopeproducedfrom83Rb whichdecaysbyelectroncapturewith
a mean lifetime of 124 days. The decay scheme [11] is shown in figure 2. The
ground state of 83Kr is not reached directly in the 83Rb decay but via nuclear tran-
sitions from intermediate excited 83Kr levels. The relevant level for drift chamber
calibration is the isomeric (metastable) 41.6keV state 83mKr. It is fed down from
two higher 83Kr levels in 76% of all 83Rb decays and has a lifetime long enough
(τ = 2.64h) for the 83mKr to be introduced and distributed in the drift chamber
before it decays. The isomeric state 83mKr decays entirely by an E3 nuclear de-
excitation transition to the 9.4keV level with a transition energy Etr of 32.2keV.
This short-lived state then decays immediately (τ = 212ns) to the 83Kr ground
state with Etr = 9.4keV. In contrast to other common calibration sources like 55Fe
or 57Co, almost no low energy photons are produced since both transitions from
the 41.6keV and the 9.4keV states proceed almost entirely by internal conversion
(IC) with ratios of electron-to-photon emission of e−/γ ≈ 2·103 and e−/γ ≈ 20
respectively.
Thekineticenergy Ee ofan ICelectron isgivenby
Ee = Etr −Eb (5)
where Eb is the atomic binding energy of the electron. The subsequent atomic de-
excitation of the electron hole produces either X-rays or Auger electrons with an
energy equal toEb.
For the 41.6keV-to-9.4keV transition the internal γ conversion occurs predomi-
nantlyonanelectroninanouter(L, M,N)shell,thetotalelectronemissionproba-
bilityfromsuchshellsbeing77%[11].Thebindingenergiesfortheoutershellsare
small (Eb ≤ 1.9 keV). Furthermore, the de-excitationof an outer-shell hole leads
in mostcases to the emissionof an Auger electron rather than an X-ray. Therefore
practically the full energy Etr of 32.2keV is carried away by electrons and can be
collected. If, however, the internal conversion occurs in the K shell (23% of the
cases) an IC electron with an energy Ee of 17.8, 18.1, 19.5 or 19.6keV is pro-
duced,correspondingtofourKsub-levelswithbindingenergiesof14.3,14.1,12.7
and 12.6keV [11], respectively, according to eq. (5). In addition, the de-excitation
of the K-shell hole yields X-rays or Auger electrons with an energy equal to the
bindingenergy.
For the 9.4keV-to-ground state transition the low transition energy allows only
internal conversion on an outer-shell electron, so that practically the full transition
energy of 9.4keV is carried away by the IC electron (95% of the cases) or by the
escaping γ (5%).
Formoredetailson thereleased energies and onthebranchingfractions, see[11].
5
s1200
e
ri
t
n1000
E
800
600
400
Fig. 3. Simulated decay
200 spectrum of 83mKr with
an assumed line width of
0 σ/E =6%. Noise is ne-
0 10 20 30 40 50 60
E (keV) glected.
The overall decay spectrum has been simulated, using the branching fractions in
table 4 of [11] and assuminga resolution of σ/E = 6%, where σ is the width ofa
Gaussian.Thesimulatedspectrumisshowninfigure3;itiscomposedoffiveparts:
• The strong peak at 41.6keV is due to the electrons from the dominant modes of
the two-step decay of the 41.6keV isomericstate via the 9.4keV state. Because
of the short lifetimeof the 9.4keV state the two decay energies of 32.2keV and
9.4keV are deposited in the majority of cases at the same location such that the
totaldecay energy iscollected ina singlecluster.
• The smaller peak around 30keV comes from two contributions: (a) electrons
with Ee = 32.2keV from the 41.6keV-to-9.4keV transition when the 9.4keV γ
escapes; (b) K-shell IC electrons (Ee ≈ 19.5keV) together with IC electrons
fromthedecayofthe9.4keVstateatthesamelocation,leadingtoatotalenergy
depositionof∼29keV.
• Theverysmallaccumulationaround20keVisduetoK-shellICelectrons(Ee ≈
19.5keV)whenthe9.4keV γ escapes.
• The peak at 12.7keV is produced by the conversion of K-shell X-rays in the
chambergas,awayfrom thelocationoftheirorigin.
• The peak at 9.4keV results from either the conversion of the escaped 9.4 keV γ
inthegasorthe9.4keVelectron energy,separated from thepreceding 32.2keV
transition.
Figure4showsamoredetailedsimulation,wherepartiallydetectedclusterscaused
bythepadgeometry,diffusion,thedetectorresolution,and cutcriteriacomparable
to those in the experiment were considered. The 41.6keV peak is clearly visible
whiletheseveralpeaksaround30keVaswellasthe9.4keVandthe12.7keVpeaks
arehardlyseparable.Moreover,theshapeofthelowenergy partofthespectrumis
stronglyaffected bythechosenthreshold.
The isomeric state 83mKr is a very useful isotope for calibrating a time projection
chamberforseveralreasons:incontrasttootherfrequentlyusedcalibrationsources
6
es1400
ri
t
n1200
E
1000
800
Fig. 4. Simulation of the
600
83mKr decay spectrum as
400 seen by the two padrows
of the test-setup. Dif-
200
fusion, partially detected
0 clusters, and a threshold
0 10 20 30 40 50 60
E (keV) cutaretakenintoaccount.
such as 55Fe, 83Kr is a gas which can be distributed over the chamber volume us-
ing the existing gas system. Therefore, no laborious unmounting of the chamber
and installation of Fe sources is necessary. The mean lifetime of 83mKr is short
enough to ensure the chamber to operate normally again after a reasonably short
time, i.e. once a few half-lives have passed after cutting off the Krypton supply
to the chamber. If required, the gas can be left within the closed gas system un-
til the radioactivity subsides. On the other hand, the mean lifetime of the isomeric
state is long enough for a sufficient number of subsequent decays to occur inside
the chamber. Krypton’s parent isotope 83Rb has a sufficiently long mean lifetime
(τ = 124d), is a solid, and thus may be mounted as a foil inside a bypass line of
thegassystem.
As it is not possible to trigger on the Krypton decays, a random trigger must be
used. The decay rate must be high enough to ensure good statistics within a rea-
sonable time (an estimate of the number of events needed is carried out at the end
of section 4). Due to random triggering there is no information on the decay posi-
tioninthedriftdirection.Therefore,toobtainacleanspectrum,absorption(mainly
caused by oxygenpollutioninthegas)mustbesmall1.
4 The Measurements
AsneithertheSTARFTPCnortheFTPCreadoutelectronicswerecompletedwhen
themeasurementswereperformed,asmalltest-setupusingtheNA49readoutelec-
tronics [12,13] was constructed. For this test-setup, the same materials and gas-
mixture (Ar/CO 50/50) as for the FTPC, and a pad geometry comparable to the
2
FTPC (two non-adjacent padrowswith 16 pads each) were used. Themajordiffer-
1
Atan oxygen concentration of5ppm and themaximum drift time of50µs,17%of the
electrons areabsorbed [14].
7
s
e220
ntri
E200
180
160
140
120
100 Fig. 5. A Fe spectrum
80 measured on a single pad.
60 A combined fit consisting
40 oftwoGaussianswasper-
20 formed and the two sepa-
0 rate Gaussians are shown
0 2 4 6 8 10 12 14 16
Charge (arbitrary units) inthisplot.
ences between the test-setup and the FTPC was the axial drift field (in contrast to
the radial drift field of the FTPC) and the planar readout chamber. However, the
charge deposited on the pads is hardly affected by the drift field geometry and the
cluster sizes were comparable to those measured in a test TPC with a radial drift
field. Fordetailsonthetest-setupand themeasurements,see[14].
Thecalibrationwas performedinthreesteps:thepulsercalibration,thecalibration
with55Fe, and thecalibrationwith83mKr.
4.1 The PreparatoryCalibrations
The pulser calibration is described earlier in section 1.1. The amplification factors
ofthe16channels perpreamplifier/shaperchipdecrease characteristically within-
creasingchannelnumberi;theydifferbyupto15%.Thisbehaviorisapropertyof
thearchitecture of the NA49 FEE chips and is not observed in the FTPC electron-
icswherethefluctuationoftheamplificationfactorisbelow3%.Thus,apad-wise
calibrationwasnecessaryforthetest-setupbutmightnotbeneededforthecalibra-
tion of the FTPC. However, the pulseshape of the FTPC electronics is affected by
thecurrent load on thechip. Therisetimeofthepulseisreduced as thenumberof
pulsedchannels increases [15]whichwillaffect thetotal integratedclustercharge.
Thus,across-check ofthepulsercalibrationhas tobeperformed.
For themeasurements with the 55Fe source, two Aluminumstrips doped with 55Fe
were mounted at the cathode plate of the drift field cage, opposite to the two pad-
rows. 55Fe decays via electron capture to 55Mn which subsequently emits an X-
ray photon from the K-shell with an energy of 6keV. This photon may ionize an
Argon atom in the gas which releases an electron from the K-shell, which has a
bindingenergyof3.2keV.Inaddition,eitherAugerelectronsoranescapingX-ray
photon with an energy of 3.2keV are emitted, the latter being responsible for the
8
)1300
s
el1200
n
n1100
a
h
C1000
C 900 Fig. 6. The measured
AD800 chargeqi(integratedADC
( channels) from the 6keV
e 700
g Fe decays of each pad
r600
a
h500 i without any calibration.
C
Anode voltages from the
400
bottom up: U = 1700V,
300
1 6 11 16 21 26 31 1725V, 1750V, 1775V,
Pad Number
1800V.
1.5
b
1.4
1.3
1.2
1.1
1
0.9 Fig. 7. The measured
0.8 charge qi from Fe decays
of each pad i (as in fig. 6)
0.7
with calibration factors
0.6
obtained by the reference
0.5
1 6 11 16 21 26 31 measurement at 1750V.
Pad Number
β = qi(U)/qi(1750V)
2.8keVescapepeak[14].Thesupplementarycalibrationusing55Fewasperformed
forseveralreasons:
• A cross-check of the pulser calibration and a first order energy calibration was
needed totunetheanodevoltagefortheKryptonmeasurements.
• Determining the widths of the 6keV main peak and of the 3keV escape peak
gavean accuratemeasurement ofthechamberresolution.
• A comparison of the gas amplification at anode voltages in the operating range
oftheFTPC (1700V to1800V)wasrequired.
• The clear Fe spectrum allowed a straightforward determination of suitable cut
criteriainorderto reducenoise.
The electron energy deposition of 55Fe roughly corresponds to the energy deposi-
tionofaminimallyionizingparticle.Thus,itwaspossibletoperformanenergycal-
ibrationattheFTPC’soperatinganodevoltageofU =1750V.Inaddition,further
measurements at voltages from 1700V to 1800V were carried out. These results
were extrapolated to lower voltages in order to find the anode voltage where the
completeKryptonspectrumcanbemeasuredwithoutelectronicssaturationcaused
by thehighdecay energies of83mKr.
9
For each pad a separate Fe spectrum was measured and a combined fit of the two
peaks was performed (see figure 5). For the calculation of the calibration factors
only the position of the 6keV peak was considered. The width of the peak and
hencethedetectorresolutionweredeterminedtobeσ/E ≈12%. Thepositionsof
thefittedGaussiansonthechargespectraofeachpad(exceptthemarginalpads)is
shown as a function of the pad number in figure 6. By applying the Fe calibration
factors from a reference measurement at 1750V to measurements at different an-
ode voltages, the consistency of the Fe calibration and thus the expected precision
for the Krypton calibration was investigated (see figure 7, where only statistical
errors are shown).Theresultinguncertaintyisdeterminedto be<2%.
The comparison of the calibration factors obtained with the pulser and the iron
source shows notable differences. The drop in amplification with increasing chan-
<
nel number is smaller for calibration using the Fe source (∼11%) than for the
<
pulser calibration (∼15%). This is due to the lower current demand of the ampli-
fyingelectronicsinthecaseoftheFemeasurementbecauseonlyafewpadsreceive
charge signals.
Athoroughinvestigationofcutcriteriashowedthatconstraintsontheclustersizein
drift direction r and in φ direction (see coordinatesystem in figure 1) significantly
improve the spectra and reduce noise and background. An effective background
rejectionisofgreatimportanceforthefollowingcalibrationwith83mKr.Moreover,
clusterslocated at themarginal padsofeach padrowwererejected.
4.2 The KryptonCalibration
Inthefirststep,Kryptonspectrawererecordedathighanodevoltages(1600Vand
1700V), focusing on the 9.4keV peak. The calibration factors from the Fe refer-
encemeasurementwereappliedtothedataandthespectrafromtheindividualpads
were combined to one spectrum. By this means, the accuracy of the extrapolation
of the gas gain measurements with Fe from high anode voltages to lower voltages
waschecked.Thisextrapolationwascorrectwithinanerror< 3%,whichincludes
statistical and systematical errors. The optimum anode voltage for collecting the
completeKryptonspectrumwas determinedto be1500V.
In thesecond step, a highstatisticsrun at 1500V was carried out inorder to allow
a pad-wise calibration with the 41.6keV peak. In some of the high energy decays,
several electrons are produced consecutively in a short time span. This results in
charge clusters with multiple peaks. As the summed electron energy is relevant, a
specialclusterfinder2 wasusedthatdoesnotdeconvoluteoverlappingclusters,but
sumsupthetotalcharge.
2
The cluster finder calculates the charge and other parameters like position and width of
eachfoundclusterfromtherawdata.
10