Measurement of isolated photon production in pp and PbPb collisions at sqrt(sNN) = 2.76 TeV

0.45 MB·English

by  CMS Collaboration

EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH (CERN)
CERN-PH-EP/2011-221
2012/01/17
CMS-HIN-11-002
Measurement of isolated photon production in pp and
√
PbPb collisions at s = 2.76 TeV
NN
2
∗
1 The CMS Collaboration
0
2

n
a
J

5
1

] Abstract
x
e
-
l
c Isolated photon production is measured in proton-proton and lead-lead collisions
u at nucleon-nucleon centre-of-mass energies of 2.76 TeV in the pseudorapidity range
n
|η| < 1.44 and transverse energies E between 20 and 80GeV with the CMS detector
[ T
attheLHC.ThemeasuredETspectraarefoundtobeingoodagreementwithnext-to-
1
leading-orderperturbativeQCDpredictions. TheratioofPbPbtoppisolatedphoton
v
3 E -differentialyields,scaledbythenumberofincoherentnucleon-nucleoncollisions,
T
9 isconsistentwithunityforallPbPbreactioncentralities.
0
3
.
1
0 SubmittedtoPhysicsLettersB
2
1
:
v
i
X
r
a
∗SeeAppendixAforthelistofcollaborationmembers

1
1 Introduction
Promptphotonswithhightransverseenergy(E )inhadroniccollisionsareproduceddirectly
T
fromthehardscatteringoftwopartons. Atlowestorderintheelectromagneticandstrongcou-
plingconstants,threepartonicmechanismsproducepromptphotonsinhadroniccollisions: (i)
quark-gluonComptonscatteringqg → γq,(ii)quark-antiquarkannihilationqq → γg,and(iii)
collinear fragmentation of a final-state parton into a photon. Prompt photons from (i) and (ii)
are called “direct”; those from (iii) are called “fragmentation”. Measured photon production
crosssectionsprovideadirecttestofperturbativequantumchromodynamics(pQCD)[1],and
constrain the proton [2] and nuclear [3] parton distribution functions (PDFs). In the case of
nuclear collisions, jets are significantly suppressed [4] but direct photons as well as W and Z
bosons [5] are unaffected by the strongly interacting medium produced in the reaction. Thus,
theseelectroweakparticlesconstituteparticularly“clean”probesoftheinitialstateofthecolli-
sion. Inparticular,thedirectcomparisonofproductioncrosssectionsofsuchprobesinppand
nuclearcollisionsallowsonetoestimatepossiblemodificationsofthenuclearpartondensities
withrespecttoasimpleincoherentsuperpositionofnucleonPDFs.
However, the measurement of prompt photon production is complicated by the presence of a
largebackgroundcomingfromtheelectromagneticdecaysofneutralmesons(mostly π0,η →
γγ)producedinthefragmentationofhard-scatteredpartons. Sincehigh-transverse-momentum
(p ) neutral mesons are produced inside a jet, they are surrounded by significant hadronic
T
activity from other parton fragments. Thus, γ backgrounds from these decays are typically
suppressed by imposing isolation requirements on the reconstructed photon candidates. The
isolationrequirementsalsosignificantlysuppressthefragmentationphotoncomponent,while
removingveryfewofthephotonsarisingfromdirectprocesses. Sincetheannihilationcontri-
bution is relatively small at the Large Hadron Collider (LHC), the result is an isolated photon
sample dominated by quark-gluon Compton photons [2]. In heavy-ion collisions, the hard
scattering thatproduces an isolatedphoton is superimposedon the considerableactivity aris-
ing from multiple parton-parton scatterings (underlying event) occurring simultaneously. A
subtractionoftheunderlyingeventisthereforenecessarybeforeapplyingisolationcriteria.
In this paper, a measurement of the isolated photon production in pp and PbPb collisions at
√
nucleon-nucleon centre-of-mass energies s = 2.76TeV with the Compact Muon Solenoid
NN
(CMS) detector [6] is reported. This constitutes the first measurement of isolated photon pro-
ductioninheavy-ioncollisions. Sections2and3describethedetectorandtriggersusedinthe
analysis, while the Monte Carlo (MC) simulation and the PbPb reaction centrality determina-
tion are discussed in Sections 4 and 5. The photon reconstruction and identification methods
√
used in pp collisions follow very closely those described in the studies at s = 7TeV [7].
Theimprovementsintroducedinordertoadaptthephotonreconstructionandisolationtothe
high-multiplicity PbPb environment are discussed in Section 6. The photon signal extraction
and corrections are discussed in Section 7. The theoretical pQCD calculations from the JET-
PHOX program[1]arepresentedinSection8. Finally,themeasuredisolatedphoton ET spectra
inppandPbPbcollisionsarecomparedtothetheoryandtoeachotherinSection9.
2 The CMS detector
Final-stateparticlesproducedintheppandPbPbcollisionsaremeasuredandreconstructedin
the CMS detector, consisting of several sub-detector systems [6]. The central tracking system
comprises silicon pixel and strip detectors that allow for the reconstruction of the trajectories
of charged particles in the pseudorapidity range |η| < 2.5, where η = −ln[tan(θ/2)] and θ
2 3 Datasamples,triggersandeventselection
is the polar angle relative to the counterclockwise beam direction. CMS uses a right-handed
coordinatesystem,inwhichthezaxisrunsalongthebeam,theyaxisisdirectedupwards,and
the x axis lies in the accelerator plane and points towards the center of the LHC ring. Electro-
magnetic(ECAL)andhadron(HCAL)calorimetersarelocatedoutsidethetrackingsystemand
providecoveragefor|η| < 3. Inthecentral(“barrel”)pseudorapidityrange|η| < 1.44consid-
eredinthisanalysis,theECALandHCALcalorimetersarefinelysegmentedwithagranularity
of0.0174×0.0174and0.087×0.087,respectively,inη andazimuthalangleφ(inradians). The
calorimeters and tracking systems are located within the 3.8 T magnetic field of the super-
conducting solenoid. In addition to the barrel and endcap detectors, CMS includes a hadron
forward (HF) steel/quartz-fibre Cherenkov calorimeter, which covers the forward rapidities
3 < |η| < 5.2andisusedtodeterminethedegreeofoverlap(“centrality”)ofthetwocolliding
Pb nuclei. A set of scintillator tiles, the beam scintillator counters (BSC), is mounted on the
innersideoftheHFfortriggeringandbeam-halorejectionforbothppandPbPbcollisions.
3 Data samples, triggers and event selection
TheresultspresentedherearebasedoninclusivephotonsamplescollectedinppandPbPbcol-
lisionsat2.76TeVwithminimum-biasandphotontriggers. Thetotaldatasamplecorresponds
to an integrated luminosity of 231nb−1 and 6.8µb−1 for pp and PbPb, respectively. Note that
the pp-equivalent luminosity of the PbPb measurement, L = A2 ×L = 294nb−1
pp-equiv PbPb
(where A=208 is the nuclear mass number for Pb), is close to that of the pp data. For online
event selection, CMS uses a two-level trigger system: a level-1 (L1) and a high level trigger
(HLT). The trigger and event selection used for the pp analysis are described elsewhere [7].
PbPb events used in this analysis are selected by requiring a L1 electromagnetic cluster with
E > 5GeV and an HLT photon with E > 15GeV, where E values do not include offline
T T T
corrections for the calorimeter energy response. The efficiency of the photon trigger in PbPb
collisions is shown in Fig. 1 for photon candidates with |ηγ| < 1.44. The efficiency is greater
than98%forphotoncandidateswithcorrectedtransverseenergy Eγ > 20GeVinbothppand
T
PbPbcollisions.
In addition to the photon-triggered data sample, a minimum-bias (MB) PbPb event sample is
collected using coincidences between trigger signals from the +z and −z sides of either the
BSC or the HF. The minimum-bias trigger and event selection efficiency in PbPb collisions is
(97±3)%[4].
To select a pure sample of inelastic hadronic PbPb collisions, the contamination from electro-
magnetic (“ultra-peripheral”) collisions and non-collision beam background are removed fol-
lowingtheprescriptionsinRef.[4]. Eventsarepreselectediftheycontainareconstructedvertex
madeofatleasttwotrackswithvertexzposition|z| < 15cmandanofflineHFcoincidenceof
at least three towers with energy greater than 3GeV on each side of the interaction point. To
furthersuppressthebeam-gasandbeam-scrapingevents,thelengthofpixelclustersalongthe
beamdirectionisrequiredtobecompatiblewithparticlesoriginatingfromtheeventvertex.
Offline selection of pp and PbPb events for further analysis requires a photon candidate, de-
fined as described in Section 6, in the pseudorapidity range |ηγ| < 1.44 and with a corrected
transverseenergy Eγ > 20GeV,definingthephasespaceofthemeasurement.
T
3
1.2
s) CMS PbPb s = 2.76TeV ∫L dt = 6.8 µb 1
a NN
Bi 1

n
Mi
r / 0.8
e
g
g 0.6
ri
T

γ
(
y 0.4 Photon Trigger
nc Uncorrected ET > 15 GeV
e
ci 0.2
Effi
0
10 20 30 40 50 60 70 80
Calibrated Photon E (GeV)
T
Figure 1: Efficiency for the photon trigger as a function of the corrected photon transverse
energy in PbPb collisions at 2.76TeV, measured with the minimum-bias sample. Error bars
representthestatisticaluncertainty.
4 Monte Carlo simulation
InordertostudythephotonselectionefficiencyandelectronrejectioninPbPbcollisions,γ+jet,
dijet, and W → eν events are simulated using the PYTHIA Monte Carlo (MC) generator (ver-
sion 6.422, tune D6T) [8], modified to take into account the isospin of the colliding nuclei [9].
ThesesimulatedPYTHIAevents,propagatedthroughtheCMSdetectorusingtheGEANT4pack-
age[10]tosimulatethedetectorresponse,areembeddedinactualMBPbPbeventsinorderto
study the effect of the underlying event on the photon reconstruction and isolation. The em-
beddingisdonebymixingthesimulateddigitalinformationwiththerecordedMBPbPbdata.
These mixed samples (denoted “PYTHIA+DATA”) are used for signal shape studies, and for
energyandefficiencycorrections.
Inordertodeterminewhetheragivenphotonisisolatedatthegeneratorlevel,anisolationcone
of radius ∆R = (cid:112)(∆η)2+(∆φ)2 < 0.4 around its direction in pseudorapidity and azimuth is
defined. A photon is considered to be isolated if the sum of the E of all the other particles
T
produced from the same hard scattering inside the isolation cone is smaller than 5GeV. The
GEANT4simulationisalsousedtodeterminetheisolatedphotonenergyandefficiencycorrec-
tions.
5 PbPb centrality determination
FortheanalysisofPbPbevents,itisimportanttodeterminetheoverlaporimpactparameterof
thetwocollidingnuclei,usuallycalledthereaction“centrality”. Centralityisdeterminedwith
theminimum-biassampleusingthetotalsumofenergysignalsfromtheHF.ThePbPbMBdata
sample is divided into three percentile ranges of the total inelastic cross section: 0–10% (most
central,smallimpactparameter),10–30%(mid-central),and30–100%(peripheral,largeimpact
parameter). ThedistributionoftheHFenergy,alongwiththeintervalsdefiningthethreeevent
classes,areshowninFig.2. DetailsofthecentralitydeterminationaredescribedinRef.[4]. The
intervals can be correlated with geometrical properties of the collision using a Glauber model
simulation [11]. The two most commonly used physical quantities are N , the total num-
part
4 6 Photonreconstructionandidentification
CMS
s
vent 1 PbPb sNN = 2.76TeV ∫L dt = 6.8 µb 1
e Min Bias Trigger
s 10 1 Photon Trigger
a
Bi
n 10 2
ction of Mi10 3 30% 100% 10% 30% 0% 10%
a10 4
r
F
10 5
0 20 40 60 80 100 120 140 160
Sum HF Energy (TeV)
Figure 2: Probability distribution of the total HF energy for minimum-bias PbPb collisions
(black open histogram). The three regions separated by the vertical dotted lines correspond
to the centrality ranges used in this analysis. Also shown is the HF energy distribution for
thesubsetofeventspassingtheHLTphotontrigger(cross-hatchedhistogram),whichisabout
3.3%ofallminimum-biasevents.
ber of nucleons in the two Pb nuclei that experience at least one collision, and N , the total
coll
numberofinelasticnucleon-nucleoncollisions. Thevariable N isoftenusedtoquantifythe
part
reaction centrality, with N = 2 corresponding to a single nucleon-nucleon interaction and
part
N =2×208correspondingtoahead-onPbPbcollisionwhereallnucleonsparticipate. The
part
variable N quantifies the total number of incoherent nucleon-nucleon collisions at a given
coll
centrality,andsincethisisdirectlyproportionaltothehigh-p particleproductionyields, N
T coll
is used to normalize the PbPb yields for comparison with the same observables for hard pro-
cessesmeasuredinppcollisions. AscanbeseeninFig.2,thecentralitydistributionassociated
withhardprocesses,suchashigh-E photonproduction(cross-hatchedhistogram),hasamore
T
pronouncedcontributionfromcentralcollisionsthanforminimum-biasevents(solidline)
6 Photon reconstruction and identification
Isolated photon reconstruction in pp collisions is detailed in Ref. [12]. The reconstruction in
PbPb collisions is very similar, although some modifications are introduced in order to deal
with the large background of particles produced in the collision. ECAL “superclusters” are
reconstructedinthebarrelregionoftheelectromagneticcalorimeterusingthe“island”energy-
clustering algorithm [13]. The first step of the algorithm is a search around the seeds, which
aredefinedascells(reconstructedhits)withatransverseenergyaboveathresholdof0.5GeV.
Starting from a seed position, adjacent cells are examined, scanning first in the φ and then in
theη direction. Cellsareaddedtotheclusteruntilthecellunderconsiderationsatisfiesoneof
threeconditions;thecorrectedenergydepositinthecelliszero,theenergyinthecellislarger
than in the adjacent cell which was already added to the cluster, or the cell is already part of
adifferentislandcluster. Inthesecondstep, theislandclustersaremergedintosuperclusters.
Theprocedureisseededbysearchingforthemostenergeticclusteraboveatransverseenergy
threshold(E > 1GeV)andthencollectingalltheothernearbyclustersthathavenotyetbeen
T
∆ ∆
used in a narrow η-window ( η = 0.07), and a much wider φ-window ( φ = 0.8). A photon
candidateisconstructedfroma“supercluster”(conglomerateofenergydeposits)withuncali-
5
brated E > 8GeV,anditsenergyiscorrectedtoaccountforthematerialinfrontoftheECAL
T
and for electromagnetic shower containment. The direction of the photon is also recalculated
with respect to the primary vertex. An additional energy correction is applied to remove the
backgroundcontributionfromtheunderlyingPbPbevent. Thiscorrectionisobtainedfromthe
γ+jet PYTHIA+DATA sampleandlistedinTable1forthe3centralityintervals. Theunderlying
PbPbactivityalsoworsensthephotonenergyresolutiontoamaximumof9%forthelowestE
T
bininthe0–10%centralevents,asshowninFig.3.
Table 1: Energy correction factors for the background energy contribution found using the
γ+jet PYTHIA+DATA sample for each centrality interval and photon ET. The reconstructed ET
ofphotoncandidateswith|ηγ| < 1.44ismultipliedbythisfactortogetthecorrectedtransverse
γ
energy E .
T
Photon E PbPbcentrality
T
(GeV) 0–10% 10–30% 30–100%
20–25 0.90 0.94 0.99
25–30 0.91 0.95 0.99
30–40 0.92 0.95 0.99
40–50 0.94 0.96 0.99
50–80 0.95 0.97 0.99
12 12 12
CMS Simulation
%) 10 Energy Resolution%)10 %)10
n ( PYTHIA + DATA n ( n (
o 8 PbPb 0% - 10% o 8 PbPb 10% - 3o0% 8 PbPb 30% - 100%
uti uti uti
ol ol ol
s 6 s 6 s 6
e e e
R R R
y 4 y 4 y 4
g g g
r r r
e e e
n n n
E 2 E 2 E 2
20 30 40 50 60 70 30 40 50 60 70 30 40 50 60 70
E of photon (GeV) E of photon (GeV) E of photon (GeV)
T T T
Figure3: Relativeenergyresolutionofreconstructedphotonsasafunctionofphotontransverse
energy,determinedusingγ+jetPYTHIA+DATAsampleforthreecentralityintervals.
Anomalous signals caused by the interaction of heavily ionizing particles directly with the
siliconavalanchephotodiodesusedfortheECALbarrelreadoutareremovedbythefollowing
requirements: (i) the signal should be consistent in time (within 3ns) with a photon from the
collision;(ii)thesumoftheenergyinthefouradjacentcellssurroundingthecentralcellshould
be at least 10% of the central cell energy. These two selections are satisfied by 99.7% of the
photonsignalcandidates.
TheselectedphotoncandidatesarerequiredtobeintheECALbarrelwithinthepseudorapidity
interval |ηγ| < 1.44, to not match with any electron candidates in a search window of |ηγ −
ηTrack| < 0.02 and |φγ −φTrack| < 0.15 with respect to the associated electron candidate track,
and to have Eγ > 20GeV. A first rejection of neutral mesons mimicking a high-E photon
T T
candidate in the ECAL is done using the H/E ratio defined as the ratio of hadronic energy to
electromagnetic energy inside a cone of ∆R = 0.15, computed from the energy depositions in
theHCALandtheECAL[7]. Photoncandidateswith H/E < 0.2areselectedforthisanalysis.
6 7 Signalextraction,corrections,andsystematicuncertainties
To measure the isolation of a given photon candidate, the detector activity in a cone of radius
∆R = 0.4 with respect to the centroid of the cluster is used. Calorimeter-based isolation vari-
ables Iso and Iso are calculated by summing over the ECAL and HCAL transverse
ECAL HCAL
energy, respectively, measured inside the cone, while a track-based isolation variable Iso
Track
is measured by summing over the transverse momentum of all tracks with p > 2GeV/c in-
T
side the cone. The total ECAL energy associated with the photon candidate is excluded in
the Iso calculation. In order to remove the contribution of hadronic activity from the
ECAL
underlying PbPb event background falling inside the isolation cone for each centrality, the
average value of the energy deposited per unit area in the η −φ phase space ((cid:104)UE(cid:105)) is esti-
mated within a rectangular region 2∆R-wide and centered on ηγ in the η-direction and 2π
wide in the φ-direction, excluding the isolation cone. The UE-subtracted isolation variables
IsoUE−sub = Iso−π(∆R)2(cid:104)UE(cid:105) are used to further reject photon candidates originating from
jets. Thesumoftheisolationvariables (SumIsoUE−sub = IsoUE−sub+IsoUE−sub+IsoUE−sub) is
ECAL HCAL Track
requiredtobesmallerthan5GeV. Theefficiencyoftheisolatedphotonidentificationcriteriain
PbPbcollisions,whichisobtainedfromthePYTHIA+DATAsample,issummarizedinTable2.
Table 2: Efficiencies of the isolated photon identification at each step: clustering, anomalous
signal removal, H/E selection, and isolation requirement. Numbers in each row are the effi-
γ
cienciesrelativetothepreviousstep. TheselectionsaremoreefficientforhighE photonsand
T
γ
for more peripheral events. The intervals given indicate the E -dependent variations of the
T
efficiencies.
PbPbcentrality
Isolatedphotonidentification 0–10% 10–30% 30–100%
Superclusterreconstruction 96–99% 97–99% 97–99%
Anomaloussignalremoval 99–100% 99–100% 99–100%
H/E<0.2 96–99% 98–99% 99–100%
SumIsoUE−sub < 5GeV4 82–84% 86–88% 96–97%
Total 77–82% 83–86% 92–95%
7 Signal extraction, corrections, and systematic uncertainties
Theselectioncriteriadescribedaboveyieldarelativelypuresampleofisolatedphotons. How-
ever, there are still non-prompt photons, such as those from isolated π0s that are carrying a
largefractionoftheparentfragmentingpartonenergy,whichcanpasstheisolationcuts. Those
remaining backgrounds are estimated using a two-component fit of the shape of the electro-
magneticshowerintheECALandseparatedfromthesignalonastatisticalbasis,asdescribed
below.
The topology of the energy deposits can be used as a powerful tool to distinguish the signal
from the background by making use of the fine η segmentation of the electromagnetic calor-
imeter. The shower shape is characterized by a transverse shape variable σ , defined as a
ηη
modifiedsecondmomentoftheelectromagneticenergyclusterdistributionarounditsmeanη
position:
∑ w (η −η¯)2 E
σ2 = i i i , w = max(0,4.7+ln i), (1)
ηη ∑ w i E
i i
whereE andη aretheenergyandpositionoftheithcrystalinagroupof5×5crystalscentered
i i
ontheonewiththehighestenergy,Eisthetotalenergyofthecrystalsinthecalculationandη¯is
7
theaverageηweightedbyw inthesamegroup[7,14]. Isolatedphotonstendtohaveasmaller
i
mean value of σ and a narrow distribution, while photons produced in hadron decays tend
ηη
tohavelargerσ meanandawiderσ distribution.
ηη ηη
Theisolatedpromptphotonyieldisestimatedwithabinnedmaximumlikelihoodfittotheσ
ηη
γ
distribution with the expected signal and background components for each E interval. The
T
signal and background component shapes used in the pp analysis are described in [7]. In the
γ
PbPbanalysis,thesignalcomponentshapeforeachE andcentralitybinisobtainedfromγ+jet
T
PYTHIA+DATA samples, and the background component shape is extracted from data using a
background-enriched SumIso sideband (6 < SumIsoUE−sub < 11GeV) sample while keeping
allotherselectioncriteriaunchanged.
5000 1800 1400
CMS 1200
4000 sNN=2.76TeV pp (cid:242) L dt = 2311 n0b0-10 PbPb (cid:242) L dt = 6.8 m b11-4610000 PbPb 1200 PbPb
30 - 100%1200 10 - 30% 1000 0 - 10%
es3000 20 - 25 GeVes800 20 - 25 GeesV1000 20 - 25 GeesV800 20 - 25 GeV
ntri DDaattaa ntri600 ntri800 ntri600
E2000 SSiiggnnaallE E E
400 600 400
1000 BBaacckkggrroouunndd 400
200 200
200
2500 show0e.r0 1shape (s0.0)2 900 show0e.r0 1shape (s0.0)21800 show0e.r0 1shape (s0.0)21400 show0e.r0 1shape (s0.0)2
h h 80 h h 160 h h h h
200 pp 70 PbPb 140 PbPb 120 PbPb
es 150 es 60 30 - 100%es120 10 - 30%es100 0 - 10%
Entri 100 40 - 50 GeVEntri 4500 40 - 50 GeEntriV18000 40 - 50 GeEntriV 6800 40 - 50 GeV
30 60
40
50 20 40
10 20 20
0 0.01 0.02 0 0.01 0.02 0 0.01 0.02 0 0.01 0.02
shower shape (s ) shower shape (s ) shower shape (s ) shower shape (s )
h h h h h h h h
Figure4: Measuredshower-shapeσ distributionforphotoncandidateswithEγ = 20–25GeV
ηη T
and 40–50GeV in pp (2 left plots) and PbPb collisions for 3 different centrality ranges. The fit
result (red line), signal (red-hatched histogram) and background components (blue shaded
histogram)arealsoshown.
Figure 4illustratestheresultsofthetwo-componentfitoftheshower-shapedistributionmea-
suredinppandPbPbcollisions. Theremainingbackgroundcontributionfromelectronspass-
ing all the photon selection criteria, estimated from a sample enriched in isolated electrons
found by reversing the electron-veto requirement described in Section 6, is also subtracted to
extract the raw signal yields (Nγ ). Typically, the contribution due to electron contamination
raw
in PbPb collisions is 3–6% for different E intervals. A bin-by-bin correction for the energy
T
γ
smearing (U), which amounts to 1.00–1.08 for different E and centrality bins, is also applied
T
to the raw signal yields to obtain the number of isolated photons. The E -differential photon
T
yieldpereventisdefinedas
dNγ Nγ
PbPb = raw , (2)
dEγ U×(cid:101)× f ×N ×∆Eγ
T cent MB T
where N isthenumberofsampledminimum-biasPbPbevents, f isthefractionofPbPb
MB cent
events in each centrality bin, and (cid:101) is the efficiency of the isolated photon identification (Ta-
ble 2). For pp collisions we normalize the yields by the integrated luminosity (L ) to obtain
pp
8 7 Signalextraction,corrections,andsystematicuncertainties
γ γ
the E -differentialcrosssectiondσ /dE :
T pp T
dσγ Nγ
pp = raw . (3)
dEγ U×(cid:101)×L ×∆Eγ
T pp T
Table 3: Summary of the contributions to the estimated systematic uncertainties on the iso-
lated photon spectra measured in pp and PbPb collisions and their total. The nuclear overlap
γ
function T isdefinedinSection9. Theintervalsindicatethe E -dependentvariationsofthe
AA T
uncertainties.
pp PbPbcentrality
Source 0–10% 10–30% 30–100%
Efficiency 1–5% 5–9% 5–7% 5–6%
Signalmodeling 3–5% 1–5% 3–5% 1–4%
Backgroundmodeling 9–13% 15–23% 14–16% 12–21%
Electronveto 1% 3–6% 3–5% 3–5%
Photonisolationdefinition 2% 7% 5% 2%
Energyscale 3–6% 9% 9% 9%
Energysmearing 1% 4% 4% 4%
Shower-shapefit 3% 5% 5% 5%
Anomaloussignalcleaning 1% 1% 1% 1%
N – 3% 3% 3%
MB
Luminosity 6% – – –
Totalwithout T 14–16% 23–30% 22–25% 23–28%
AA
T – 4% 6% 12%
AA
Total 14–16% 23–30% 23–26% 26–31%
The systematic uncertainties of the measured photon spectra are summarized in Table 3. The
totalsystematicuncertaintiesare22–30%forPbPband14–16%forppcollisions. Thesystematic
γ γ
uncertaintyofthephotonyielddN /dE inPbPbcollisionsisdominatedbytheuncertainty
PbPb T
on the background modeling. Since the transverse shape variable σ may be correlated with
ηη
thenumberofparticlesintheisolationcone(characterizedbySumIsoUE−sub),non-promptpho-
tonsfrom PYTHIA+DATA samplesareusedtoexaminethepossibledifferencebetweenthe σηη
distributionintheSumIsoUE−subsignalandinthesidebandregions. Systematicchecksareper-
formed using the differences in the mean and width seen in the MC to vary the background
component shape in the fit. The estimated uncertainty is in the range of 12–23%, where the
γ
given interval indicates the E and centrality-dependent variations of the uncertainty. The
T
uncertainty due to the σ distribution of isolated photons is estimated by comparing the dis-
ηη
tributions of electrons from MC and data. Given the small number of Z → e+e− events in
√
the PbPb data sample, Z → e+e− events from the 2010 pp run at s = 7TeV are mixed with
MB PbPb data. The differences in the measured mean and width from those obtained in the
MC(Z → e+e−)+PbPbdataareusedtovarytheσ distributionsofisolatedphotons. Suchsys-
ηη
tematic changes result in a final propagated uncertainty of 1–5% in the isolated photon yield.
Theuncertaintyduetotheenergyscalepropagatestoanuncertaintyof9%inthefinalspectra.
Theuncertaintyduetotheenergysmearingcorrectionisobtainedbyvaryingtheassumediso-
latedphotondifferentialcrosssectionatlowphotonE (usedtoobtaintheunfoldingcorrection
T
factors)by ±50%,andisfoundtobe4%. Theuncertaintyofthetwo-componentfitischecked
byusingdifferentbinningwidthsinthefit,andisfoundtobe5%. A3–6%uncertaintyisasso-
ciated with the electron contamination subtraction. The difference between experimental and
theoretical photon isolation definitions as described in Section 8 due to the detector response