ATLAS calorimeter performance at low p$_T$ for $\gamma$, $\pi^0$ and $\eta$ particles identification in the B-physics context

HAL Id: in2p3-00011014

http://hal.in2p3.fr/in2p3-00011014

Submitted on 25 Oct 2001

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

and η particles identification in the B-physics context

F. Ohlsson-Malek, M. Melcher

To cite this version:

ATL-COM-PHYS-2001-021 ISN-01.75

September 10 th

, 2001.

ATLAS Calorimeter performance at low p T

for ,  0

and  particles identication

in the B-Physics Context

F.Ohlsson-Malek 1

,M. Melcher

Institut desSciences Nucleaires,UniversiteJ.Fourier, IN2P3-CNRS,F-38026Grenoble,France

Abstract

In ATLASCP violation is studied in the Bmeson decays. Some of the decay modes are subject to an extensive study in the liquid argon calorimeter such as B

s

decay to J=  where the  particle decaystotwo 'sandB

d !J= K 0 wheretheK 0 decaysto K 0  0 . Theobservabilityofthese channels issubjecttothecapabilityofthe calorimeterto detect thelow p

T

clusters produced fromthe decayof  and 

0

particles to two 's . It is also important to clean up the datafrom theelectronic noiseandthe pile-upeects sincethislatter producesmostlysoftchargedandneutralpions.

Onecan achievean detectioneÆciency of about2.26% withan amountoffakerateofabout24%whichmakesthestudyofB

s

!J=  feasible. A

0

detectioneÆciency of9%is obtainedbut thefakerate amountsto73%. Subsequently,itisshownthatB

d

!J= K 0

mode isratherdiÆcultto study.

1 Introduction and Motivations

CPviolation isstillone oftheexperimentallyleastconstrained phenomena intheStandardModeldescriptionofelectroweakinteractions. Avery fruit-ful testing ground for this, is the exploration of Physics with b- avoured

hadrons. At hadron machines, likethe LHC, very precise measurementsof CPviolationcan bedoneand clearanswersonthevalidityoftheStandard Model andthe existenceof New Physics willbeachieved.

TheB s

decaytoJ= isamodesimilartoB s

decaytoJ= (golden de-caymodefortheB

s

meson)withanupperlimitbranchingratioof3:8
10 3

.

Therefore,we expectsmallCP-violationeects withintheStandardModel and New Physics contributionsto B

s  B s

mixing, leadingto largeCP asym-metries. The B

s

decay to J=  is a pure CP eigenstate decay. Asa

con-sequence,there isno needto performheavy angularanalysisto distinguish between nal states as it will be the case for the J=  decay. Similar-ily to the work presented for B

d;s

! J= K s

in [1 ], it has been proposed

to extract the angle of the Unitarity triangle through measurements of CP asymmetries inthe B

d

! J=  process and measurements of the CP averaged widthsof theB

d;s

!J=  processes[2 ].

Argumentsinfavourofthemeasurement ofB d

!J= K 0

decaymode are numerous. It has a branching ratio of (1:580:27)
10

3

, interesting

compared to the golden plated mode (B d ! J= K s ) branching ratio of 5: 10 4

. Similarilyto this later mode, it can measure the CP asymmetry, helptoextracttheUnitarityangleandprobetheaccuracyoftheStandard Model. Let us remind that K

0 decays to (K 0 ! K s ; 0 ) and that B d ! J= K 0

is not a pure nal CP eigenstate mode. Fortunately, nal states withdierentCPquantumnumberscanbeseparatedbymeasuringangular distributions.

The observability of these channels is subject to the capability of the calorimeterto detect thelowp

T

clustersproducedfrom thedecayof  and

 0

particles to 2 's . It is also important to distinguish between clusters producedfrom single photons and those producedfrom the two 's decays which end superimposed onto 1 cluster in the calorimeter. We need also

to distinguish between  and  0

particles and to clean the data from the electronicnoiseand thepile-upeects sincethislatter producesmostlysoft charged and neutralpions.

TheLiquidArgonCalorimeterparticleidenticationpowerisstudiedin theoineanalysislevel.

2 Data samples and Analysis

and K 0 ! K 0 X (K 0 !K s

) were selected. The K 0 !K 0  0 branching rate ist next to 100%, but also a few K

0 ! K

0

decays appear [6 ]. The LVL1cuts(P

T

() >6GeV/c,j 2

j<2:5)werealsoappliedatthegeneration

level. TheeventgenerationisperformedusingPYTHIAversion5.7[3 ]where we chose the production of b

 b, forcing  b to decay to J=  (respectivelyto J= K 0

). The sets of data were used as inputs for a full simulation in

the ATLAS detector using the ATLAS Geant3 based program DICE, srt release 1.3.0. Thereconstructionwasperformed usingtheATLASprogram ATRECON, srt release 1.3.0. A total of 17000 (respectively 20000) fully

simulatedandreconstructed B s !J=  (resp. B d !J= K 0 )events were obtained. Threesamplesof about10000 singleparticles (; ;

0

) havealso beenusedforthedetailedstudyoftheparticleidenticationandtoperform

eÆcientcutsforpile-uprejection. TheCombinedNtuples(CBN-Tuples)[5 ] facilitywasused foranalysiswith PAWand ROOT.

IntheElectromagneticCALorimeter(ECAL),aclusterisdenedasthe energydepositinsidetheregion

 =0.075x1.25, whichcorresponds to 3x5cellsinthe2nd sampling. Tondclusterscandidatesthecellsofthe

dierentsamplingsaremappedontotowersofsize0.025x0.025 andasliding window algorithm is used, taking a 3x3 towers and a transverse energy thresholdof1 GeV.The electronic noiseandthe pile-upwereaddedat the

reconstructionlevel asdescribedin[4].

An algorithmbasedontheselectionofthetwoclustersofhighest

trans-verse energies was applied to build the clusters invariant mass using their kinematical characteristics (E cl uster ,  cl uster and  cl uster ). Only 20% of the generated  particles decayinginto two 's have a openingangle higher

than 5 Æ

and a energy higher than 1 GeV. Therefore, the  detection ef-ciency will be limited to 20% due to the resolution of the ECAL and the clusterizationconditions.

Another algorithm hasbeenbuilt to collect events where theshower of the two 's from the  decayare overlapping and which are reconstructed

as one cluster in the ECAL. This is also the case for the  0

's where most events, dueto thesmall openingangle, give1 cluster.

For the following analysis, only the ten electromagnetic clusters with highesttransverse energyare taken into account.

In order to obtain reconstruction eÆciencies and the fake rates, the reconstructed particles had to be identied with the generation data, the 'truth'. Forthis,thedirectionofthegeneratedparticleandthedirectionof

thereconstructed one were compared. A reconstructed particlewas saidto bea trueparticle, ifa generatedparticlesatisfying

\( ~p generated ;~p reconstructed )<2 Æ (2.1) 2

was found. When there were more than one true particle satisfying this relation,theparticle withtheclosest angle wasselected.

3 Denitions: EÆciency and fake rate

Themaincriteriaforjudgingthereconstructionqualityarethe

reconstruc-tioneÆciency andthe fake rate,denedasfollows:

 N X

: Numberofreconstructed X particles

 N true X

: Numberofreconstructed Xparticles associatedto agenerated X particle

 N fake X

: Number of reconstructed X particles not associated to a

gen-erated X particle: N X =N true X +N fake X  N gen X

: Numberof generated Xparticles

 N true X from Y

: Numberof reconstructed particles associatedto a

gener-ated X particlebelonging to thedecaychannelY

 N gen X from Y

: Number of generated X particles belonging to the decay channelY

 Overall Xreconstruction eÆciency:

(X)= N true X N gen X (3.1)  X from Y eÆciency (X fromY)= N true X from Y N gen X from Y (3.2)  fake rate: f X = N fake X N X (3.3) 4  reconstruction 4.1 The visibility

 particles decaying into two ...

All  from B

s

(1) pointingto two clusters 1967 =^ 12.7% 1809 =^ 14.5%

(2) pointing to the same cluster(superimposition)

447 =^ 2.9% 428 =^ 3.4%

(3) Onepointingtoacluster, one withoutcluster

5028 =^ 32.4% 4765 =^ 38.1%

(4) Notpointingto any

clus-ter

8060 =^ 52.0% 5516 =^ 44.1%

Total 15502 12518

Table1:  visibility

 The openinganglehastobehighenough(>5 Æ

)sothattheclusters intheECALarenotsuperimposed.

 The energies have to be suÆcient (p T

> 1 GeV) for the clusterto befoundbytheclusterizationalgorithm.

Theratioofparticlessatisfyingthesetwoconditionsgivesanupperlimitto

thereconstructioneÆciencywhichcanbeachievedbythemethodpresented here. Asshownintable 1,themaximum reconstructioneÆciency thatcan beachieved is14:5%. Mostofthe  particlesarelostbecausetheenergy of

at leastone of the particlesis lessthan1 GeV.

4.2 Reconstruction without pile-up

Thedistribution oftheinvariant massforallpossibletwo-cluster-pairs, g-ure1(a), shows theetapeakabovethe combinatorial background. In order

to suppressthiscombinatorialbackground, weonlylookat thetwoclusters ofhighesttransverse energy. Choosingtheclustersbytotal energy(instead of transverse energy) was shown to give worse results (smaller eÆciencies

and higherfake rates). The resulting invariantmass distributionis plotted in gure 1(b). 4.0% of the  particles from B

s

decays were found in an invariant masswindowof 2 around the mass(0:547 GeV [6 ]),butstill

42% of the reconstructed  arefakes. The parameters of the gauss-t and theobtainedeÆcienciesare summedup intable 3 (leftside).

For all thefollowing analysis,an invariant mass windowof 2 willbe

used.

4.3 Reconstruction with pile-up

GeV

0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

0

100

200

300

400

500

600

700

800

900

(a)Datawithoutpile-up,combinatorial

GeV

0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

0

20

40

60

80

100

120

140

160

180

200

η

π

(b)Datawithoutpile-up,nocuts

Figure 1: Two clusters invariant mass. (a) combinatorial, (b)2 clustersof

highestp T

,no pile-upand no cuts.

GeV

0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

0

20

40

60

80

100

η

π

(a)Datawithpile-up,nocuts

GeV

0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

0

20

40

60

80

100

η

π

(b)Datawithpile-up,cutsapplied

Figure 2: 2 clusters invariant mass: 2 clusters of highest p T

into account, each cluster not originatingfrom the  decay may hide an  cluster. Consequently,the reconstructioneÆciency falls to 2.9%, whilethe fake rate increases to 45%. The detailed results are summed up in table

3(right side).

Using thegeneration data, thereal originsof clusters which were iden-tiedas from  (in the2 invariant masswindow)weresearchedfor. As

listed up in table 2, the main contributions are coming from charged and neutralpions, photonsandnoise.

 785 =^ 72%  0 46 =^ 5%   30 =^ 3% 116 =^ 11%

other particles andnoise 102 =^ 9%

Table 2: Originof clusters being identied as from  (2 invariant mass window)

The taskis nowto suppresseverything thatdoesn'toriginate from real  decays.

4.4 Cluster rejection

Clusters originating from hadrons or noise can be identied and rejected

usinggeometricalpropertiesof thecluster:

 In order to reject background photons,which are situated at low p T

, andtheelectronicnoise,clusterswererejectedifthetransverseenergy islessthan2GeV.Asthephotonsfromarealsosituatedatratherlow p

T

,thiscutalonereducessignicantlythe reconstruction eÆciency.

 In the ECAL, hadrons are not absorbed as easily as e 

and

par-ticles. Therefore, they deposit a higher part of their energy in the thirdsamplingoftheECALand intherstsamplingof thehadronic calorimeter. Therefore, clusters were rejected ifmore than4% of the

total energy is deposited in the third sampling or if more than 10% ofthe total transverse energy isdepositedinthe rstsampling ofthe HCAL,see gures3(a)and 3(b).

 Clustersoriginatingfrom photonsarethinner thanhadronicclusters,

includingclustersfrom 0

,which arethesuperimpositionoftwo pho-tonswithslightlydierentdirections. Therefore,clusterswererejected ifthetotal clusterwidth,calculatedusingtheenergydepositionin40 stripsoftherst sampling,islarger than8 strips,seegure 3(c).

 reconstructionwithoutpile-up #events 17039 N gen  20795 N gen  from Bs 17094 N  1164 N true  684 N true  from B s 676 ( from B s ) 4.0% f  42%

Gausstmean 0.552 GeV Gausst 0.032 GeV

 reconstructionwith pile-up

#events 16697 N gen  20368 N gen  fromBs 16749 N  833 N true  479 N true  fromB s 475 ( fromB s ) 2.9% f  45%

Gausst mean 0.551GeV Gausst  0.039GeV

Table 3: Reconstruction eÆciencies without cuts, data without and with pile-up. Forthedenitionofthe given numberssee section3 on page4.

isdepositedina3x3cellsregioninthesecondsampling. E 33

beingthe energy deposited inthis3x3 windowof thesecond sampling, E

37 the energydepositedina3(dir.)x7(dir)windowandE

77

theenergyin a 7x7 towers window, clusterswere rejectedif

E 37 E 33 E 37 >20% or E 77 E 37 E 77 >20% (4.1)

thusdemanding thatlessthan20% oftheenergyisdeposited outside the 3x3windowinboth and directions,seegure 3(d).

Thecutswereappliedinsuchawaythatthersttwoclusters(ofhighest E

T

) passing the cuts were combined for the  reconstruction. While the eÆciency fellto 2.26%, as also some real  clusters were rejected, the fake rate could be lowered to 24%. In theobtained invariant massdistribution,

asshowningure 2(b),next to all thenoisehasdisappeared. Thedetailed reconstructionresultsareshownintable4andthedenitionofthevariables

usedintable 5. Ifwe usethe same cuts onthe datawithoutpile-up, the reconstructioneÆciencyisfoundto be equalto3.28% whilethefake rate isequal to 20%. Therefore, the pile-upisseen to decreasesignicantlythe

reconstructioneÆciency whileit does notincreasethefake rate somuch.

4.5  !1 cluster events

Aslistedupintable1,insomeoftheparticledecaysthe openingangle issosmallthat thetwoclustersare superimposedand identiedasasingle

0

0.02 0.04 0.06 0.08

0.1

0.12 0.14

1

10

10

2

10

3

0

0.05

0.1

0.15

0.2

0.25

0.3

1

10

10

2

10

3

Strip units

0

2

4

6

8

10

12

14

1

10

10

2

10

3

(c)Totalclusterwidth

GeV

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1

10

10

2

10

3

10

4

Leakage E37 vs E33

GeV

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1

10

10

2

10

3

10

4

Leakage E77 vs E37

(d)Energyleakagein(top)anddirection

Figure 3: Cluster properties for clustersfrom  ({),  

( ),  0

(


) and

therest(
),aswellasappliedcuts(verticallines). Onlytwo clustersof highestE

T

Final cutsfor fromB s selection E T > 2.GeV E3 35 E 35 < 4 % Etha1 E T < 10 % Wtots1 < 8.strips E37 E33 E 37 < 20 % E 77 E 37 E 77 < 20 %

 reconstruction withpile-up

#events 16697 N gen  20368 N gen  from Bs 16749 N  461 N true  371 N true  from B s 367 ( from B s ) 2.26% f  24%

Gausstmean 0.55 GeV Gausst 0.033 GeV

Table4: (Left side)Summaryofnalcuts appliedforthe selection;(right side)ReconstructioneÆcienciesafter cutapplication,datawith pile-up.

E 35

Total EnergyintheECAL(3x5 cell)

E T

Transverse Energyin theECAL(3x5cell) Wtots1 Total clusterwidth(over40 strips)

E3 35

Energyinthirdsamplingof theECAL(3x5 cell) Etha1 ETinrst samplingof HADcalorimeter

Table 5: Variablesusedfor selection

 The transverseenergy hasto beat least10 GeV.Thisrejects mostof the pile-upclusters. Astheclustersnow containtheenergyof both particles, the typical cluster energy is much higher than in 2 cluster

decays.

 The total cluster width (over 40 strips) has to be between 2 and 9 strips. It was shown that mostof the clusters have a width within

thiswindow.

 Duetothehighpseudorapidityresolutioninrstsampling,twoenergy maxima are found corresponding to the two , enclosing an energy

minimuminbetween. Clusterswere rejected,if thisenergyminimum is more than0.4% of thetotal transverse energy.

 The energyleakageinanddirectionsarecorrelated. Clusterswere

rejected, if(E 37 E 33 )=E 37 >15%and(E 77 E 37 )=E 77 >15%. Also, clusters wererejected if(E

37 E 33 )=E 37 >50%or(E 77 E 37 )=E 77 >

50%. This corresponds to a selection of the clusters insidethe frame showningure 4.

direction

φ

0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

1

direction

η

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Leakage correlation

Figure 4: Energyleakage in  and  direction of ! 1 cluster eventsand appliedcut. Seecomments intext page10,thirdbullet.

#events 16697 N gen  20368 N gen !1 cl uster 477 N !1 cl uster 1073 N true !1 cl uster 166 (!1 cluster) 35% f !1 cl uster 85%

Table6:  !1clusterreconstruction withpile-up

eÆcienciesareshown intable 6.

Here (andonlyhere),forthe identicationwiththetruth,thecluster

energy was forced to be withina windowof 30% aroundthe generated  particle energy. As before, the direction of cluster and generated particle had to matchwithin2

Æ .

TheReconstructioneÆciencyandthefakeratewerenotaectedby pile-up,because theclusterenergiesarenowsuperiorto theenergiesof clusters comingfrom pile-up.

Nevertheless, the obtained fake rate was very high (85%), while the

 ! 1 cluster eÆciency was 35%, much higher than the  ! 2 cluster eÆciency.

#events 16679

# 25370

# !1 cluster 9369 =^ 36.9%

#converted 675 =^ 2.7%

#converted ,stillgiving1 cluster 361 =^ 1.4% #converted ,no clusterin dir. 314 =^ 1.2%

Table 7: Reconstructed photon conversions. Only photonsfrom the decay

channelB s

!J=  ( ! ) werecounted.

4.6 Conversions in the inner detector

In the inner detector, approx. 20% of the photons convert into e +

=e

-pairs. Depending on the radius where the conversion takes place, some of theclustersfromthee

+

=e areseparated(notsuperimposed)anddeviated from the original photon direction by the applied magnetic eld. In this

case, the photon cannot be found using only the ECAL data. However, photonconversionscan bereconstructed usinginnerdetector tracks.

ThenumberofreconstructedphotonconversionsforphotonsfromB s

!

J= ( ! ) isshownintable 7.

Thealgorithmusedtotakeinto accounttheconversionsforthe recon-struction tries the following combinations, until an invariant mass within

the2 is found:

1. TwoECALclusters ofhighest E T

2. ECALclusterof highestE T

and conversionof highestp T

3. Twoconversions ofhighest p T

Inordertosuppressfakeconversions,conversionswererejectedifthe 2

t parameter was superior to 5, value taken arbitrarily. Cluster rejection wasnotchanged.

Taking into account the conversions, the  from B s

eÆciency slightly risesfrom2.26% to2.4%, whereasthefakerate risesa littlebit. Therefore, theconversionswere notconsideredfortheB

s

reconstruction,too.

4.7 Consequences for B s

!J=  reconstruction

Withoutpile-up,we caneasilyndthe peakat 0.547GeVwhen calculat-ingthe two clusters invariant mass. The combinatorial backgroundcan be mostlyeliminatedbycombiningonlythetwo clustersof highesttransverse

energy. The eÆciency of  reconstructionfor  particles from B s

particles (such as electrons and positrons) and electronic noise. They are also responsiblefortheratherhuge fakerate (45%). Thepionscan be sup-pressed by applying cuts on the energy deposited in the 3

rd

EM-sampling

andthe1 st

HCALsampling,ontheclusterwidthandontheenergyleakages inanddirections. Electronicnoiseandapartofthephotoncontribution can be eliminated by applying a 2 GeV transverse energy cut. An  from

B s

eÆciencyof 2.26% wasachieved. The fake ratebeingat 24%, thesignal isratherclean.

 ! 1 cluster events are uninteresting for the reconstruction, as the clusterscannot beeasilyidentied.

As we are at low E T

, only few photon conversions into e 

-pairs were

reconstructed. Consequently, adding conversions to the analysis doesn't changethereconstruction eÆciencybymuchand givesa higherfake rate.

After having applied the reconstruction algorithm (without 1 cluster events and conversions) on the physical background, consisting of B ! J= X events, the signicance of the signal, which is a measure for the

cleanliness, was calculated. The signal was found to be visibleenough for thephysics analysis[7 , 8]sincethesignicance wasof theorder of100.

5 K

0

reconstruction

The reconstructionof the decaychannelK 0 !K 0  0 (K 0 !K 0 s ! +  )

consists of two tasks: K 0 s

reconstruction by identifying the  

pairs (in the inner detector) and 

0

reconstruction (in the ECAL). For this work, the focus was drawn to the 

0

reconstruction; in order to decide wether

an inner detector track came from a  

, the truth was used. Still, forthe reconstruction,themeasuredmomentum and directionwere used.

5.1  0

= separation

The opening angle of  0

! decays is so small that next to all  0

giveonlyoneclusterintheECAL,astheclustersfromthetwo photonsare superimposed. Therefore,somecriteriafor

0

= separationhadtobefound.

In order to do this, a set of fully simulated single  0

and single events withoutelectronicnoiseandpile-upwasusedandtheclusterpropertieswere analyzed and compared. The transverse energy given to the 

0

's/photons

wasvariedfrom 1to30 GeV;thetransverse energyoftheresultingclusters isshowningure 5(a).

As the  0

clusters are the superimpositionof two photon clusters with slightlydierent directions, thefollowing convention can be used for a dif-ferentiation:

 Clustershape:  0

0

5

10

15

20

25

30

35

1

10

10

2

(a)TransverseenergyinGeV asmeasured in ECAL

0

10

20

30

40

50

1

10

10

2

10

3

(b)Width(40strips)/Width(3 strips)

0

0.2

0.4

0.6

0.8

1

1

10

10

2

10

3

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

1

10

10

2

10

3

sampl:) E(min:btw :2 maxima) E

T (1

st sampl:)

Figure5: Clusterpropertiesforclustersfrom and 0

(dashedline): Study froma sampleof simulatedsingle 

0

5.2  selection 15

Cuts applied Surviving  0

rate Surviving rate E2tsts1 E1 35 =cosh ( 2 ) >0:06 E2ts1 Emins1 E135=cosh (2) >0:02 45.0% 4.4% Wtots1>3 E2tsts1 E135=cosh (2) >0:03 E2ts1 Emins1 E1 35 =cosh ( 2 ) >0:015 Emins1=E1 35 >0:01 E2ts1=E1 35 >0:02 22.8% 2.9% Table 8:  0 = separationat p T

=0::30GeV. Thevariablesareexplainedin table 9.

E2tsts1 Transverse energyof 2nd max (1 strip)in rstsampling

E2ts1 Transverse energyof 2nd max (3 strips)inrst sampling Emins1 Energy minimumbetweentwo maximainrst sampling Wtots1 Total clusterwidth(over40 strips)

E1 35

Energy inrst sampling(3x5 cell) 

2

Pseudorapidityas measuredinsecondsampling

Weta1 Cluster width(over 3strips)

Table 9: Variablesusedfor 0

= separation

seegure 5(b).

 Second max: The rst sampling of the ECALhas a high granularity in the  direction. Therefore, two energy deposition maxima can be identied,correspondingtothetwophotons,gure6. Useful

informa-tion, such as the transverse energy of the second maximum and the minimumbetween thetwo maxima is providedbythe reconstruction softwareandcan easilybeusedfordierentiation,seegures5(c)and

5(d).

High photon rejection and acceptable  0

eÆciencies could be achieved by applying simple cuts on the mentioned cluster properties, as shown in

table 8. 5.2  0 selection In order to nd the  0

coming from the K 0

decay, the ECAL cluster of

highest p T

which passesthe chosen cuts was selected. Comparedto the  0

studied above, the  0

concerned here are at lower transverse momentum, andthecutspresentedaboveappeartobenexttouseless. Additionally,the

Figure 6: Substructure of a hadronic cluster in the 1 st

sampling of the ECAL[9 ]: Twoenergy maximaarevisible

gure7,wichmakesthesearch even more diÆcult.

Figure 8(a) shows the correlation between generated and reconstructed

transverse momentum: Theymatch withinanacceptable range,astheroot mean square of (p reconstructed T p generated T )=p generated T was 0.18, as shown in gure8(b). The  0

rejection was optimized in order to obtain the lowest possible

fakerate. Table10 shows theappliedcutsand theeÆciencyobtained: The achieved

0

from K 0

eÆciencywas 9.0%,whilethefake rate was73%.

Thecuts werefoundbyusinga simplesearch algorithmwhichndsthe cutvaluegivingtheminimalfakerate,providedthatnotmore than70%of the real 

0

's are rejected. The cuts slightly improve the  0

fake rate, but

the  0 from K 0 (B d

) eÆciency has dropped. With and without the cuts applied,approximatelythesameamountoftherealreconstructed

0 (38%) isnotcomingfrom K

0 decays. 5.3 K 0 S reconstruction FortheK 0 S ! +

 reconstruction,theinnerdetector tracks wereusedfor

 

reconstruction.

Several approaches were made inorder to reconstruct theK 0 S

particles. In all cases, the truth (data from generation) was used in order to decide whether a track iscoming from a 



. However, forthe reconstruction,the

momentumobserved bytheinnerdetector wasused. In allcases, theinner detectortracks witha t parameter

2

>6 were rejected. Two ofthe used methodsarepresentedbelow

1. Forreference: Use of thesimulation truth. Usingthedatafrom

generation, the 

5.3 K S reconstruction 17 Cutsfor 0 fromK 0 selection E T > 2.5 GeV E 37 E 33 E37 > 0:02 1:8 < Wtots1 < 7:8 0:04 < E2tsts1 E35=cosh(2) < 1:44 3:2 < Wtots1 Weta1 < 16:8 E2tsts1 Emins1 E 35 =cosh( 2 ) > 0:016  0

selectionwithpile-up

#events 9016 N gen  0 60104 N gen  0 from B d !J= K 0 8963 N  0 4811 N true  0 1300 N true  0 from B d !J= K 0 805 ( 0 from B d !J= K 0 ) 9% f  73.0% Table 10:  0 from K 0

reconstruction: Applied cuts (left side) and eÆ-ciencies (right side). The meanings of the variables are listed up on page 15.

GeV/c

0

2

4

6

8

10

12

14

1

10

10

2

generation

T

P

0

5

10

15

20

25

30

reconstruction

T

P

0

5

10

15

20

25

30

(a) pT from generation (x axis) and recon-structedp T (y axis) for  0 from K 0 , GeV/c units.

-0.5 -0.4 -0.3 -0.2 -0.1 -0 0.1 0.2 0.3 0.4 0.5

0

5

10

15

20

25

30

35

fromgenerationand reconstruction

were searchedfor. Consequently, the K 0 S

fake rate is practically 0.

The K 0 S fromK 0 (B d ) eÆciencywas31%. 2. Select  

with best invariant mass. Finally, each  +

was com-bined with each  , and the combination giving the best invariant

mass(closesttotheK 0

mass)waschosen. Thefakerate risesto85%, and the eÆciencydrops to 11%.

The detailed results of K 0 S

reconstruction using these two methods are given intable11. Asbefore,an invariantmasswindowof2 wasusedfor allanalysis. The invariant massdistributionsare shownin gure9.

AnoptimizedK 0 S

reconstructionmethodusingspeciccutsonthetracks

andvertexingisdescribed in[10],where aneÆciency of41% wasachieved.

5.4 K 0

reconstruction

The last step is to combine the reconstructed K 0 S

with the  0

particles

in order to reconstruct the K 0

particle. The best results were obtained by combining the best invariant mass method for the K

0 S

with the cuts for the

0

as described above; The K 0 S

 0

5.4 K reconstruction 19

K 0 S

reconstruction withpile-up

#events 11274 N gen K 0 S 19213 N gen K 0 S fromB d 11274

Usingthesimulationtruth

N K 0 S 3502 N true K 0 S 3484 N true K 0 S from B d 3484 (K 0 S fromB d ) 31% f K 0 S 0.5%

Gausstmean 0.499 GeV Gausst 0.057 GeV

 

withbestinvariant mass

N K 0 S 9139 N true K 0 S 1359 N true K 0 S fromB d 1193 (K 0 S from B d ) 11% f K 0 S 85%

Gausst mean 0.498 GeV Gausst  0.014 GeV Table 11: K 0 S ! +  reconstruction

GeV

0

0.2 0.4 0.6 0.8

1

1.2 1.4

0

100

200

300

400

500

600

700

0

S

K

 invariantmass: Usingthetruth

GeV

0

0.2 0.4 0.6 0.8

1

1.2 1.4

0

1000

2000

3000

4000

5000

0

S

K

withbest in-variantmass

Figure9: K 0 S

GeV

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0

10

20

30

40

50

60

70

80

90

K

*0

invariant mass

obtained eÆciency with thismethod was1.2% with a fake rate of 75%. If wetrytoselectthe

0

clustersclosetotheK 0 S

momentumdirection,wewill

probablyreduce the background and improve the reconstruction eÆciency signicantly. Thistaskhasnotbeendoneyet.

5.5 Consequences for B d !J= K 0 reconstruction Whereas 0

= -distinctionappearsnottoposebigproblemsathigher trans-verse momentum, no eective distinction criterium could be found which

could have helped to identify the neutral pions from K 0

decays. Some cutswerefoundwhichslightlyamelioratethefakeratebutwhichrejectalso many real 

0

particles. The cuts had the same impact on the K 0

recon-struction: The fake rate dropped by some percent, but a great amount of K

0

was lost, too. Table 13 gives a short overview of the K 0

eÆciencies obtainedbythedierentmethods.

ItappearsthattheK 0

reconstructioncouldstillberescuedbyabetter

K 0 S

reconstruction method, as the K 0

fake rate is at a rather low level if thetruthis usedfortheK

0 S

,even ifthe 0

fake rateis veryhigh. Selecting 

0

's onlyaround the K 0 S

direction(instead of taking thewhole  0

angular

spectrum), might reduce the background signicantly, thus improving the reconstructioneÆciency. In addition,abigpart ofthe K

0

shouldpossibly be suppressed ifthe combination withthe J= particle is donein order to

K 0

reconstructionwith pile-up

#events 11274 N gen K 0 fromB d 11274 N K 0 569 N true K 0 139 N true K 0 fromB d 137 (K 0 fromB d ) 1.2% f K 0 75%

Gausstmean 0.902GeV Gausst 0.147GeV

Table 12: K 0

reconstruction; with pile-up; K 0 S

reconstruction using the bestinvariant massmethod;

0

selection aftercut application.

 0 selection K 0 S reconstruction K 0 (B d ) eÆciency K 0 fake rate

no cuts usingthetruth 5.0% 12%

no cuts bestinv. mass 2.2% 77%

cuts applied usingthetruth 3.0% 11%

cuts applied bestinv. mass 1.2% 75%

Table 13: K 0

reconstruction: Summary

Making the most optimistic assumptions, and assuming additionnally that the J= particles can be reconstructed with a good eÆciency and a very lowfake rate,theupperlimittotheB

d

signal signicanceof60can

beachieved.

6 Conclusion

The goal of this work was to nd strategies to identify low p T ,  and  0 particles from B s ! J=  and B d ! J= K 0

decays and to give estimates on the reconstruction eÆciencies which will be achieved in the

ATLASdetectorat LHC.

Using cuts on cluster width, energy leakage and energy deposition in thethird ECALand rst HCALsampling,the clustersfrom  ! could be identied, rejecting the major part of hadronic jets and noise. An 

reconstruction eÆciency of 2.26% could be achieved; the ratio of fake  particlesbeingat24%,thereconstructioncanbesaidtobesuÆcientlyclean. IncombinationwiththeJ= reconstruction[7 ,8 ], theexpectedsignicance

ofthe B s

signal wasfoundto besuÆcientlyhigh (100). =

0

separation, needed to identify the  0

particles from K 0

decays,

theclusters (second maximum, cluster width). Unfortunately, the 0 from K 0 are at low p T

(some GeV) and energetically at the same level as the pile-up. Therefore, the identication is much more diÆcult. Nevertheless,

a reconstruction eÆciency of 9% could be achieved, whereas 73% of the reconstructed 

0

were fakes. Thereby, the measurement of the transverse momentumsatisestherequirements. It wasshownthatalargepartofthe

fake pionscan be rejected by combining them with the K S

particles (K 0

reconstruction),usinganoptimalK S

reconstructionalgorithm. Despitethe diÆculties we encountered for obtaining a clean 

0

signal, the signicance

oftheB d

signalwas foundto be satisfying(60).

Acknowledgements

This work has been performed within the ATLAS Collaboration and we

have made useof the software framework and tools whichare the resultof collaboration-wide eorts. We thank collaborationmembers forthehelpful discussionswithspecialacknowledgementstoJohannCollot,DavidRousseau

andMonikaWielersforreading,commentingandmakingsuggestionsto im-prove thisnote.

References

[1] R.Fleischer,CERN-TH/99-78, hep-ph/9903455.

[2] P.Z.Skands, Int.J.HighEnergyPhys.0101 (2001) 8,hep-ph/0010115

[3] T. Sjostrand, PYTHIA 5.7 and Jetset 7.4, hep-ph/9508391 and

Computer Physics Commun.82 (1994) 74.

see also the ATLAS B physics simulation web Page: http://msmizans.home.cern.ch/msmizans/production/0.html.

[4] S.Simion,ATLASnote: ATL-SOFT-99-001

[5] CBN-Tuples: http://droussea.home.cern.ch/droussea/cbnt/cbnt.html

[6] Reviewof Particle Physics, Europ.Phys. Journ.C15,(2000) 1-4.

[7] C. Driouichi, P. Eerola, M. Melcher, F. Ohlsson-Malek and S. Viret, scientic note inpreparation,2001.

[8] J= reconstruction in the inner detector, S. Viret, rapport de stage

DEA, matiereetrayonnement,2001.

[9] M.Wielers, ATLASnote: ATL-PHYS-99-016

[10] K 0 S