1. Electron identification
M. Jaffré
LAL Orsay
Outline :
L1 and L2 EM objects, L3 tools
EMReco algorithms (new from preco05)
Energy, position corrections
Match with PS and track
H-matrices, Likelihood
Performances with Zee events
SEMReco algorithm
EMAnalyze Ntuple
Calibration Workshop Sep 25, M.Jaffré

2. Mrinmoy Bhattacharjee
L1 & L2 major challenges
Input Rate to L1 ~ 7.6MHz at 1032cm-2s-1
<pT> e- for J/ee <pT> e- for W,Z,top decays
~ 2.7 GeV/c central ~ 30-40GeV/c
~ 3.1 GeV/c forward Although threshold high
Low threshold in CAL high QCD rate S/B ~ 1/50,000
Calibration Workshop Sep 25, M.Jaffré

3. Definition of EM object (e-/)
Central Region
e = CFT trk + CPS cluster + CCEM Trigger Tower
 = No Trk + CPS cluster + CCEM Trigger Tower
Forward Region
e = MIP + FPS cluster + ECEM Trigger Tower
 = No MIP + FPS cluster + ECEM Trigger Tower
Calibration Workshop Sep 25, M.Jaffré

4. L1 electron algorithm (high pT)
Central (1.6)
1 CPS cluster  high matched to 1 track
5GeV/c in 4.50
1 Calo EM tower  7-10GeV (1.6)
matched to CPS cluster by quadrant
Forward (1.62.6)
1 FPS cluster  high + MIP confirmation
1 Calo EM tower  7-10GeV (1.62.6)
matched to FPS cluster by quadrant
Track/MIP matching to PS optional;
perform only if rates high
 1.0
Cal EM>7GeV
Calibration Workshop Sep 25, M.Jaffré

5. L1 EM trigger rates (high pT)
Rates at L=21032cm2/s
Cluster/Track & CCAL Quadrant match
Cluster & ECAL Quadrant match
CEM(1,10,C) W mass, QCD 
CEM(1,7,C)CEQ(1)TNQ(1) QCD 
CEM(1,10,C)TEL(1,5) W mass, WZ 
CEM(1,10,N/S) EC W mass
CEM(1,10,N/S)FQN(1) FWD EM
CEM(1,10,N/S)FPQ(1) EC W mass
Rates highly dependent on thresholds in CAL
Quadrant matching give 2-4 in rates rejection
1.0
Calibration Workshop Sep 25, M.Jaffré

6. L2 EM triggers Advantages at L2 are: (1) Large decision time 100sec
(2) Finer detector information available
-- clustering in PS
-- clustering in CAL
(3) Due to more time finer matching can be
performed
-- L1 CAL/PS matching in quadrant
-- L2 CAL/PS match within 0.20.2
in 
Calibration Workshop Sep 25, M.Jaffré

7. L2 CAL Preprocessor L1 EM Trigger based on following reference set
Marc Buehler
L2 CAL Preprocessor
L1 EM Trigger based on following reference set
2.5, 5, 7 & 10 GeV ET
(1) L2 uses TT’s above low threshold ref. set
(2) Find 2nd. Maximum in 33 around seed
(3) ETEM = ETEMseed + ET2nd > Thr
(4) EMF = ETEM/(ETEM+ETHAD)
(5) TISO =  ET(EM+Had)/ETEM ( 33  - seed )
= ETEM/ET(EM+HAD) (33  includes seed)
EMF>0.85
TISO < 0.4
PT =20GeV
Calibration Workshop Sep 25, M.Jaffré

8. L2 CAL Efficiency, Rates & Timing
L1 seed tower 7GeV / 10GeV
ETEM  10GeV / 12GeV
EMF  TISO  0.4
pT Gev/c GeV/c GeV/c
L2/L % % %
“ % % %
Dijet Rates
L1(1,7GeV) L2(1,10GeV) L2(1,12GeV)
@2E Hz Hz Hz
L2 CAL Timing (available 50sec)
L2 Seed cut 0.5GeV 1.0GeV 1.5GeV 2.0GeV
# of seeds
Time s
Calibration Workshop Sep 25, M.Jaffré

9. L2 electron algorithm (high pT)
Central (1.6) (L2 Global 50sec)
1 CAL tower 7GeV
1 CPS Axial cluster  5MIPs + CFT Track tag
CPS 3D match of X,U,V  , 
(1) Axial ,U, V = 5MIPs
(2) Axial = 5MIPs & U,V = 3MIPs
Different detector info combined at L2 Global
CPS-CAL match within 0.250.25 in 
(1) efficiency loss for ET <40GeV
(Z  ee, 93.7%)
(W+Jets, 84.0%)
(2) regains lost electrons at low ET
(Z  ee, 99.0%)
(W+Jets, 95.4%)
Calibration Workshop Sep 25, M.Jaffré

10. L2 electron algorithm (high pT)
Forward (1.62.6) (L2 Global 50sec)
CAL Cluster (ETEM)> 10GeV,
CAL EMF  CAL Isolation  0.4
FPS 3D match of U,V  , 
(1) U, V = high
(2) U = high & V = low OR vice versa
Different detector info combined at L2 Global
FPS, CAL matching within 0.250.25 in 
Z  ee 94.3% % (HH / HL.OR.LH)
@ 21032cm2/s
900Hz
145Hz
80Hz
Calibration Workshop Sep 25, M.Jaffré

11. B Physics: di-electron trigger
L1 & L2 Trigger Performance:
Central Region (CAL EM>2.5GeV, CPS>3MIPs)
eff (pT>1.5GeV) = 10%, Rates = 50Hz
Forward region (CAL EM>2.5GeV, FPS>5MIPs)
L1 eff (pT>1.5GeV) = 10%, Rates = KHz
L2 eff (pT>1.5GeV) = 4-5%, Rates = 50Hz
Calibration Workshop Sep 25, M.Jaffré

12. L1 & L2 summary
L1 Trigger
electron (pT>5GeV) efficiency > 95% (7-10GeV CAL)
background rates Hz (7GeV CAL thr)
Hz (10GeV CAL Thr)
di-e (pT>1.5GeV) efficiency ~ 20%(cen), 10%(fwd)
background rates kHz (cen+fwd)
PS/CAL Quadrant matching 2-3 in rates
L2 Trigger
electron (pT>5GeV) efficiency > 95% (10GeV ETEM)
background rates Hz (10GeV ETEM)
di-e (pT>1.5GeV) efficiency ~ 10%(cen),5%(fwd)
background rates Hz (cen+fwd)
CAL EMF, Isolation, Invariant mass helps
PS/CAL 0.250.25 in  match = 2-3 in rates
Calibration Workshop Sep 25, M.Jaffré

13. Volker Buescher Juan Estrada Vishnu Zutshi
L3 Components
Cal.Unpacker
Charles Leggett
CPS.Unpacker
Andre Turcot
FPS Unpack
U.Michigan
L3TPrVtx
G. Lima
Cal. Cluster
Volker Buescher
L3TCps
Andre Turcot
L3TFps
U.Michigan
L3TEle, L3TPhoton
Volker Buescher Juan Estrada Vishnu Zutshi
L3track framework
Daniel Whiteson

14. L3 Preshower Clustering
CPS ( A. Turcot )
a review of algorithm performed last summer
Algorithm : - contiguous strips above threshold (2 MIPs) clustered in
each layer ( Single Layer Cluster )
- 3D matching of SLC with some energy correlation
Localized clustering in  or - implemented to be used by L3 electron tool
CPS resolution for single e-
(r) = 0.54 mm, (z) = 1.9 mm
Calibration Workshop Sep 25, M.Jaffré

15. EMRECO input cluster Cluster « finding » algorithms
3 Clustering available in calreco based on Towers : simple cone, cone and emNN
Default now is simplecone R=0.4 (fast)
4th algorithm based on cells : CellNN
Selection of the clustering Since t (  preco05)
New parameter in EMReco.rcp
“standard” emreco RCP clusterer = <emreco CalScone0401>
cellNN clusters RCP clusterer = <cellNN CellNN>
Imply different algorithms for building the EMCluster
Both could be “called”  2 different Emparticle Chunks
Calibration Workshop Sep 25, M.Jaffré

16. EMRECO algorithms EMCluster from cellNN clusters
Serban Protopopescu
M.J.
EMRECO algorithms
EMCluster from cellNN clusters
Ptmin > 1.5 GeV ; e.m. fraction > 90%
Cluster in UNCHANGED
EMCluster for “standard” emreco
Isolation : (Etotal(cone 0.4) - Eem(cone0.2))/ Eem(cone0.2) < 0.2
EMCluster definition (EM1,EM2,EM3,EM4,FH1)
CC : all channels in a (3x3) window
EC : all channels in a cone of 10 cm radius in EM3 around hottest channel in EM3
RCP parameters
rather independent of clustering algorithm for isolated electrons
Calibration Workshop Sep 25, M.Jaffré

17. EMRECO (cont ’d) EMcluster position corrections (DooKee Cho)
CAL cell is given a weight : w=w0 + log(Ecell/Elayer)
w0 varies for each layer
Additional corrections depending on z and Pt (CC),
on r (EC)
EMcluster energy corrections (Serban)
CPS : x ECPS GeV
FPS : x EFPS
Overall correction :
E=E/(1.3 –SQRT( /E))
DE/E uncorr
CPS energy fraction
Calibration Workshop Sep 25, M.Jaffré

18. Cluster position resolutions
DooKee Cho
Calibration Workshop Sep 25, M.Jaffré

19. EMRECO (cont ’d) Preshower matching Track matching
input : PS 3D clusters from cal_reco and fps_reco
select the most energetic cluster within a window :
x = 0.05x0.05 CPS x = 0.1x0.1 FPS
Track matching
input : ChargedParticleChunk
P>1.5 GeV/c; x window = 0.05x0.05
Window sizes can be reduced without loss of efficiency
Emparticle direction and ID
1 - track (if it exists )direction at the vertex electron  12
2 - vertex - CPS cluster direction photon
3 - vertex - cluster position in EM photon
Calibration Workshop Sep 25, M.Jaffré

20. HMRECO status It uses the package hmatrix to store/read H-matrix files
F.Touze, A. Abdesselam,M.J.
(cvs) Package used by emreco to
- build different flavours of H-matrices
- calculate the covariance parameter 2 as a measure of the compatibility of the EMCluster object with an electron shower
The choice between 1 and 2 is done by a RCP parameter in emreco
 Emreco provides the same cluster preselection
It uses the package hmatrix to store/read H-matrix files
Calibration Workshop Sep 25, M.Jaffré

21. Available H-matrices in preco04
Run I H matrix (41 parameters)
Energy fractions in floors [EM1, EM2,EM4]
Energy fractions in floor EM3 (6x6) channels
log(Etot),Zvertex/28.
H matrix (9 parameters)
Energy fractions in floors [EPS,EM1, EM2, EM3,EM4]
DETA, DRPHI in floor EM3
log(Etot), Zvertex/28.
Exists for 2 energy ranges : < > 30 GeV(RCP parameter)
 2 2 values attached to each EM candidate
Calibration Workshop Sep 25, M.Jaffré

22. Latest results with Hmatrices
30GeV
Calibration Workshop Sep 25, M.Jaffré

23. EMReco Efficiencies in CC
F.Fleuret
CC fiducial region
CPS fiducial region
TRK fiducial region
+ opposite sign
86% (93% per e)
98% (99% per e)
10K Pythia Zee
<1.1>minBias
Preco
98% (99% per e)

24. EMReco Efficiencies in EC
F.Fleuret
EC fiducial region
FPS fiducial region
TRK fiducial region
+ opposite sign
67% (82% per e)
92% (96% per e)
10K Pythia Zee
<1.1>minBias
Preco
97% (99% per e)

25. Serban
Calibration Workshop Sep 25, M.Jaffré

26. Serban
Calibration Workshop Sep 25, M.Jaffré

27. Calibration Workshop Sep 25, 2000 M.Jaffré
Frédéric Fleuret

28. Conclusions EMReco provides a framework to identify electrons
There is some degree of flexibility
RCP parameters
Clustering choice
SEMReco exists for soft electrons
EMAnalyze/SEMAnalyze provides a Ntuple with blocks of data for each choice for an easy comparison
Need user feedback
Calibration Workshop Sep 25, M.Jaffré