EMCal showers in data and MC
Gustavo Conesa Balbastre
(extracted most of it from Nicolas Arbor thesis)
σ/E [%]
Erec/Einc
Electron energy resolution
• Good agreement data/ MC for E > 1 GeV
With APD poissonian fluctuations adjusted in
MC with pi0’s
Electron non-linearity :
• linear for E > 3 GeV
• different corrections data / MC
• a bit different performance G3/G4
See effect of clusterization,
fluctuations in back-up
Shower shape :
• clusters  ellipse
• λ0,λ1  ellipse axis weighted by the cells energy distribution
See formulae in back-up
Data
pp 7 TeV
λ02
25 GeV/c
6 GeV/c
Photons
prompts
π0
g+g
Eπ0
π0
g
• Beam test with electrons  shower shape
• Difference data/Monte Carlo from 6 GeV, increase with energy
MC studies
• Problem in the EM GEANT simulation?
• Many things can affect: Physical processes, energy cuts, fluctuations, reconstruction
biases …
Clusters have few more cells on average and energy is a bit more spread in data
compared to MC
Is the Geant transport cuts removing too much energy? Is the shower cut too soon?
Is the tower Molière radius properly considered?
10 GeV electrons
No significant difference in G3 and G4
G4 Physics list can affect,
• EMV list, worse description
• QGSP-BERT, and default have similar performance,
maybe a bit more tail, cut =0.01 mm disfavoured
vs 1 mm
Fluctuations seem to introduce the bias we look for?
-
No reco, add all
cells with signal
Clusterization
Add noise
Add APD
fluctuations
0.06
Data, LHC11c+d
g
0.05
MC, GJ+JJ p >7 GeV/c
0
T
0.05
0.04
0.03
0
T
0.06
12 < E < 14 GeV
1/integral d N / d l 2 ; Normalized to max
g
MC, GJ+JJ p >7 GeV/c
1/integral d N / d l 2 ; Normalized to max
Data, LHC11c+d
0.07
0
1/integral d N / d l 2 ; Normalized to max
10 < E < 12 GeV
0.08
0.04
0.03
0.02
0.06
14 < E < 16 GeV
Data, LHC11c+d
0.05
g
MC, GJ+JJ p >7 GeV/c
T
0.04
0.03
0.02
0.02
0.2
0.3
0.4
0.5
0.6
0
0.7
l2
0.1
0.2
0.3
0.4
0.5
0.6
g
MC, GJ+JJ p >7 GeV/c
18 < E < 20 GeV
0.06
Data, LHC11c+d
g
MC, GJ+JJ p >7 GeV/c
0.05
T
0
T
0.04
0.03
0.02
0.04
0.03
0.02
0.01
0
0.1
0.3
0.4
0.5
0.6
0
0.7
l2
0.1
0.2
0.3
0.4
0.5
0.6
MC, GJ+JJ p >7 GeV/c
30 < E < 40 GeV
0.12
Data, LHC11c+d
g
0.1
MC, GJ+JJ p >7 GeV/c
T
0
T
0.06
0.05
0.04
0.03
0.07
0.5
0.6
0.7
l2
20 < E < 25 GeV
Data, LHC11c+d
0.06
g
MC, GJ+JJ p >7 GeV/c
T
0.05
0.04
0.03
0
0.7
l2
0.1
0.2
0.3
0.4
0.5
0.6
0.08
0.06
0.7
l2
0
0
g
0.4
0
1/integral d N / d l 2 ; Normalized to max
Data, LHC11c+d
0.07
0
1/integral d N / d l 2 ; Normalized to max
25 < E < 30 GeV
0.08
0.3
40 < E < 60 GeV
0.14
Data, LHC11c+d
g
0.12
MC, GJ+JJ p >7 GeV/c
T
0.1
Photon region seems
more affected than
pi0 region.
0.06
0.04
0.04
0.02
0.02
0.01
0.1
0.2
0.3
0.4
0.5
0.6
0.7
l2
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
l2
0
Not quite the same
Distributions as in
beam test.
Similar effect
anyhow.
0.08
0.02
0
We are forced to
select photon
clusters in the blue
line limit to properly
calculate efficiency
0.01
0
0.09
0.2
0.02
0.01
0.2
0.1
0
0
Data, LHC11c+d
0.05
0
0.7
l2
0
1/integral d N / d l 2 ; Normalized to max
16 < E < 18 GeV
0.06
0
1/integral d N / d l 2 ; Normalized to max
0
1/integral d N / d l 2 ; Normalized to max
0.1
1/integral d N / d l 2 ; Normalized to max
0
0.01
0.01
0.01
MC: Add Gamma-jet
and jet-jet Pythia (jet
triggered by decay
photon in EMCal
acceptance with
pT>7 GeV)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
l2
0
• Probability that a photon hits some material and converts ~ 50% (in MC)
• At high E the e+e- are too close and form a single cluster, effect on shower shape
Plot: Clusters with photon as main contributor
- N_over: Number of other particles in
cluster
Squares Jet-Jet triggered MC,
pp@7TeV
Bullets: Single photon simulations
In reality:
Do we have more material?
Is hadron response under control?
These contributions and beam test
discrepancies could go together?
Is beam test contaminated by
hadrons and not rejected properly?
• A discrepancy in shower shape distributions and non linearity is
observed in data and MC
• Both in pp data and beam test, although discrepancies are not
the same
• No clear origin, going to G4 does not seem to solve the discrepancy
• Possible effects that could contribute
• APD gain fluctuations
• Material conversions
• Hadronique response of the calorimeter
• Geant settings?
•…
• Aim to set a task force but
• The usual man power issue
• Need some fresh ideas to test
• Any input is welcome
Back-up
Width
sp 0 (GeV/c2 )
mp 0 (GeV/c2 )
Mass
0.15
Data
0.145
0.03
Data
0.025
Simulation
Simulation
0.02
0.14
0.015
0.135
0.01
p0
p0
0.005
0.13
5
10
15
20
25
p (GeV/c)
T
0
5
10
15
20
25
p (GeV/c)
T
• Beam test with electrons  shower shape
• Difference data/Monte Carlo from 6 GeV, increase with energy
MC studies
• Problem in the EM GEANT simulation?
• Many things can affect: Physical processes, energy cuts, fluctuations, reconstruction
biases …
• Matrix eigen-values (cluster) :
l02 = 0.5* (d xx + dzz ) + (0.25* (d xx - dzz )2 + d xz2
l12 = 0.5* (d xx + dzz ) - (0.25 * (d xx - dzz )2 + d xz2
with dxx, dzz and dxz weighted by the tower energy
d xx = å
i
dzz = å
i
d xz = å
i
wi
w
(colcell (i ) )2 - (å i (colcell (i) ))2
wtot
i wtot
wi
w
(rowcell (i ) )2 - (å i (rowcell (i ) ))2
wtot
i wtot
wi
w
w
[(rowcell (i) )(colcell (i ) )] - å i (colcell (i ) ) * å i (rowcell (i ) )
wtot
i wtot
i wtot
w = TMath :: Max(0., w0 + log(Ecell / Ecluster ))
wtot = å wi , w0 = 4.5
i
1 particle cluster
2 particles cluster
10 GeV photon, Geant 4
Parametrize w0 in simulation to match data?
10 GeV, Geant 4