ECAL: Analysis and Test Beam results

1
ECAL Analysis and Test Beam results

• Marco Incagli - INFN Pisa
• on behalf of the ECAL group
• IHEP Beijing
• INFN Pisa Univ. of Siena
• LAPP Annecy GAM Montpellier

2
Results on

• Calibrating with MIPs
• Linearity, energy and angular resolution
• Comparison with MC
• Problems in calibration with MIPs
• Results on e/p separation
• The plots presented summarize the work of people
  from all the institutions which contribute to the
  ECAL construction

3
Main goals of ECAL
g
e, p ?

• e/p discrimination at 10-3 level, using ECAL
  standalone, and 10-4 level, using E-p matching
• Detection of non-interacting photons (75)
• Trigger on photons and on e which deposit energy
  in Veto Counters

Electromagnetic Calorimeter (ECAL)
4
Construction technic
e?
Lead foil (1mm)

• Sampling calorimeter with lead foils and
  scintillating fibers
• Pile up of 9 superlayers with fibers which run in
  x or y direction

Fibers (?1mm)
z
particle direction
18.5mm
y
x
5
Supporting structure and readout

• The active part is inserted in a supporting
  structure which connects ECAL to ISS
• Light is readout with PMTs connected to fibers
  through light guides
• PMTs have 4 active cahodes 9?9 mm2

18.5 mm
6
Granularity

• The key to separate e/p is to sample the em
  cascade with fine longitudinal and lateral
  granularity
• 18 longitudinal samplings (9 superlayers 18
  layers)
• Readout cell dimensions 1X0 ? 0.5 rM

superlayer
layer

• Granularity 0.9 cm ? 0.9 cm
• 0.5 Moliere radius in X-Y
• 1 X0 in Z

7
Data used

• Test Beam data of summer 2002
• 105 clean (tagged) 120GeV protons
• 5104 clean (tagged) electrons in energy range
  6-120 GeV
• MC events generated April 2003

Active part
8
Calibrating with MIPs

• Look for MIPs in 120GeV proton events (cut on
  number of hit cells and energy deposited)
• Subtract pedestal and fit Landau distribution
• The average MIP peak is at 7.0 ADC channels (MIP
  deposit ?8MeV)

m 7.0 s 1.4
9
Spread of MIPs

• The MIP peak is very sensitive to the details of
  the fit and to pedestal subtraction ? a
  systematic error of 5 will be included

MIP values found by two independent analysis
groups
MIP peak - group 2
MIP peak - group 1
10
Longitudinal profile - Leakage

• Above 10GeV some energy is loss due to
  longitudinal leakage
• Two technics have been tested to correct for this
  effect

• Fit of longitudinal profile event by event
• Last layer method (suggested by Choutko) ? next
  slide

Active part
leakage
11
Last Layer method

• The energy loss due to leakage depends only on
  the energy deposited in the LAST layer of the
  calorimeter

EBEAM beam energy EVIS visible energy ELAST
energy deposited in last layer
EVIS/EBEAM
a -3.65 b 1.035 (should be 1)
ELAST/EVIS
12
Linearity
1
-1

• ? Leakage correction
• ? No leakage correction

? Leakage correction ? No leakage correction

• Linearity within 1 up to 120GeV.

13
Energy resolution

• Fitting the points up to EBEAM80GeV we obtain
• The errors are dominated by fluctuations in the
  MIP calibration
• Errors are strongly correlated

fit
14
Angular resolution

• Fit of shower CoG on each layer, excluding the
  first two

50 GeV
??68
Reconstructed q (deg)
15
Angular resolution
??68 (7.8 0.2)º/?E ? (0.8 0.1)º

• The 3D angular resolution is lt2.5o for energies
  gt10GeV
• It improves for tracks impinging with an angle ?0

16
X0 determination

• Fit of the longitudinal shape with
• Total ECAL thickness

tMAX (layer units)
X0 (1.05 0.05) layer units (9.7 0.5)
mm X0(MC) 9.75 mm
17
Comparison data/MC

• 120 GeV protons
• data/MC normalized to same number of events

Longitudinal profile
Total energy deposit
18
Data/MC - electrons
6 GeV
10 GeV
Longitudinal profile
Total energy deposit
19
Data/MC - electrons
50 GeV
120 GeV
Longitudinal profile
Total energy deposit
20
Data/MC - comments

• Disagreement data/MC
• The disagreement depends upon layer number
• The observed longitudinal profile shows some
  strange (step like) behaviour, in particular in
  the last layers
• Maybe some problems in calibrating with MIPs ?

120GeV
21
Calibrating with MIPs
Extended showers
MIPs
E0
E1lt E0

• The presence of defects in the construction is
  NOT observed in MIPs, but it affects the electron
  showers
• This is more evident in last layers where we put
  the worst quality planes

22
Photomultiplier with 4 cathods Cell dimensions
Last superlayer
23
Calibrate with MC

• Assume X0(MC) - justified by data
• Correct the ADC count of layer j in order to
  reproduce the MC shape
• The correction factors are evaluated at
  Eele120GeV and used at all energies

Deposited energy (GeV)
120GeV
Layer number
24
Calibrate with MC

• Using the new calibration the step behaviour
  disappears

Energy deposit (GeV)
50GeV
80GeV
Layer number
25
Calibrating with MC

• and part of non linearity is recovered

180GeV
180GeV
After 172.5 GeV
Before 170.4 GeV
26
Conclusion 1

• The calibration procedure cannot rely on MIPs
  only
• A correction to the calibration constants using
  electrons is under study
• The difference with MC is, at least in part, due
  to calibration problems and to ECAL non
  uniformities
• In the new ECAL (flight model) particular care
  will be put in uniformity
• The machining of the lead foils is in progress
  14 superlayers will be built and the best 9 used

27
e/p rejection

• Test beam data have been used to study of e/p
  rejection
• Results are still preliminary also due to the
  aforementioned problems in calibration
• Two metodologies have been investigated
• Neural network (AMS note 2003-07-02)
• Statistical analysis (AMS note 2003-01-01
  updated in phd thesis of P.Maestro)

28
e/p neural network

• Entry points are discriminant variables
• Energy 2 cm around the shower axis
• Energy in the last three superlayers
  
• Variance for hit cells position
• Energy in the first five superlayers
• Thrust
• Neurons are linked by weights (w) determined by
  training
• Exit is a number between -1 (pure proton) and 1
  (pure photon)

Neural network feed-forward
iii
w
ij
i
w
jk



Entry
Layers Exit
29
Test Beam results

• Half of the collected data sample has been used
  to train the net
• Protons misidentified as positrons leave half
  of their energy in ECAL (?E/p0.5)
• ee 0.95 , ep 0.0052
• R ee/ep 183

50GeV electrons 120GeV protons
95 of e-
5.2 of p
30
e/p Fishers discriminant

• Analysis based on a MIP hunter algorithm
  (preselection) and 4 discriminant variables
• Energy fraction in 2 Moliere radii around shower
  axis
• Shower collimation (thrust)
• Area of shower projection on (x,y) plane
• Position of shower maximum
• The discriminant variables are combined by using
  Fishers Linear Discriminant analysis

31
Results from TB data
electrons
electron efficiency
cut
Fisher discriminant
120 GeV protons
50
Deposited energy
Deposited energy

• Tested with electrons of 6-120 GeV and 120GeV
  protons
• Electron efficiency 88.5 flat with energy
• ep0.62 (no cut on proton energy) ? R ee/ep
  143

32
Energy match

• The background to positrons of energy Ee comes
  mostly from protons of energy Ep2?Ee
• (e.g. 75 of 120GeV protons identified as
  electrons have visible energy 40ltEplt60 GeV 20
  have visible energy 20ltEplt40 GeV)
• Given the cosmic rays falling spectrum, the
  Rejection Factor R is further increased by a
  factor (Ep/Ee) 22.787
• R 150?7 103

33
Energy/momentum match

• Tracker information can be combined with ECAL to
  determine the E/p (energy/momentum) ratio
• For TB data a dummy momentum p is associated

• to both proton and electron by smearing the
  beam momentum with the tracker resolution
  estimated by MC.
• The resulting proton efficiency becomes 0.62 ?
  0.033 (a factor of 20 gain) applying the cut
  E/pgt0.8

34
Conclusions - I

• Performances of ECAL in agreement with design
• s(e)/E 10.6/sqrt(E) 2.6
• s(q) 7.8deg / sqrt(E) 0.8deg Syst error
  5
• X0 9.7mm ? XECAL 166.5mm 17.2 X0
• ECAL general characteristics reasonably well
  described by MC (X0, s(E)/E, s(q),)
• Detailed shower characteristics not well
  described because (at least in part) of
  inhomogenities in ECAL construction ? MIP
  calibration must be integrated with electrons?
  MC?

35
Conclusions - II

• Discrimination e-p studied with Test Beam data
• A rejection factor 200 can be reached by looking
  at shower characteristics (studied with data)
• A factor of 7 is gained with Energy match
• A factor of 20 is gained with energy-momentum
  matching (studied with data at 120GeV)
• The analysis can be (and has been) repeated with
  MC, but it relies on detailed shower propagation
• New Test Beam in 2004 with flight module will be
  necessary for fine tuning MC