Shower recognition algorithm

1
Shower recognition algorithm For LCAL
LCAL-group K. Afanaciev, V. Drugakov, E.
Kouznetsova, W. Lohmann, A. Stahl
LUMI Workshop, Zeuthen November 14, 2002
2

• Sandwich LCAL
• geometry

• Z - Segmentation
• Tungsten 3.5 mm
• Layer
• Diamond 0.5 mm

• (R,f) - Segmentation

12 Radial Layers Square cells of about 0.5?0.5 cm
3
GUINEAPIG BRAHMS ( for vs 500 GeV )

• Per bunch crossing
• 15000 e hits
• 20 TeV of total deposited energy
• (x,y)-distribution of the beamstrahlung energy

Background averaged for 500 bunch crossings
The bulk of energy is deposited in the inner
region (radial layers 1, 2 and 3)
4

• Energy distribution
• in background

Total number of particles corresponds to 500
bunch crossings
Most particles have energy of up to few GeV A few
particles have energy greater than 50 GeV.
5

• 250 GeV particle

• Energy deposition by 250 GeV e-

Total energy deposited by 250 GeV electron is
about 30 GeV
6

• Sandwich LCAL
• background

• Average background for 10 bunchcrossings
• Longitudinal energy deposition profiles are a bit
  different for particle and background
• Distribution of energy deposition of BG defines
  good and bad regions

250 GeV e- BG
7

• Particle recognition
• algorithm

• 1. Calculate average background and its RMS
• 2. Subtract average BG from data
• 3. Compare result with 3sBG (RMS) (only for
  long. layers 4 - 17)
• 4. Find columns with gt 10 (of 14) such cells

• 5. Check neighbor columns to contain at least
  7 suspected cells

8

• Why 3 ? ?

• Number of recognized particles with 2? and 3?
  threshold
• (100 real particles of 250 GeV)

• 2 ?

• 3 ?

9

• Fake rate due to
• high energetic background

• Fake rate
• (BG of high energy BG fluctuations)

( 500 bunchcrossings )
10

• Efficiency

• Efficiency vs radius

11

• Energy resolution

• Energy resolution vs radius

12

• Calibration curve

• Calibration curve for different conditions.

• Though quite simple, this algorithm provides good
  efficiency and energy resolution.
• We should find a way to improve its performance
  in bad regions.