Pattern Recognition in the MEG Drift Chambers

1
Pattern Recognition in the MEG Drift Chambers

• Presented by
• Matthias Schneebeli
• on behalf of the MEG collaboration

2
Outline

• Introduction to the experiment
• Discussion of the pattern recognition algorithm
• Results of the pattern recognition

3
Decay topology
m ? e g
Main background m ? e nn g
g
Accidental
m
180º
radiative decay
e

• m ? e g signal very clean
• Eg Ee 52.8 MeV
• qge 180º
• e and g in time

Good energy resolution Good spatial
resolution Excellent timing resolution Good
pile-up rejection
4
The MEG Detector

• Solenoidal magnetic spectrometer
• LXe for efficient g detection
• Drift Chamber for Positron detection
• Timing Counter for time measurement

5
COBRA Magnet
Constant bending radius
Quick sweep-out

• Constant Bending RAdius(COBRA) positron
  spectrometer
• Special graded B field
• - Constant projected bending radius for
  monochromatic positron independent of
  emission angle
• -gt Easily define momentum window
• - Quick sweep-out of positron
• -gt Stable operation of chamber system in high
  rate muon beam

6
Drift chamber

• 16 radial chambers with two planes of wires
• Staggered cells measure both position and time
• He C2H6 gas to reduce multiple scattering
• Vernier pattern to determine z coordinate

Field Wire
Sense Wire
sz 1cm
sz 200mm
7
MC Hits
Cyan line mc track Blue line rec track
Red line mc signal track
Green circle mc hit circle Black circle rec
hit circle
8
MC Track
Cyan line mc track Blue line rec track
Red line mc signal track
Green circle mc hit circle Black circle rec
hit circle
9
Tasks

• Find all hits belonging to a track
• Calculate the drift time for each hit
• Fit a track through the hits

t2
t1, t2, t3, t4 measured tdi ti t3 tx
drift time of earliest hit tdrift_i tdi tx
td2
t4
tx
td4
tx
tx
td1
tx
t1
t3
td30
10
PatRec Algorithm

• Kalman filter
• Implemented but too slow (1 Hz)
• Only used to fit pre-selected hits
• Customized PatRec algorithm
• no B-Field
• no material
• List of hits belonging to a track
• Drift distance for each hit
• Track seed for kalman filter

11
Find Clusters

• Combines every 2 hits in a chamber which fulfill
• Hits in same plane
• ?z lt 0.2cm z0.02cm
• cell_1-cell_2 1
• Hits in different plane
• ?z lt 0.6cm z0.046cm
• cell_1-cell_2 2

12
Find Track Seed

• Find a track seed
• Find 3 clusters in consecutive chambers which
  fulfill
• z-zproj lt 0.6cm z0.046cm

zproj z1 ?z01l12/l01
13
Calculate Drift Time
Find drift time

• Find smallest hit time of hits in the seed,
  define as t0

• Calculate a track circle by using wire
  position/xy hit points

• Calculate 2 xy points for each hit using t-xy
  relation.

• Choose best combination of xy hit points and
  calculate deviation from track

• Decrease t0 and repeat steps ii to v until the
  minimal deviation is reached

t0
t1
radius t1-t0
14
Extrapolate Track

• Extrapolate track
• Look in the next chambers for additional clusters
  or single hits which fulfill
• zproj z lt 0.6cm z 0.053cm
• Calculate a new track circle including new
  cluster/hit
• Check deviation ??r lt 0.4 cm
• Add additional hits
• Add hits to clusters for tracks which made more
  than 2 hits in a chamber

?r1
?r0
15
Reconstructed Track
Cyan line mc track Blue line rec track
Red line mc signal track
Green circle mc hit circle Black circle rec
hit circle
16
Reconstructed Track
Cyan line mc track Blue line rec track
Red line mc signal track
Green circle mc hit circle Black circle rec
hit circle
17
Reconstructed Track
Cyan line mc track Blue line rec track
Red line mc signal track
Green circle mc hit circle Black circle rec
hit circle
18
Reconstructed Track
Cyan line mc track Blue line rec track
Red line mc signal track
Green circle mc hit circle Black circle rec
hit circle
19
Performance

• Compared hits found during
• track finding with mc hits and
• mc tracks.
• Statistics for signal tracks
• correctly found 96
• 1 hit missed 1
• 2 hit missed 2
• not found 1

Error in drift time calculation 4.3ns
10 times faster than kalman filter
20
Summary

• We presented a pattern recognition algorithm for
  the MEG drift chambers
• Only MC studies
• Will be tested on the beam data at the end of
  this year