Improvement%20on%20vertex%20reconstruction

1
Improvement on vertex reconstruction
S. Moriyama 6th March 2004 XMASS meeting

• Motivations
• Many peaks in vertex distributions
• Further study for the lt0.1MeV events

Nambas plots
2
Problem in the vertex reconstruction

• We are using maps which have value on a 2.5cm
  spacing grid. ? F(ix,iy,iz,ith PMT)
• Between grid points, F(probability to observe
  photoelectron by ith PMT) is obtained by a linear
  interpolation. ? F(x,y,z, ith PMT)
• Namba pointed out that there must be
  discontinuity of dF/dx, etc, on the grid edge.
• When I saw likelihood distribution, there is a
  valley on the grid edge due to the dF/dx
  discontinuity.

3
How to solve?
One example
Likelihood
Zoom up
Likelihood
Actual likelihood
Best fit here
2nd order polynomial fit
Concentration to the grid is now understood.
Z(cm)
Z(cm)

1. Adopt higher order interpolation for the maps. ?
   Technically complex for me.
2. Slightly modify likelihood distribution near the
   current best fit point. ? looks good.

4
Principle of the improvement

• For the events wallgt2cm, vertex improvement code
  is called.
• In the code, likelihood along x, y, and z axis is
  calculated by 1mm spacing for /- 1cm region.
• For each likelihood, 2nd order ploynomial fit is
  performed based on the 1mm spacing data, and
  adopt minimum value of the 2nd order polynomial
  as a new fitted result.
• No modification for energy information
• (/home/xmass/src/reconst_v6.0)

5
Effect on the source run analysis (Namba)
MC
Data
Improvement succeeded.
Further improvement needs maps with more fine
spacing.
6
Performance check revisited

• Minamino?

7
Background data and MC
Almost no change observed

• Note There was a bug in MC treatment after
  applying reconst_v4.4 (improvement on lt0.1MeV
  analysis), but there is almost no change by the
  bug fix.

8
How about lt0.1MeV events?

• Detail check of wall distribution
• Horizontal axis (15.5-wall)3 ? uniform volume
  plots can be seen.
• 0.6kg/bin

9
Background distributions
Data MC
0.4-0.5MeV
0.5-1MeV
0.1-0.2MeV
0-0.1MeV
0.2-0.3MeV
0.3-0.4MeV
1-2MeV
2-3MeV

• Horizontal cm3 (125cm310cmFV, 1000cm320cmFV)
• Vertical DRU

10
0-0.1MeV
0.1-0.2MeV
0.2-0.3MeV
0.3-0.4MeV
11
Summary

• Improvement on the vertex reconstruction done.
• Still there is small grid effect. It will be
  improved by making fine maps.
• lt0.1MeV problem and Kr like signals seems to be
  related. Further study is needed why event
  concentration at the detector center occurs.