Physics Analysis

1
Physics Analysis

• W. Ootani
• on behalf of the physics analysis working group
• Feb. 18th, 2009

2
Outline

• Physics Analysis in MEG
• Preselection and Blinding
• Radiative Decay Analysis
• µ?e? Analysis
• Summary

3
Physics Analysis in MEG
4
Physics Analysis in MEG
Analysis Scheme

• Physics Analysis Working Group
• Coordination F.Cei/W.Ootani
• Preselection/Blinding
• Coordinator R.Sawada
• LXe R.Sawada/G.Signorelli
• TC P.Cattaneo/G.Cavoto
• DC H.Nishiguchi/W.Molzon
• Normalization
• P.-R.Kettle/H.Nishiguchi
• Beam intensity optimization
• F.Cei/T.Iwamoto
• MC production
• G.Cavoto
• Analysis
• Everybody

We are here!
5
(No Transcript)
6
Radiative Decay Analysis
7
Radiative Decay Analysis

• Its quite important to identify radiative decay
  (RD) events, in order to demonstrate the quality
  of our experiment.
• Time calibration with real coincident events
• Two types of data samples for RD study
• Dedicated RD runs
• Ultra-low beam intensity (1.2x106 µ/s)
• Low accidental BG (better S/N)
• live time 4.88x105 s
• Lower energy threshold, no back-to-back
  requirement
• µ?e? runs
• Normal beam intensity (2.8x107 µ/s)
• Higher accidental BG
• live time 1.7x106 s (1st half of data 2008)
• Higher energy threshold, back-to-back requirement

8
Dedicated RD Run

• Data sample
• Run 23017-39963
• Total live time 4.88x105 s
• Selection criteria
• Geometrical cuts, track quality cuts, time and
  energy cuts
• Kinematical constraint greatly improved S/N
  (0.83?2.8)
• Found 428 RD events on 152 BG events in 2.5s
  (S/N2.8)
• Peak width 28718 ps (s)

Effect of kinematical constraint
Kinematical Constraint
9
Comparison with Expectations

• E? spectrum shape is well reproduced by MC.
• Angular dependence of observed number of events
  is in agreement with the expectations.

Angular dependence of of events
E? spectrum for selected events (No absolute
normalization)
10
RD in µ?e? Run

• Data sample
• Run 24002-31989 (first half)
• total live time 1.7x106 s
• Selection criteria
• Geometrical cuts, track quality cuts, time and
  energy cuts, kinematical constraint
• 608 signal events 1796 BG events in 2.5s
  (S/N0.34)
• Peak width 17829 ps (s)
• If we select E? gt40MeV,
• Peak width 11430ps (s)
• Well centered at T 0

Blinding box edge
11
RD Analysis Summary

• We observed a clear peak from radiative decay
  events in both dedicated RD run at lower beam
  intensity and µ?e? run.
• E? spectrum shape and angular dependence of the
  number of events are in agreement with the
  expectations.
• Time resolution seems improved as the gamma
  energy increases.
• Obtained time resolution (s150ps) for
  high-energy gamma still has large uncertainty,
  but it is close to the combined resolution
  between each detector.

Data E? cut Peak width (s)
RD 25 lt E? lt 40MeV 26319 ps
RD 20 lt E? lt 55MeV 28718 ps
µ?e? (1st half) 25 lt E? lt 45MeV 17829 ps
µ?e? (1st half) 40 lt E? 11430 ps
suffering large time drift (500ps)
12
µ?e? Analysis
13
µ?e? Analysis

• We started µ?e? analysis with blinding box
• Testing likelihood analysis
• BG study
• Data sample
• Run 24002-40997 (all µ?e? trigger data for
  2008)
• Live time 3.44x106 s
• Analysis region for likelihood fit
• Z? lt 26cm, F? lt 60deg, 2cmlt depth, Xvertex
  lt 3.5cm, Yvertex lt 3.3cm, ZVertex lt 9.4cm,
  Fe lt 60deg, cosTe lt 0.342
• 46 lt E? lt 60MeV, 50 lt Ee lt 56MeV, Te? lt 2.1ns,
  -1 lt cos?e? lt -0.9982
• Additional cuts
• XEC pileup rejection Npeakld 1
• XEC cosmic ray veto Qinner / Qouter gt 0.4
• DCH tracking quality cut ?2/DOF lt 20
• Selection criteria still to be optimized

14
Likelihood Analysis

• Likelihood function used in the analysis
• Partial probability to measure i-th event with
  observable Xi
• P(Xi) (Nsignal S(Xi) NRD S(Xi) NBG B(Xi))
  / (Nsignal NRD NBG)
• Likelihood function
• L(Nsignal, NRD, NBG) (NobservedN
  exp(-Nobserved) / N!) ?P(Xi)
• Observable Xi
• Xi (E?, u, v, w, Ee, cos?e, Fe, x, y, z, Te?)
• Describe non-uniform detector response
• PDFs
• Signal S(Xi) from measured response functions.
• RD S(Xi) from theoretical spectrum convoluted
  with response functions
• BG B(Xi) from measured BG spectra

15
BG E? Distribution in µ?e? Run

• Run24002-31989
• µ?e? trigger (0)
• Z? lt 23cm, F? lt 60deg, 2cmlt depth
• Te? gt 1.5ns (outside blinding box)
• Additional cuts
• Pileup rejection Npeakld 1
• Cosmic-ray veto Qinner / Qouter gt 0.4
• BG E? PDF was formed from the BG distribution
  (blue).
• Selection criteria still to be optimized.
• Uncertainty of energy scale

Geometrical cut Pileup rejection CR rejection
E? GeV
16
BG Ee Distribution in µ?e? Run

• Run24002-31989
• µ?e? trigger (0)
• Geometrical and track quality cuts
• Xvertex lt 3.5cm, Yvertex lt 3.3cm, Zvertex lt
  9.4cm, ?2/DOF lt20
• Te? gt 1.5ns (outside blinding box)
• Fe lt 60deg, cosTe lt 0.342
• Nhits gt 7, T0 lt 30ns, sp2 lt0.25(MeV/c)2
• Tighter quality cuts which give better resolution
  but lower efficiency
• BG Ee PDF was formed from the BG distribution
  (green).
• Selection criteria still to be optimized.

Ee GeV
17
(No Transcript)
18
(No Transcript)
19
BG Study
? single BG in µ?e? run

• We study BG using
• Single-BG trigger mixed with µ?e? trigger
• Side-band data outside blinding box
• ? single BG
• XEC self trigger (10, run24082-26503)
• Live time 8.4s
• Uncertainty in energy-scale
• e single BG
• TC self trigger (22, run24082-26534)
• Live time 0.0425s
• BG rate estimated at side-band is consistent with
  each single-BG rate.
• BG is predominantly accidentals.

e single BG in µ?e? run
BG distribution at side-band
20
Summary

• Finally we got the data!
• We started the physics analysis although the
  calibrations and resolutions are still being
  optimized.
• The tools to process and to analyze the physics
  data were developed and are working fine.
• Preselection/blinding
• Likelihood analysis, confidence level
  calculation,...
• We observed a clear peak from RD events both in
  dedicated RD runs and µ?e? runs.
• Time resolution of 150ps (s) (still measured at
  lower energy than signal energy)
• Spectrum shape and angular dependence of number
  of events are in agreement with the expectations.
  
• We started the µ?e? analysis with the blinding
  box.
• BG rate seems consistent with the assumption that
  they are accidentals.
• We still need to understand the BG improving the
  calibrations and resolutions.

21
(No Transcript)
22

• End of Slides

23
Physics Run 2008

• Sep. 12th - Dec. 16th (12 weeks)
• µ?e? trigger run
• Mixed with various pre-scaled triggers for
  calibration data
• PMT calibration every day
• Full set of calibration runs on Mon. Wed. and
  Fri. ( a few hours)
• Radiative decay run at reduced beam intensity
  once a week (1 day)

24
Preselection and Blinding

• Windows large enough to contain signal events
  even with preliminary detector calibration
• Preselection
• Selection with loose criteria for data reduction
  and efficient analysis
• Criteria
• DC track(s) with TC hit
• TDC - TTC lt 50 ns
• -6.9 ns lt T? - TTC lt 4.4ns
• Blinding
• Hidden signal box on (E?, Te?)
• 48 lt E? lt 57.6MeV
• T? - Te lt 1.5ns

Comparison of data size
25
What to Improve Before Opening the Box?

• The numbers of the detector performance quoted
  here are not final.
• Calibrations and resolutions are still being
  improved.
• We are trying to understand systematic
  uncertainty on some parameters.
• Some are hoped to be improved even in 2008 data.
• ? energy (s 2.3 should be compared to 1.2 in
  the large prototype)
• Improve uncertainty on energy-scale,
  no-uniformity, stability over time
• No depth correction yet
• We got 1.6 at the best position, which is
  consistent with 1.2 in the prototype if scaled
  with the light yield.
• ? position (s 6.5mm compared to s 4.3mm in
  the large prototype)
• s 6.5mm was obtained at the collimator slit
  (see XEC analysis talk), whereas s 5mm was
  obtained at the collimator edge.
• Improve PMT equalization (QE, gain)

26
What to Improve Before Opening the Box?

• ? timing (s 100ps compared to 53ps in the large
  prototype)
• Correction of time drift
• Positron
• DC tripping problem
• Efficiency
• DC tripping problem
• Improve the reconstruction algorithm for the
  acceptance edge events at the LXe detector
• Many others...
• Improvement expected for run 2009
• DC tripping problem to be solved.
• New waveform digitizer chip (DRS4)
• Further improvement of light yield of LXe?
• Still 30 lower than the light yield in the large
  prototype