Brief Description

1
Performance of least-squares based muon fitter
(FitTrack3)Jenny Thomas for Ryan Nichol

• Brief Description
• Performance on Monte Carlo
• Comparison To CandFitTrackSR
• Summary and Future Plans

2
Brief Description

• Trying to measure muon momentum from curvature.
• Based on reco_muon fortran package.
• Minimise ?2 with respect to 5 parameters (initial
  momentum, position and direction.) and iterate
  until convergence.
• Multiple scattering is taken into account in the
  correlated weight matrix.
• See NUMI-L-301 and NUMI-L-536
• Throw away ambiguous events to avoid miss
  identification.

3
Example Fit To Cosmic Muon (MC)

• Real Momentum
• -7.82GeV
• Fitted Momentum
• -7.68 GeV
• Raw Hits in blue
• Swam Hits in pink

4
Performance With Cosmic MC

• Initial studies with sub 10Gev cosmic muon monte
  carlo show
• 97.3 correct charge identification.
• But only fit 18.7 of events.

5
True And Fitted Spectra

• The momentum spectra from the fit is more spread
  out than the spectra from truth.

6
Angular Distribution Of Events

• There is a reasonable angular distribution for
  events which pass the fitter.

7
Performance of FitTrackSR

• Using the same monte carlo with the standard
  reconstruction we have.
• 18.4 incorrect charge determination
• 88 of events are fitted.

8
More on FitTrackSR

• If we only consider those events which pass the
  tighter cuts required for the new fitter we find.
• Only 3.5 incorrect charge determination.

9
Even More On FitTrackSR

• If we consider only those events which are not
  fitted by the new package we now have.
• 33.6 incorrect charge determination.

10
Comparing FitTrack3 to FitTrackSR

• There is good agreement between the two fitting
  packages.

11
Comparison With Truth

• Generally good agreement.
• LS fit method has at present charge bias in the
  mis-identification.
• Due to initial guess
• SR fitter more symmetric in the mis-
  identification.

12
Comparison With Truth (II)

• If we look at the events which do not pass one or
  other of the fitters.
• Very few missed by SR
• Lots of charge mis-identification in those events
  ignored by LS fit.

13
Summary (and Future Work)

• New package designed to limit occurrences of
  mis-identification.
• Only 2.7 incorrectly assigned in lt10GeV cosmic
  sample.
• Comparable results to standard reconstruction on
  sample good events.
• Things to do.
• Increase 18.6 acceptance.
• Look at charge bias in mis-identification.
• Fix bug in range calculation.