reconstruction issues in cosmic ray muons maury goodman/gavril giurgiu & jurgen reichenbacher
TRANSCRIPT
Reconstruction Issuesin Cosmic Ray Muons
Maury Goodman/Gavril Giurgiu & Jurgen Reichenbacher
January 6, 2006 Maury Goodman 2
• Monte Carlo Distributions• Chi-squared distributions• Release 18/14 comparisons• Hook events in releases 1.14, 1.16 and
1.18• Maximum analyzable momentum• Initial look at Cambridge reconstruction• Our concerns about publishing 1099
January 6, 2006 Maury Goodman 3
R1.14 Monte Carlo, NO CUTS(Problems are more evident before cuts are made)
pFit(GeV)
Cha
rge
ratio
Forward
Reverse
0.5(Forward+Reverse)
pFit(GeV)
January 6, 2006 Maury Goodman 4
Reconstructed/True charge ratioversus reconstructed/true p in MC
pTruth(GeV) pTruth(GeV)
Tru
e C
harg
e R
atio
pTruth(GeV) pTruth(GeV)
Rec
onst
ruct
ed C
harg
e R
atio
pFit(GeV) pFit(GeV)
Tru
e C
harg
e ra
tio
pFit(GeV)
Cha
rge
ratio
Forward
Reverse
pFit(GeV)
Low p dip
High p randomization
January 6, 2006 Maury Goodman 5
• The low energy dip appears in the Monte Carlo and is a problem with high momentum tracks being reconstructed at low momentum.
• The problem is worse in the data than the Monte Carlo.
• (In addition, there are some differences between forward and reverse MC we don’t fully understand.)
January 6, 2006 Maury Goodman 6
Monte CarloAlec cut applied: (q/p)/( Sigmaq/p) > 2.5
pFit(GeV) pFit(GeV) pFit(GeV)
pFit(GeV) pFit(GeV)
pTru
e(G
eV)
pTru
e(G
eV)
pTru
e(G
eV)
pTru
e(G
eV)
pTru
e(G
eV)
No plane cut 20 planes 60 planes
100 planes 150 planes All plots -> Log z scale -> Vertical tails smaller than they seem
High momentum tracks with few planes are reconstructed as low mometum tracks and poor Chi2/ndof ?
pFit(GeV)
Mea
n C
hi2/
ndof
pFit(GeV)M
ean
Chi
2/nd
of
NO CUTS All cuts + 20 plane cut
Bad Chi2 for low reconstructed momentum
Monte Carlo
Red = positive muons Black = negative muons
January 6, 2006 Maury Goodman 8
Release 14/18 comparisons18 data
18.2
MC 14
14 data
2
p(GeV)
Before Alec cut After Alec cut
Black 1.14
Blue 1.18
momentum(GeV)
momentum(GeV)
char
ge r
atio
char
ge r
atio
20 plane cut
60 plane cut
R1.14 data
Black: no momentum correction
Red: with momentum correction q/p -> q/p (1.01 - 0.1 q/p)
Last minute slide
“tweak”
January 6, 2006 Maury Goodman 10
New plot 14 vs 18Red 1.14 tweak added
Blue 1.18 has tweak by default
Black 1.14 (no tweak)
January 6, 2006 Maury Goodman 11
Scanning
• It would be desirable to have a display package that showed the hits, showed the strips, and showed the fit, that we could use from or at ANL.
• Why does the on-line display package show such unphysical behavior in the x-y plane?
January 6, 2006 Maury Goodman 12
Event from Release 16; hit outside the detector !#@
January 6, 2006 Maury Goodman 13
Rel 1.14
Looks like used a wrong hit (from a brem?) near end of track
January 6, 2006 Maury Goodman 14
Looks like used a wrong hit (from ???) near end of track)
January 6, 2006 Maury Goodman 15
Fit to a 3 GeV muon track with a 7 GeV shower
Does not fail PhUSE cut.phUse = 0.577406planeUse = 0.695652stripUse = 0.223301digitUse = 0.295302
January 6, 2006 Maury Goodman 16
Cambridge Reconstruction
• Recently John Marshall made available standard ntuples with a new Kalman fitter
• Only reverse field data exists so far.
• We need to also reconstruct forward field and Monte Carlo to see if the same problems exist or not.
standard Cambridge
January 6, 2006 Maury Goodman 17
Maximum Analyzable Momentum
• For a muon going through good field, multiple scattering is ~15% of (1/p)/(1/p)
• The sagitta for a track with 0 impact parameter is 8 times that for 3.5 m impact
• Therefore, for muons which “skim” the detector, multiple scattering provides more curvature than bending.
• This can be quantified as a function of track direction.
January 6, 2006 Maury Goodman 18
MAM (2)
• For an infinite cylindrical detector, it should depend only on impact parameter and d(cosz)
• For a finite detector, it also depends on L• Maury, Phil and Tom are all trying to work this
out analytically and then apply to the detector• This may be part of what the 3.5m 60 plane cut
is accomplishing, but it could be made quantitative.
January 6, 2006 Maury Goodman 19
Concerns about publishing nowUsing MIC (which might/might not solve the problem)
• It is a recent circumstance (November 2005) that two separate effects have been identified in the charge ratio analysis: the low energy dip (a charge ratio randomization effect) and previously identified bias(es). We now have a good opportunity to solve the former problem in a more robust analysis, particularly since it appears in the Monte Carlo.
• It is plausible that the MIC reduces the possibility that hook events contribute to the charge ratio. Reconstruction problems don’t appear to be the motivation for the MIC
• The analysis reported in note 1099 cannot be repeated in release 1.18, and the reasons for this are not clear. There is strong evidence that some reconstruction issues are much worse in release 1.18 than in release 1.14, with or without the MIC. (maybe the tweak??)
• Systematic errors should be (and are) being calculated.
January 6, 2006 Maury Goodman 20
Next Steps
• Set up scanning in the Monte Carlo
• Pursue MAM calculation/cut
• Figure out what the problem with Release 18 is.
• Finish looking at Cambridge reconstruction.