CPSReco Deghosting

1
CPSReco Deghosting
John Gardner University of Kansas Feb. 2003
Since p13, deghosting is in CPSReco but turned
off. Should it be turned on?
2
Ghosts
f
z
- four 3D clusters formed three from tracks, one
is a ghost
- remove ghosts by limiting number of times a
subcluster can be used
- use cuts based on position and energy matching
quality to decide what to remove
3
2 Matching Qualities for 3D clusters
phix - phiuv
matchquality
position)
resolution
u subcluster
phix
x subcluster
phiuv
f
z
v subcluster
2
r2 - 1/3
Ö
)
(
E
?
i
i
, r
energy) Ematchqual
r (1-r )
E
i
i
i
total
3 layers
4
Parameters and Cuts
ttbar MC event
- in cases like this, limiting a subcluster
to one use is too harsh
- to ease cut
-for every 3D cluster, define
mq_frac(eq_frac)
best position(energy) match quality of all 3D
clusters sharing a subcluster with the 3D cluster
position(energy) match quality
keep 3D cluster if
deghostpwr
deghostpwr
mq_frac eq_frac gt 1
deghostpwr is a parameter
5
Plots of mq_frac vs eq_frac for Z-gtee monte
carlo -red lines represent deghostpwrs of 0.5 and
1 -keep everything above red line -large of 3D
clusters cut, but not many electrons
3D clusters matched to electron
all 3D clusters
6
Effects of deghosting vs pT
no deghosting
deghostpwr 0.5
(these plots use 3200 p13.03.00 ttbar monte carlo
events)
7
Effects of deghosting vs isolation
ratio of electron matches kept vs. number of
other tracks within 8cm
deghostpwr 0.3/no deghosting
deghostpwr 0.5/no deghosting
8
fakes removed vs isolation
deghostpwr 0.5
deghostpwr 0.3
9
Choosing deghostpwr
Results for p13.03.00 ttbar MC
deghostpwr 3d clusters matching efficiency
of e- removed
total gt20GeV lt20GeV 0.3
21 99.5 100.0
99.1 0.4 47
97.4 99.7 95.9 0.5
62 95.2 99.4
92.6
10
deghostpwr is the same for mq_frac and eq_frac.
Why?
- it simplifies things and looks close to optimal
plot comparing the best points(o's) in the left
plot and the equal deghostpwr(x's)
plot made by varying deghostpwr separately for
mq_frac and eq_frac
11
monte carlo
real zee candidates