Electron and Photon Identification

1
Electron and Photon Identification

• Drew Alton, University of Michigan
• Harald Fox, Northwestern
• Yurii Maravin, Fermilab
• Jan Stark, Grenoble

Winter Physics Workshop, February 22nd, 2004
2
Electron-ID Performance and documentation
The performance of our p14 electron
identification machinery is documented in great
detail in the Z ? e e- and W ? e ? cross
section note by Chapin, Fox, Gardner,
Illingworth, Lyon, Steinbrück, Verzocchi and
Zhu - Cluster reconstruction efficiency -
Preselection efficiency - H-matrix
efficiency, - Track-matching efficiency, -
Fake-rates, - This analysis note also
serves as a standard reference on EM-ID
performance. For this session of the workshop,
we will follow the same logic standard
performance plots will be shown by John in the
first analysis talk this afternoon (right after
Patrice on missing ET). The present talk will
focus on more general issues.
3
p14 improved geometry and scale corrections
Geometry corrections (Magnan, Crepe)
EM scale corrections (Kermiche, Mendes) Based
on Z. Also studying J/Psi now.
4
p14 all new electron likelihood
For p14, a new electron likelihood was
implemented by Kozminski, Park, Gardner, Jabeen,
Kehoe, Quadt and Weerts. A draft D0 note is
available. It is used with loose precuts and
exploits all the remaining information (HMx8,
E/p, track-matching, track-isolation, ). It
comes in two flavours one without CPS and one
with CPS information.
For p13, we were on this curve. Now we are on
this curve.
Unfortunately, some CPS information was lost
during the first pass of TMB fixing. The upcoming
second pass (from the original TMBs) will recover
this information. Looking forward to using the
CPS in physics analysis !
5
New for p14 H-matrix 7
Blue Monte Carlo Red data
HMx8 CC
HMx8 EC
HMx8 data/MC agreement lets something to be
desired, especially in CC. One of the worst
offenders shower shape in z. Removed one of the
corresponding input variables from the H-matrix
? HMx7.
HMx7 CC
HMx7 EC
Rejection power of HMx7 and HMx8 in the CC for
different cuts. HMx7 does not only have better
data/MC agreement, it also gives better
rejection. HMx7 will be included in electron
likelihood.
Greg L. HF
6
Energy resolution
Blue Monte Carlo Red data
Z ? e e-
At the conferences our Z ? e e- peak will look
like this.
Di-EM mass (GeV)
But we to keep in mind that, before energy
smearing (S. Jain) it looks like this. We have a
task-force working on this problem, your
contribution is welcome.
Z ? e e-
Di-EM mass (GeV)
7
Track matching
The cluster-track matching (R. Zitoun ? M. Titov)
comes in two flavours spatial only, and spatial
E/p.
centered around 1.05 ? 0.15
fake rate
efficiency
E/p
With E/p, the efficiency is somewhat low. But
keep in mind computing a ?2 probability with
non-gaussian distribution does not make
sense. Recommended solution use electron
likelihood. Another issue spatial
track-matching efficiency drops pretty badly at
high detector ?. This is mostly a tracking
problem, not a matching problem. More on next
slide.
Spatial track-matching efficiency as a function
of detector ?
8
Track matching at high ?
Z ? e e- candidate, right at the centre of the
mass peak.
Track ?
Track !
Cannot conclude anything from one event. But we
have a whole load of such events.
9
A reminder DST repro vs. reco from RAW
This is the crappiest t-tbar candidate from the
ejets/soft-muon-tag analysis. The default
analysis uses run 165777 reprocessed from a p13
DST. On the right, the kinematics of the
candidate event.
When the candidate event is processed from
RAW (p14.06.00, event display on the right), it
fails the electron likelihood cut. The
likelihood includes a track isolation. Most
likely, that suspicious track close to
the electron candidate is not found in DST
reprocessing. Such subtle differences are
expected. This is just a reminder that such
oddities can occur.
10
EM objects and data quality
J. Gardner
High-pT EM objects are a great tool to find
electronics problems in the calorimeter precision
readout. Unfortunately that means that analyses
using objects are very sensitive to the
electronics problems. ?/? distribution of all
candidate EM objects pT gt 25 GeV EM
fraction gt 0.9 isolation lt 0.15 HMx8 lt
20 pink boxes tower two problem blue boxes
energy sharing problem Green boxes cable swaps
These three problems are taken care of by
the TMB fixing. But keep in mind that we
cannot fix events that were rejected at
L3. red boxes other stuff
Checkerboard
The calorimeter hard- and software experts are
working hard to get a checkerboard fix ready for
the Second pass of TMB fixing.
11
Photon ID

• First-order photon identification a photon is
  an electron candidate without track-match.
• Sophisticated tools for next-order photon ID are
  in the final stages of development
• Cal-CPS-TRK (not-trk) matching (YM Askew)
• Trying to choose best method
• Line from Vertex-EM3 choose closest
• Fit CPSEM1-4 find best fit
• Line from 0,0-EM3 choose highest energy
• Prefer not to use vertex
• Hits on the Road(Atraventov YM)
• Separate electrons, photons, fakes
• Recover/remove electrons without track match
• Separate photons from EM-jets
• Information
• c2 for both roads
• Number of hits on each road from each detector

12
Photon ID
For electron-ID, it is relatively easy to get a
control sample of real electrons (di-EM
resonances). Trying to get a control sample of
real photon (DA et al.) Wg, QCD-g, p0,
cc, D, etc Other projects pursued by the
photon ID subgroup - Implement standard
track isolation - Test/study other cuts and
construct likelihood - MC studies comparing e and
g - Efficiency measurements in data (system 8?)
13
Conclusions
Electron and photon ID are in a reasonably good
shape during this afternoon, we will see that
nice analyses can be done based on these
objects. Powerful new developments have been
introduced in p14 - TMB fixing - improved
geometry and scale corrections - HMx7 - new
track-matching - new electron likelihood -
Many parts of this particular talk had sort of
a negative note. This is because the speaker
believes that he can leave the excitement for the
analysis talks and that it is his role to point
out areas with room for improvement. Over the
past months, a lot of effort has shifted from ID
work towards analysis. We are seeing The fruits
of this analysis effort (and also our past
detector and ID work !) today, and we hope For a
large-scale re-involvement in ID-work after
Moriond.