Post on 02-Apr-2015
Electron and muon reconstruction in single lepton tT events
(Analysis Note 2005-024) (short!) update of previous session new definition of some observables less sensititive to Pile-Up effects updated all results in discussion with referees ... focus on main updated items
Jorgen D’Hondt (Vrije Universiteit Brussel)on behalf of also S.Lowette and J.Heyninck
PRS General Meeting, CERN, 6th of December 2005
Referees : Anne-Sylvie Giolo-Nicollerat, Martin Grünewald
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 2
Illustration
Select correctlepton
Used 5.41fb-1 of ttbar events
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 3
Classifying the reconstructed lepton
non-W-like
Classify the reconstructed objects identified as leptons in two categories via the Monte Carlo truth information. Determine for each lepton candidate in the final state.
distribution of the R of all reconstructed lepton candidates (e/) in the final state robustness against the used metric was checked: () versus ()
R=0.2 → where one distribution goes up on top of the background
non-W-likeW-like
0.2Takes into account double counted reconstructed electron candidates
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 4
Differentiating the two classessensitive observables
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 5
Observables for the W→ decay
Eiso
iso
log(pTiso)
log(pT)
SZ
• Distribution on the left• Likelihood Ratio S/(S+B) on the right
2.0 GeV (PU)
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 6
Observables for the W→ee decay
Eiso
iso
log(pTiso)
log(pT)
SZ
• Distribution on the left• Likelihood Ratio S/(S+B) on the right
Same tendencies for electrons and muons, as expected
3.6 GeV (PU)
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 7
Combined Likelihood Ratio Variable
linear scale
logarithmic scale
muons electrons
Second bump has disappeared due to new definitions
Correct lepton is the one with the highest Combined LR value
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 8
Comparing with other methods
muons electronscut on the LR-value andscan this LR-cut
Event selection efficiency of the Likelihood Ratio method much less sensitive to Pile-Up effects, which can reach 4% when using a sequential cut based approach.
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 9
Pile-Up influence on observables
about 0.5GeV effect
no effect
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 10
Energy deposits in the calorimeter
Jet clustering algorithms usually start from the collection of calorimeter deposits, but also the lepton creates these deposits and can radiate photons. These calorimeter deposits should be removed before performing the jet clustering algorithm.
It can be seen that there is an asymmetry between the deposits in and , due to the magnetic field along the z-direction. The deposits around the chosen lepton within an asymmetric cone of opening 0.1rad in and 0.05rad in should be removed.
muon electron
Plots have changed due to the propagation of the lepton to the calorimeter surface
Pile-Up included
6th of December 2005 Jorgen D'Hondt (Vrije Universiteit Brussel) 11
Outlook in use cases
Update of previous presented studies (PRS meeting November)
Variable created which differentiates W-like and non-W-like leptons
also differentiates between fake electrons or muons several sensitive observables are combined with a Likelihood Ratio method it was shown that one can increase the event selection efficiency (wrt.HLT) the definition of some variables were updated (= simplified)
Comments from referees were implemented and discussed thoroughly in the updating process
thanks to both Anne-Sylvie and Martin for their constructive and valuable comments