1. semileptonic ttbar + jet events
Kenneth Wraight
semileptonic ttbar + jet events
pt spectra of extra jets
comparison of matching methods
preliminary reconstruction results

2. Motivations
Top physics is an opportunity to test multiscale QCD calculations by comparing tt+j and tt+jj channels, as well as cross checking generator (e.g. Alpgen) predictions with W+jet events.
Glasgow’s interest lies in associated Higgs production.
Understanding the extra jet pt is a major step to controlling the background to ttH, Hbb and SUSY channels (10s of ttbar events per second for σ=833pb at L=1034cm-2s-1). 1

3. matching is used as a measure of merit
In physics: hard processes hadronise and interact with the detector to produce jets which are the input of analysis algorithms
In simulation: generated samples reflect this tripartite structure:
Generator Level (GL) -> Shower jets -> Simulation Level (SL)
to measure theoretical success (i.e. efficiency) of an analysis a connection between GL partons from the hard process and SL jets is therefore required
aim is to unambiguously match a parton to jet through the simulation process by a generator independent procedure
Hence kinematic (e.g. η, pt, Et) properties are used rather than individual mapping
if an analysis product is composed of the matched jets it is well-matched
What if there is more than one matching procedure?
What if they do not give the same results? 2

4. Nikhef method (ATL-COM-PHYS-2008-046 M. Gosselink et al.)
top daughter partons = 2 b quarks & 2 light quarks
Two protocols
Nikhef method (ATL-COM-PHYS M. Gosselink et al.)
Compare reconstructed top vector to generated top vector. If ΔR<0.4, then declare top well-matched
Alternative methods
A: Match generated top daughter partons to shower jets (ΔR<0.4) then on to simulated jets (ΔR<0.4). The event is then matched. If top is reconstructed from jets matched to generator level partons, then declare top well-matched
B: Match generated top daughter bs and hadronic W to shower jets (ΔR<0.4) then on to simulated jets (ΔR<0.4). The event is then matched. If top is reconstructed from jets matched to generator level partons, then declare top well-matched
C: Match generated top daughter bs and hadronic W to simulation jets (ΔR<0.4). The event is then matched. If top is reconstructed from jets matched to generator level parton, then declare top well-matched
Which is ‘The best’ method of matching?
(I will not answer this) 3

5. defining efficiencies to compare matching
ξsel = #events passing trig.&sel. cuts / #events in sample
depends on triggering and selection criteria
ξmatch = #events matched / #events passing trig.& sel. cuts
depends on matching criteria  matching dependent
ξwell = #events with ‘well-matched’ top chosen / #events matched
combinatorics (, some selection)  matching dependent
ξpotential = #events with ‘well-matched’ present / #events matched 4

6. relevance of efficiencies
Product gives a global efficiency:
ξglobal = ξsel * ξmatch * ξwell
= #events with ‘well-matched’ top chosen / #events in sample
In comparing matching methods ξmatch and ξwell are used.
ξmatch to assess how many sample events can be used to check analysis
ξwell to assess the success of the analysis according to the matching
In addition, ξwell / ξpotential provides a measure of success of reconstruction algorithm. 5

7. standard commissioning
* electron or muon with Ptlep>20GeV within |η|<2.5 fulfilling isolation requirement Et,ΔR=0.2<6. GeV. Also ignore electrons in crack region 1.37<|η|<1.52.
** jets with Ptlep>20GeV within |η|<2.5. Also remove jets with good electron nearby i.e. ΔRj-el<0.2.
standard commissioning
trigger
isolated electron Ptel>25GeV or isolated muon Ptmu>20GeV or
analysis cuts
exactly one good* lepton remove dilepton events
Etmiss>20GeV neutrino
>3 good** jets with Ptjet>20GeV  four quarks from tops
>2 good** jets with Ptjet>40GeV  hadronic top products
(n b-tagged jets (n=1 or 2), not used here)
reconstruction
simple: take highest Pt tri-jet combination
Wmass constraint: take highest Pt tri-jet combination provided it has di-jet combination with Mjj-MW<10.GeV... used here
centrality: take highest Pt tri-jet combination with jets in |η|<1
both: take highest Pt tri-jet with both Wmass and centrality constraints 6

8. efficiencies from standard commissioning
events from 5200 sample fullsim) normalised to 100pb-1
ξsel= 19.5% (es:16.6%, μs:21.9%)
efficiencies from standard commissioning
method
matched ξmatch(%)
well-recon. ξwell(%)
potential ξpot.(%)
ξglobal(%)
Nikhef
4707
100.
1063
22.6
1523
32.4
4.3 A
2682
60.8
346
12.1
698
24.4
1.4 B
2936
62.4
341
11.6
678
23.1 C
3434
73.0
397
832
24.2
1.6
ξmatch
Nikhef matching criterion fulfilled by default
other methods more complex criteria  non-maximal matching
especially A&B where two matching steps are required
events passing other methods are subset of Nikhef passed events (ΔRsimtop<0.4)
Not all (~74%) of Nikhef tops are within 10GeV of other matched top (in case C).
ξwell & ξpotential ... comparing recon. algorithm
Loss of events from ξmatch to ξpotential is due to kinematic cuts (pt, η, jetcut=3)
By Nikhef method, 1/3 of events incorrectly reconstructed (combinatorial)
By, other methods, ~1/2 of events are incorrectly reconstructed (combinatorial)
ξglobal
Nikhef method defines ~3 times as many selected and well-matched events 7

9. * quoted values are not exclusive
Non-unique matches exist in the case of each protocol...
multiple matching
Nikhef: 26% 
A*:
55% 1st step
44% 2nd step 
B*:
74% 1st step
35% 2nd step
C: 31% 8

10.  more top combinations making reconstruction more difficult
Concentrate on Nikhef and C methods...
5212 sample
5212 sample fullsim) has at least two extra jets in each event
selection efficiencies:
ξsel= 29.1% (es:25.1%, μs:32.6%)
method
ξmatch(%)
ξwell(%)
ξpot.(%)
ξglobal(%)
Nikhef
100.
20.0
31.7
5.8 C
72.0
10.1
25.3
2.1
multiple matches are more common in this higher multiplicity sample:
Nikhef: 35% and method C: 36%
more jets are present to pass the jet selection and fake lepton selection.
 more top combinations making reconstruction more difficult
& more jets combinations to pass matching criteria. 9

11. properties of matched jets using method C
Hadronic W daughter in green & red, hadronic b in blue
properties of matched jets using method C
b-weight:
ΔEjet->parton:
5200 sample
5212 sample 10

12. reconstructed top mass
reconstructed top (black) with non-well-matched proportion (red)
reconstructed top mass
Nikhef method:
method C:
5200 sample
5212 sample 11

13. Nikhef method in black, method C in red
area normalised
Nikhef method in black, method C in red
extra jet pt spectra
X≥1
X≥2
5200:
5212: 12

14. Calculating cross-section
Split sample in two
sample A
sample B
use half to determine efficiencies and define acceptance:
Purity= #well-matched / #reconstructed
matching dependent
Efficiency: #well-matched / #sample events
Acceptance = Efficiency / Purity
= #reconstructed / #sample events
 matching independent
apply acceptance to other half to get cross-section plots
Ngen = (Ndet / Lum.) * (1/ Acc.)
and compare to generated cross-section

15. cross-sections (unscaled)
Generated mass & pt plots from shower level
cross-sections (unscaled)
tri-jet mass:
tri-jet pt:
calculated
generated
5200 sample
5212 sample 13

16. Matching independent method used to calculate cross-sections
Summary
Comparison of two different matching procedures shows disparities between results.
Different matching procedures give different degrees of confidence in analysis method.
This ambiguity makes testing multiscale QCD calculations and generator predictions more difficult.
These difficulties are particularly important for extra-jet studies where the identity of the simulated jets is crucial.
Matching independent method used to calculate cross-sections
Next step is to investigate fully leptonic channel where it is easier to identify extra jets and not as reliant on jet matching. 14

17. method ξmatch(%) ξwell(%) ξpot.(%) ξglobal(%) Nikhef 100. 22.6 33.4
Three parton (hadronic side) efficiencies
BACKUP
5200 sample:
selection efficiencies: (as before)
ξsel= 29.1% (es:25.5%, μs:32.6%)
method
ξmatch(%)
ξwell(%)
ξpot.(%)
ξglobal(%)
Nikhef
100.
22.6
33.4
4.3
C_3
76.9
11.3
23.7
1.7
multiple matches are more common in this higher multiplicity sample:
Nikhef: 35% and method C_3: 23%
with only three partons to match:
fewer opportunities to double cross match partons (so fewer events rejected)
fewer opportunities to match parton to multiple jets 9