1. Report from SUSY working group
MPI ATLAS meeting
F.Legger, J.v.Loeben, XA.Zhuang, V.Zhuravlov

2. SUSY in ATLAS Inclusive search (jets + MET) Exclusive measurements
data-driven background estimation
QCD , top, W
Exclusive measurements
measurements of endpoints
dilepton edge, lepton-jet edge, di-tau endpoint, light stop signatures
Exotics
non-pointing photons
long lived charge particle, R-hadrons
LF violated tau decay

3. Inclusive search 0-lepton mode MET > 100 GeV 4 jets
no interaction with detector  ET,miss
0-lepton mode
MET > 100 GeV
4 jets
Transv. Spher. > 0.2
MET > 0.2 Meff
1-lepton mode
MET > 100 GeV
4 jets
Mt > 100 GeV
Transv. Spher. > 0.2
1 rec. lepton
MET > 0.2 Meff
2-lepton mode (OS)
MET > 100 GeV
4 jets
2 rec. leptons
Transv. Spher. > 0.2
MET > 0.2 Meff
Missing Et (GeV)

4. Estimate background from data: Why?
Bad knowledge of:
Underlined Event
Parton Showering
Cross-sections
Parton Distribution Functions
Detector Calibration (jets, MET)
statistics of Monte Carlo

5. Control Sample as close as possible to the signal region
free of SUSY signal
sufficient statistics
unbiased w.r.t to the signal region
small theoretical uncertainty

6. 2-lepton search: data driven estimation of top background

7. dileptonic top background: clean bb lνlν sample
Quartic equation: 0, 2 or 4 solutions
no solutions: SUSY event, semi-leptonic ttbar, …
2 or 4 solutions: dileptonic top event
4 and more jet in event: loop over all jet combinations.
Clean dileptonic top sample
Number of b-jet pairs

8. 2-lepon search: result Systematic errors: Total syst. error: 16%
biased control sample: 15%
norm. factor stat: 5%
JES 3%
MET scale 3%
JER 1%
Total syst. error: 16%
Stat. error 12%
SU3 over-estimate 30%

9. Why Control sample is biased?
Mainly from unreconstructed low pt b (depending on MET cut)
Others: fakemet, tau decay and gluon radiation (only 10%) 8

10. Cut optimization
2l OS
2l OS
MissEt cut (GeV)
Eff. Mass cut (GeV)

11. 1-lepton search: data driven estimation of top background

12. Di-leptonic Top in 1 lepton Mode: Why? (1)
Transverse mass:
Minv(Missing Pt and PtLepton)
W mass
bbqql
bbll
MT (GeV)
MT (GeV)
Because cut Mt>100 GeV kills almost all semi-leptonic top

13. Why second lepton is not visible?
it is tau: 50 %, hadronic or leptonic with low Pt lepton
lepton Id is not perfect: 20%
it is close to the jet: 17% (we requre 4 and more jets)
events with two taus, one decays leptonically: 3%
lepton out of acceptance: 9%

14. Method
Use the same Control Sample as used by the 2-lepton search (not really the same: no cuts on MET, Meff, ST, 3 jets…)
Contribution of events with 1 tau:
each events of the Control Sample is a “seed event”
one of the two lepton replaced by tau and a set of tau decays is simulated 1000 times
same is repeated for the second lepton, yielding a total of 2000 events for a seed event
make efficiency correction: efficiency of identification of 2 leptons (Control Sample) is less than the efficiency of identification of 1 lepton, factor 1/ε
recacaculate the SUSY selection variables:
MET: take into account one or two neutrino from tau decay
Meff, ST: take into account one replaced lepton
Njets: additional jet produced by hadronic tau

15. Contribution of events with misidentified lepton:
again, the same events from the control sample are used
if lepton is an electron, it is replaced by a jet with the same momentum
if lepton is muon, it is replaced by stand-alone muon (already taken into account for MET)
apply the procedure to each of two lepton of seed event
efficiency correction is ε/(1-ε)

16. How well does it work
Systematic error: 10%

17. 1 lepton search: result Syst. errors Stat. error 12%
replacement: 10%
RMC: 10%
normalization(stat): 9%
normalization(syst): 8%
JES 9%
Stat. error 12%
overestimation due to SU3: 47%

18. τ→μμμ Allowed by SM, but suppressed by δmν2/mW2
Resent limit (B-factories) Br < 5x10-8
Sensitive to new physics
R-parity conserved MSSM with see-saw neutrino masses: 4x10-10
R-parity violated MSSM:
Little Higgs
technicolour
Expected 2x108 taus/10 fb-1 from W decay
Main background Ds and Bs mesons
pp→cc→Ds + X → φ+μ+X → μμμ+X

19. Selection PtMiss > 20 GeV 3 muons in narrow cone EtCone < 10 GeV
veto φ mass
window μμμ mass
3 background events, Br < 1.4x % CL 10fb-1

20. Conclusion 1-lepton and 2-lepton (OS) inclusive searches:
top background: data-driven top background estimation
SUSY CSC note Data-driven determinations of W, Z, and top background
no-lepton mode: in progress
LFV tau decay: on-going

21. Backup slides

22. 0 lepton signature Main background (1 fb-1): Meff>800GeV
3800 (60% ) events:
312 (43%)
top, mostly semileptonic with tau (65%)
1100 (17%) events:
225 (31%) W
800 (13%)
24 (3%)
QCD
600 (10%)
156 (21%)
Z invisible decay: dominant at large MET
Meff (GeV)
SU3 discovery significance:
2.9 (nom.cuts)
12.9 (Meff>800GeV)

23. 1 lepton signature Main background (1 fb-1): Meff>800GeV
130 (91% ) events:
36 (87%)
top, mostly dileptonic, with tau (50%)
10 (7%) events:
5 (12%) W
2 (1%)
0.2 (0.5%) Z
SU3 discovery significance:
7.3 (nom.cuts)
12.3 (Meff>800GeV)

24. 2 lepton signature, opposite sign
Main background (1 fb-1): MET>140GeV, Pt (Leading jet) > 200 GeV
82 (96% ) events:
23 (100%)
top (90% dileptonic)
2 (2%) events:
0 (0%) W
1 (1%)
0 (0.5%) Z
SU3 discovery significance:
7.3 (nom.cuts)
11.4 (Meff>800GeV)

25. Statistical errors Errors are complicated…
One seed event contributes to several bins of estimated distribution: bin-to-bin correlations
covij = Σkmkimkj
mik is a contribution of i-th event to k-th bin
error bar of bin i is
sqrt(covii)
Covariance matrix of MET distribution

26. Another method: top decay resimulation (Sheffield)
Select top seed events
Njet≥2
MET< ½(PT(l1)+PT(l2))
one and only one combination M(jet, l)<155 GeV
Reconstruct top’s momenta (6 equations)
Simulate top decay (atlfast)
Remove original top decay product from event
Merge new decay products with seed event
Perform SUSY cuts

27. τ→μμμ