1. Supersymmetry Searches with the ATLAS Experiment at the LHC
Sven Vahsen, LBNL
for the ATLAS collaboration
DPF 2006 Meeting, Honolulu, Hawaii
October 31, 2006

2. The ATLAS detector at the LHC
2006
2007
2008
Full physics run
(14TeV)
First beams and
collisions (0.9TeV)
LHC (Large Hadron Collider)
Under construction at CERN
14-TeV proton-on-proton in  7 x Tevatron
ATLAS (A Toroidal LHC apparatus)
One of the two general purpose experiments at LHC, to be completed in time for first beam
Supersymmetry (SUSY)
Exciting possibility for physics beyond the standard model
ATLAS sensitive to previously un-probed SUSY-particle masses with very little 14-TeV data
Honolulu, October
Sven Vahsen, LBNL

3. Standard Model at 14 TeV
Expect most events written to tape from high cross-section SM processes
Use to calibrate and measure detector performance
Constitute main backgrounds to SUSY searches
The SUSY-discovery challenge
Use trigger and event selection to reject SM by factor of ~109 (for S/B ~ 1/10)
Demonstrate that we understand SM events that survive SUSY selection
Channel
# of Events, 1pb-1
(17 minutes at 1033cm-2 s-1 !)
W -> µ
7000
Z -> µµ
1100
t tbar -> µ + X 80
QCD jets PT>150 GeV
1000(for 10% of trigger bandwith)
Minimum bias
Trigger limited
gluino-gluino, M~ 1TeV
1-10
SUSY?
Honolulu, October
Sven Vahsen, LBNL

4. Beyond the SM: Supersymmetry?
Supersymmetry (SUSY) is one of the more well-motivated extensions to the standard model
Keeps corrections to higgs mass finite. Requires wino and stop masses ~ few hundred GeV
Unifies gauge couplings at large Q2
Provides cold dark matter candidate via LSP if neutral, weakly interacting, stable. Cosmological arguments prefer mass ~ a few hundred GeV
MSSM: Minimal Supersymmetric extension of the Standard Model.
particles
sparticles
=“Neutralino”
=“Chargino”
Honolulu, October
Sven Vahsen, LBNL

5. Sparticle production
Cross-section for producing SUSY particles in proton-proton collisions calculated at NLO
Production of the strongly interacting sparticles dominates: pp gg,qq,qq,gq
√s=14TeV, mSUSY ~ TeV  σppSUSY ~ pb
 Potential for copious sparticle production at the LHC! _
~~ ~~ ~~ ~~
Honolulu, October
Sven Vahsen, LBNL

6. Sparticle Decays ~ c01 ~ qR ~ g q p p ~ _ c01 ~ ~ ~ qL ~ q c02 g l q q
Event selection guided by typical decay chain of SUSY particles
c01 ~ qR ~ g ~ q p p
c01 ~ q _ ~ qL ~ ~ ~
c02 g l q q l l
Two strongly interacting sparticles produced
If “R-Parity” conserved
Each sparticle decays into lighter sparticle + SM particle ( jets + leptons)
cascade decays down to stable, undetected LSP (usually neutralino in mSUGRA)  large ETmiss
 Canonical SUSY signature: ETmiss, high-pT jets, often leptons
Honolulu, October
Sven Vahsen, LBNL

7. Missing ET + jets Event selection (0-lepton case)
Jets 1,2 with pT > 100 GeV
Jets 3,4 with pT > 50 GeV
ETMISS > max(100 GeV,0.2Meff)
Transverse sphericity ST > 0.2
No isolated muon or electron with pT > 20 GeV
Plot Effective Mass variable:
Meff = Σ|pTi| + ETmiss
SUSY events typically have higher Meff than SM backgrounds
Z(νν) + jet
W(μν)+ jet
QCD
Initial SUSY search strategy tbd. Counting experiment after cut on Meff?
Jets + ETmiss + 0 leptons
ATLAS TDR, 1999
Honolulu, October
Sven Vahsen, LBNL

8. SM background: Monte Carlo predictions
Isajet (1999)
MC prediction of SM events surviving SUSY selection uncertain by a factor 2-5
Need to measure normalization of these backgrounds!
Strategy to do so early ( pb-1) is the topic of much current work in in ATLAS SUSY group
Different SUSY signals!
ALPGEN
(2006)
Pythia
(2006)
Honolulu, October
Sven Vahsen, LBNL

9. Optimizing the event selection
0-leptons
If more Standard Model background than expected
 Re-optimize event selection
lower signal efficiency
higher S/B ratio
Two options
tighten cuts in ETMISS + jets + 0 leptons
require extra particles in final state (next slide)
0-leptons
Honolulu, October
Sven Vahsen, LBNL

10. Missing ET + jets + 1 lepton
Require an additional lepton
 higher S/B, and different BG composition than 0-lepton mode
(0 leptons)
(1 leptons)
ATLAS
preliminary
ATLAS
preliminary
Semi-leptonic top decays now dominate BG. May be possible to directly measure normalization and shape (see later slide)
Typical SUSY cut
NJet>=4 (PT1st>100GeV, pT4th>50GeV)
MET>100GeV and MET>0.2xMeff
ST>0.2
Honolulu, October
Sven Vahsen, LBNL

11. Missing ET + jets + 2 leptons
Example: Opposite sign, same flavor di-leptons from single neutralino decay
Less general search, but edges in di-lepton mass distribution smoking gun
Low Background
Estimate BG using opposite sign, opposite flavor di-leptons
 Robust, i.e. less prone to fake discovery in early searches?
Position of mass-edge sensitive to SUSY masses
First step in larger program beyond discovery: constraining sparticle masses in decay chain using partial reconstruction l+ l-
e+e-, μ+μ- SUSY
e+μ-, e-μ+ SUSY
e+μ-, e-μ+ top-pair
ATLAS
preliminary
Honolulu, October
Sven Vahsen, LBNL

12. Early SUSY discovery possible!
Reach gluino / squark masses of
1.0 TeV with 0.1 fb-1
1.5 TeV with 1.0 fb-1
2.0 TeV with 10 fb-1
To be repeated with more realism: full simulation, non-perfect detector, systematic from background estimation
Different final states may “win” at different times, depending on tradeoff between systematic and statistical uncertainties (i.e. understanding of detector, SM)
Honolulu, October
Sven Vahsen, LBNL

13. Only the finest ingredients!
Main ingredients to SUSY search: jets, ETmiss,leptons
 Bad ingredient  spoil the dish!
(= false discovery…)
Example of ETmiss
can spoil ETMISS resolution with large large tails due to
Miss-measured jets in cracks
Colorimeter noise & inefficiencies
CDF
Honolulu, October
Sven Vahsen, LBNL

14. Estimating background from W+jet, Z+jet
Example of directly measuring SM background: Z(nn)+jets, especially in region with high ETMISS
W+jet ~ 10x higher cross-section than Z+jet, similar kinematics
Select W(mn)+jets and “forget the lepton” to estimate shape of Z(nn)+jets background
Looks promising, and might be feasible already at 100 pb-1
Need to demonstrate that it still works with other SM backgrounds present
Missing ET [GeV]
Z(nn)+njets
Estimated [W(mn)]
Events/50GeV/1fb-1
normalization
ATLAS
preliminary
signal region
Result (Missing ET>300GeV, 1fb-1)
Z(nn)+njets : 157+/-13
Estimated : 134+/-10
SUSY cuts
NJet>=4 (PT1st>100GeV, pT4th>50GeV)
MET>100GeV and MET>0.2xMeff
Nm>0 with Pt(m)>10GeV
Honolulu, October
Sven Vahsen, LBNL

15. Conclusion LHC projected to produce 14-TeV proton collisions in 2008
With 0.1–1 fb-1 of data, expected by end of that year, ATLAS sensitive to sparticle masses in the 1.0 – 1.5 TeV range
However, convincing SUSY search requires good understanding of
Detector
Jets, ETMISS, leptons
SM events passing SUSY selection
Lots to do, including defining the full strategy
If SUSY is actually found, there will be even more work!
2006
2007
2008
Full physics run
(14TeV)
First beams and
collisions (0.9TeV)
Honolulu, October
Sven Vahsen, LBNL

16. BACKUP SLIDES
Honolulu, October
Sven Vahsen, LBNL

17. Estimating background from top-pairs
Top mass is largely uncorrelated with ETMISS (calibration variable).
Select semi-leptonic top candidates in a mass window GeV
Using early data: no b-tagging availablecombinatorial BG to top mass  estimate this from the sideband (Mtop= GeV) and subtract
Normalize ETMISS distribution in low ETMISS region, where SUSY contribution is small
Extrapolate to high ETMISS region to estimate the background after SUSY selection
Needs to be demonstrated that this still works when both SUSY and rest of SM (especially W+jets) present in data sample
How to apply this in 0-lepton mode?
Top Mass (GeV)
ttbar sideband
ttbar signal T1
Missing ET (GeV)
ATLAS
preliminary
Missing ET (GeV)
Honolulu, October
Sven Vahsen, LBNL

18. Only the finest ingredients!
Main ingredients to SUSY search: jets, ETmiss,leptons
 control their quality, so as to not compromise the final product (= avoid fake discoveries!)
Example of ETmiss
Ideally only from undetected particles (neutrino, out of acceptance, LSP)
But “fake” contribution can spoil ETMISS resolution and give large tails
Miss-measured jets in cracks
Colorimeter noise & inefficiencies
Monitor resolution using control sample, e.g. W(mn)+ jet
Clean up by rejecting
bad runs (e.g. large average ETMISS)
events with ETMISS in direction of jets
events with high-pT jets pointing at cracks
Critical issues for SUSY analysis
Jets: resolution, jet energy scale
ETMISS: resolution and scale
Leptons: purity, efficiency, isolation
ATLAS wide challenge to get the above under control with early data samples
Special challenge for SUSY group
pT higher than SM processes
High multiplicity typical of SUSY events
Tails of resolutions
CDF
Honolulu, October
Sven Vahsen, LBNL

19. Optimizing the event selection
BG level in data similar or worse than high-background MC prediction  not show-stopper for discovery
Can re-optimize selection for lower signal efficiency and higher S/B  discovery will take only “a bit” longer
Two options
tighten cuts in ETMISS + jets
require extra particles in final state, e.g.
ETMISS + jets+ 1 lepton
ETMISS + jets+ 2 leptons (SS, OS)
Suitability of particular final states for discovery also depends on feasibility of measuring the SM BG in that channel, so as to achieve to smallest possible σBG
Note that optimum selection cuts depend on SUSY model as well
0-leptons
0-leptons
1-lepton
Honolulu, October
Sven Vahsen, LBNL