1. Mass Reconstruction Methods in ATLAS
S. Laplace
On behalf of the ATLAS collaboration
Physics at LHC – Cracow, Poland
SUSY Session, July 4th 2006
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

2. S. Laplace, "Mass Reconstruction Methods"
Outline
Introduction: ATLAS Activities in SUSY
SUSY Phenomenology and Meas. Strategies
Discovering SUSY
Mass Measurements:
Masses: Endpoint Method
Masses near Dilepton Endpoint
and Mass Relation Method
From Measurements to Model Parameters
Conclusion
(note: no time to talk about stop mass measurement and
other methods than endpoints like Mass Relation Method…)
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

3. S. Laplace, "Mass Reconstruction Methods"
Introduction
ATLAS activities in SUSY:
TDR (1998): fast simulation studies  discovery potential
Currently:
Full simulation studies (preliminary results)
Commissioning, systematics
Background estimation (from latest MC and plans to
measure it from data)
New measurement techniques
Note: in this talk,
MET = Missing Transverse Energy
Sleptons = selectrons and smuons (will explicitly call a stau a stau)
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

4. SUSY Phenomenology and Mass Measurement Strategies
If R-parity ( ) is conserved, then:
Lightest Supersymetric Particle (LSP) is stable
LSP not detected thus large MET (few x 100 GeV)
Event is not fully reconstructed: no mass peak
Sparticles produced in pairs: both sides of event are not reconstructed !
Mass measurement strategy: exploit kinematics of long decay chains
Production of SUSY at LHC: strong interactions dominates:
 decay chain starts from a gluino or a squark:
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

5. S. Laplace, "Mass Reconstruction Methods"
(RPC) SUSY Models
Focus point
(m0  3 TeV)
Simple benchmark:
mSUGRA
SUSY Parameters (SM = 19):
M.S.S.M
(note: if RPV )
Constrained models:
mSUGRA
m0, m1/2, A0, tan β, sgn μ
G.M.S.B.
λ, Mmes, N5, tan β, sgn μ, Cgrav 6
A.M.S.B.
m0, m3/2, tan β, sgn μ
+ funnel region
at large tan
bs
g-2
WMAP
Bulk
(SPS1a)
Stau
coannihilation
Ellis et al., Phys. B565 (2003) 176
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

6. Discovering SUSY and Evaluating MSUSY
RPC models signature: MET + several high-pT jets
 Build discriminating variable Meff:
where
Coannihilation point
Full sim
20.6fb−1
SUSY signal
SM Bkg
(Herwig)
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

7. Mass Measurement: Endpoint Method
Example: dilepton endpoint
mll has a kinematic endpoint that
depends on the masses of the
sparticles in the chain
Does not need a-priori knowledge
of any sparticle mass
Backgrounds:
SM & uncorrelated (not Z) SUSY:
use Same Flavour (SF) –
Different Flavour (DF)
Edge fit: stat. error = 0.05%, syst. error
dominated by lepton energy scale (0.1%)
SPS1a
Fast sim
300 fb−1
B.K. Gjelsten et al, J. High Energy Phys. JHEP12(2004)003 
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

8. A Variety of Endpoint Measurements
Sequential:
Branched:
SPS1a
Fast sim
300 fb−1
Bulk
Full sim
4.20fb−1
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

9. Di-lepton Endpoint in Various mSUGRA Scenarii
Depending on point: different shape, number of edges, 2-body vs 3-body decay, …
Coannihilation
Focus Point
ATLAS
MC truth lL
MC truth lR
Full sim
6.9fb−1
signal
Full Sim
20.6fb−1
2 edges for
left and right
slepton
m0 large, heavy scalars
 no sleptons in  decays
direct 3-body decay:
small BR
at least 1 lepton with
small pT
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

10. Extraction of Sparticle Masses from Endpoints
100 fb-1
MC toy of ATLAS experiments, use inversion formulae to get masses from edges:
SPS1a
All masses are strongly
correlated with
B.K. Gjelsten et al, J. High Energy Phys. JHEP12(2004)003 
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

11. Right-Handed Squark Mass
mSUGRA: 1 essentially a bino: Br( )  100%
If both gluino decay to right-handed squarks:
require 2 high-pT jets, MET
Discriminant: Cambridge variable MT2 endpoint gives
the right squark mass:
(low pT) q
(high pT)
SPS1a
Fast sim
30 fb−1
Coannihilation
Full sim
20.6 fb−1
True
Mass
520 GeV
True: 735
Fit: 7115
Fitted edge:
512 GeV
Lower than true
because of
SUSY bkg
SM bkg
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

12. S. Laplace, "Mass Reconstruction Methods"
Staus Signatures
SPS1a: dominant decay is
(because of relatively high tan value)
Look at hadronic  decays (dedicated algorithms for -jets)
Background (QCD jets misidentified as  ) evaluated from
same signs events:
All:
Same sign
substracted:
(Z+j, tt)
(signal)
SPS1a
Fast sim
30 fb-1
(background)
B.K. Gjelsten et al, ATL-PHYS
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

13. Sbottom and Gluino Masses: Near The l+l- Endpoint
Near l+l- endpoint: LSP and l+l- are at rest in frame,
thus can evaluate momentum (approximation):
where and
are known from endpoints b b
Add 1 or 2 b-jet to get sbottom and gluino masses: and
Correlation between
and
SPS1a
Fast sim
300 fb-1
=2.2 GeV
Wrong
associated
b-jet
SUSY bkg
Spread from p(2)approximation is
common to both masses
Gluino mass
Gluino – sbottom masses
B.K. Gjelsten et al, ATL-PHYS
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

14. Sbottom and Gluino Masses: Mass Relation Method
Alternative method to previous one using ALL data set (not only near endpoint)
Each event = 4D surface in 5D space
In principle: 5 events to determine
the 4 unknowns !
In practice: know
so have following constraint:
5 parameters
4 unknowns (4-momentum)
Endpoint only:
Not obvious to
resolve the 2 peaks !
SPS1a
Fast sim
300 fb-1
Two possible solutions
(2 lepton assignments) b1 b2 b1 b2
 The two b-peaks are
well resolved
Mass Relation
Method
Kawagoe et al, hep-ph/
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

15. Obtaining the Fundamental Model Parameters
LHC Measurements
SUSY Model
Ex: mSUGRA
m0, m1/2, A0, tan, sgn()
Spectrum
Generator
(Ex: SUSPECT,
SoftSUSY, …)
Ex: endpoints
Fit: 2
Mes.
Note: better to exploit edges than masses (correlations)
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

16. An Example SFITTER program: List of measurements (300 fb-1)
mSUGRA Parameter
determination
SPS1a
ΔLHC edges m0
100
1.2
m1/2
250
1.0
tanβ 10
0.9 A0
-100 20
Sign(μ) fixed
Note: m(ll) most powerful input (m0 driven
by 1st and 2nd generation slepton sector)
R. Lafaye, T. Plehn, D. Zerwas, hep-ph/
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

17. S. Laplace, "Mass Reconstruction Methods"
Conclusion
New era for SUSY studies in ATLAS is currently starting:
large scale productions to prepare for real data analysis
study detector systematics
SM background: latest MC and plans to measure it from data
new models studied
new techniques developed
Discovery potential: in most models, a few fb-1 are sufficient to:
observe squarks and gluons below 1-2 TeV and sleptons below 300 GeV
accurately measure squark, slepton and neutralino masses using cascades
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

18. S. Laplace, "Mass Reconstruction Methods"
Backup
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

19. mSUGRA Excluded by b s (CLEO,BELLE) Favored by gμ−2 at the 2σ level
WMAP: 0.094<Ωχh2<0.129
Excluded by b s
(CLEO,BELLE)
Favored by gμ−2 at the 2σ level
Muon g−2 coll.
Stau1=LSP
Funnel region
s-channel Higgs-exchange.
Focus point
Bulk region
t-channel slepton exchange.
(ATL-PHYS )
Stau coannihilation
(Ellis et al., Phys. B565 (2003) 176)
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

20. S. Laplace, "Mass Reconstruction Methods"
SPS1a Point
mSUGRA fundamental parameters :
Mass spectrum :
Main branching ratios :
(note: )
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

21. Meff: Parton Shower vs Matrix Element for Bkg Simulation
TDR:
LHC Point 5
Isajet (PS)
Fast sim
10 fb-1
Recently:
Alpgen (ME)
Fast sim
10 fb-1
Parton Shower (only good in
collinear region)
Matrix Element (more correct)
 Background increases by factor 2 to 5 !
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"

22. S. Laplace, "Mass Reconstruction Methods"
Stop Mass Measurement
SPS5: light stop
Reconstruct stop mass via
Signature: 2 b-jets, MET, 3 light-quark jets
Fit m(tb) distribution endpoint:
Fast sim
300 fb-1
M(tb)fit= ± 0.3(stat.) ± 2.6(syst.)
July 4, 2006
S. Laplace, "Mass Reconstruction Methods"