Supersymmetry at ATLAS

11 NIKHEF, January 2007NIKHEF, January 2007Dan ToveyDan Tovey

ConstructionConstruction

Supersymmetryat ATLAS

Supersymmetryat ATLAS

Dan Tovey

University of Sheffield

Dan Tovey

University of Sheffield


SupersymmetrySupersymmetry• Supersymmetry (SUSY) fundamental

continuous symmetry connecting fermions and bosons

Q|F> = |B>, Q|B> = |F>

• {Q,Q}=-2pgenerators obey anti-

commutation relations with 4-mom– Connection to space-time symmetry

• SUSY stabilises Higgs mass against loop corrections (gauge hierarchy/fine-tuning problem)– Leads to Higgs mass 135 GeV

– Good agreement with LEP constraints from EW global fits

• SUSY modifies running of SM gauge couplings ‘just enough’ to give Grand Unification at single scale.

LEPEWWGWinter 2006

mH<207 GeV (95%CL)


e

H-d

~H+

u

~H0

d

~H0

u

~0

4~0

3~0

2~0

1~

SUSY SpectrumSUSY Spectrum

• Expect SUSY partners to have same masses as SM states– Not observed

(despite best efforts!)

– SUSY must be a broken symmetry at low energy

• Higgs sector also expanded

• SUSY gives rise to partners of SM states with opposite spin-statistics but otherwise same Quantum Numbers.

H±H0A

G

e

bt

sc

du

g

W±Z

h

W±Z~ ~ ~

g~

G~

± 2~± 1

~

e

e

bt

sc

du~ ~

~

~

~ ~

~ ~ ~

~

~

~


SUSY & Dark MatterSUSY & Dark Matter• R-Parity Rp = (-1)3B+2S+L

• Conservation of Rp (motivated e.g. by string models) attractive– e.g. protects proton from

rapid decay via SUSY states

• Causes Lightest SUSY Particle (LSP) to be absolutely stable

• LSP neutral/weakly interacting to escape astroparticle bounds on anomalous heavy elements.

• Naturally provides solution to dark matter problem

• R-Parity violating models still possible not covered here.

Universe Over-Closed

mSUGRA A0=0, tan() = 10, >0

Ellis et al. hep-ph/0303043

01

01

l

llR

~

~~

01

1

/Z/h1

~

~~

Disfavoured by BR (b s) = (3.2 0.5) 10-4 (CLEO, BELLE)

0.094 h2 0.129 (WMAP-1year)


SUSY @ ATLASSUSY @ ATLAS

• Much preparatory work carried out historically by ATLAS – Summarised in Detector and

Physics Performance TDR (1998/9).

• Work continuing to ensure ready to test new ideas with first data.

• Concentrate here on more recent work.

• LHC will be a 14 TeV proton-proton collider located inside the LEP tunnel at CERN.

• Luminosity goals:– 10 fb-1 / year (first ~3 years)

– 100 fb-1/year (subsequently).

• First data this year!• Higgs & SUSY main goals.


Model FrameworkModel Framework• Minimal Supersymmetric Extension of the Standard Model (MSSM)

contains > 105 free parameters, NMSSM etc. has more difficult to map complete parameter space!

• Assume specific well-motivated model framework in which generic signatures can be studied.

• Often assume SUSY broken by gravitational interactions mSUGRA/CMSSM framework : unified masses and couplings at the GUT scale 5 free parameters(m0, m1/2, A0, tan(), sgn()).

• R-Parity assumed to be conserved.• Exclusive studies use benchmark

points in mSUGRA parameter space:• LHCC Points 1-6;

• Post-LEP benchmarks (Battaglia et al.);

• Snowmass Points and Slopes (SPS);

• etc…

LHCC mSUGRA Points

1 2

3

45


SUSY SignaturesSUSY Signatures• Q: What do we expect SUSY events @ LHC to look like?• A: Look at typical decay chain:

• Strongly interacting sparticles (squarks, gluinos) dominate production.

• Heavier than sleptons, gauginos etc. cascade decays to LSP.• Long decay chains and large mass differences between SUSY states

– Many high pT objects observed (leptons, jets, b-jets).

• If R-Parity conserved LSP (lightest neutralino in mSUGRA) stable and sparticles pair produced.– Large ET

miss signature (c.f. Wl).

• Closest equivalent SM signature tWb.

lqq

l

g~ q~ l~

~

~

p p


ATLAS

5

• Use 'golden' Jets + n leptons + ETmiss discovery channel.

• Map statistical discovery reach in mSUGRA m0-m1/2 parameter space.

• Sensitivity only weakly dependent on A0, tan() and sign().

• Syst.+ stat. reach harder to assess: focus of current & future work.

Inclusive SearchesInclusive Searches

ATLAS

5


Inclusive SUSYInclusive SUSY• First SUSY parameter to be measured

may be mass scale:– Defined as weighted mean of masses of

initial sparticles.

• Calculate distribution of 'effective mass' variable defined as scalar sum of masses of all jets (or four hardest) and ET

miss:

Meff=pTi| + ET

miss.

• Distribution peaked at ~ twice SUSY mass scale for signal events.

• Pseudo 'model-independent' measurement.

• With PS typical measurement error (syst+stat) ~10% for mSUGRA models for 10 fb-1 , errors much greater with ME calculation an important lesson …

Jets + ETmiss + 0 leptons

ATLAS

10 fb-1


Exclusive StudiesExclusive Studies• With more data will attempt to measure weak scale SUSY parameters (masses etc.) using exclusive channels.

• Different philosophy to TeV Run II (better S/B, longer decay chains) aim to use model-independent measures.

• Two neutral LSPs escape from each event – Impossible to measure mass of each sparticle using one channel alone

• Use kinematic end-points to measure combinations of masses.

• Old technique used many times before ( mass from decay spectrum, W (transverse) mass in Wl).

• Difference here is we don't know mass of neutral final state particles.

lqql

g~ q~ lR

~

~

~p p


Dilepton Edge MeasurementsDilepton Edge Measurements

~~02

~01

l ll

e+e- + +-

~

~

30 fb-1

atlfast

Physics TDR

Point 5

e+e- + +- - e+- - +e-

5 fb-1

SU3

ATLAS ATLAS

•m0 = 100 GeV•m1/2 = 300 GeV•A0 = -300 GeV•tan() = 6•sgn() = +1

• When kinematically accessible

canundergo sequential two-body decay to

via a right-slepton (e.g. LHC Point 5).

• Results in sharp OS SF dilepton invariant mass edge sensitive to combination of masses of sparticles.

• Can perform SM & SUSY background subtraction using OF distribution

e+e- + +- - e+- - +e-

• Position of edge measured with precision ~ 0.5%

(30 fb-1).


Measurements With SquarksMeasurements With Squarks• Dilepton edge starting point for reconstruction of decay chain.• Make invariant mass combinations of leptons and jets.• Gives multiple constraints on combinations of four masses. • Sensitivity to individual sparticle masses.

~~

~

l ll

qL

q

~

~

~

bh

qL

q

~

b

llq edge1% error(100 fb-1)

lq edge1% error(100 fb-1)

llq threshold2% error(100 fb-1)

bbq edge

TDR,Point 5

TDR,Point 5

TDR,Point 5

TDR,Point 5

ATLAS ATLAS ATLAS ATLAS

1% error(100 fb-1)


• Full Geant4 simulation of signals + backgrounds now routine.

• Invariant mass of dilepton+jet bounded by Mllq

min=271.8 GeV and Mllq

max =501 GeV.• Leptons combined with two jets

with highest pT • Flavour subtraction & efficiency

correction applied

Full SimulationFull Simulation

Smaller of M(llq)

Larger of M(llq)Smaller of M(llq)4.20 fb-1

2j100+2l10

4.20 fb-1

2j100+2l10

M(qql) (GeV)200 400 6000

600400200

0

10

Eve

nts

0

30

ATLAS

Preliminary

ATLAS

Preliminary

4.20 fb-1

No cuts

Dilepton Edge = (100.3 0.4) GeV

ll endpoint

ATLAS

Preliminary

SU3

•m0 = 100 GeV•m1/2 = 300 GeV•A0 = -300 GeV•tan() = 6•sgn() = +1


‘Model-Independent’ Masses‘Model-Independent’ Masses

• Combine measurements from edges from different jet/lepton combinations to obtain ‘model-independent’ mass measurements.

01 lR

02 qL

Mass (GeV)Mass (GeV)

Mass (GeV)Mass (GeV)

~

~

~

~

ATLAS ATLAS

ATLAS ATLAS

Sparticle Expected precision (100 fb-1)

qL 3%

02 6%

lR 9%

01 12%

~

~

~

~

LHCCPoint 5


Sbottom/Gluino MassSbottom/Gluino Mass• Following measurement of squark, slepton

and neutralino masses move up decay chain and study alternative chains.

• One possibility: require b-tagged jet in addition to dileptons.

• Give sensitivity to sbottom mass (actually two peaks) and gluino mass.

• Problem with large error on input 01 mass

remains reconstruct difference of gluino and sbottom masses.

• Allows separation of b1 and b2 with 300 fb-1.

lbb

l

g~ b~ lR

~

~

~p p 300 fb-1

~

~ ~

m(g)-0.99m(01)

= (500.0 ± 6.4) GeV

300 fb-1

ATLAS

ATLAS

SPS1a

SPS1a

m(g)-m(b1) = (103.3 ± 1.8) GeV

m(g)-m(b2) = (70.6 ± 2.6) GeV

~ ~

~ ~

~ ~


Stop MassStop Mass• Look at edge in tb mass distribution.• Contains contributions from

– gtt1tb+1

– gbb1bt+1

– SUSY backgrounds

• Measures weighted mean of end-points• Require m(jj) ~ m(W), m(jjb) ~ m(t)

• Subtract sidebands from m(jj) distribution• Can use similar approach with gtt1tt0

i

– Di-top selection with sideband subtraction

• Also use ‘standard’ bbll analyses (previous slide)

~

~ ~

~ ~

~

120 fb-1

ATLAS

LHCC Pt 5 (tan()=10)

120 fb-1

ATLAS

LHCC Pt 5 (tan()=10)

mtbmax = (443.2 ± 7.4stat) GeV

Expected = 459 GeV

mtbmax = (510.6 ± 5.4stat) GeV

Expected = 543 GeV

~

~

~


RH Squark MassRH Squark Mass• Right handed squarks difficult as rarely decay via ‘standard’ 0

2 chain

– Typically BR (qR 01q) > 99%.

• Instead search for events with 2 hard jets and lots of ETmiss.

• Reconstruct mass using ‘stransverse mass’ (Allanach et al.):mT2

2 = min [max{mT2(pT

j(1),qT(1);m),mT

2(pTj(2),qT

(2);m)}]

• Needs 01 mass measurement as input.

• Also works for sleptons.

qT(1)+qT

(2)=ETmiss

~ ~

~qR

q~

ATLAS

ATLAS

ATLAS

30 fb-130 fb-1 100 fb-1

Left sleptonRight

squark

Right squark

~

~

SPS1a

SPS1aSPS1a

Precision ~ 3%


Heavy Gaugino MeasurementsHeavy Gaugino Measurements

• Also possible to identify dilepton edges from decays of heavy gauginos.

• Requires high stats.• Crucial input to reconstruction of MSSM

neutralino mass matrix (independent of SUSY breaking scenario).

ATLAS100 fb-1 ATLAS

100 fb-1

ATLAS100 fb-1

ATLAS

SPS1a

SPS1a


Measuring Model ParametersMeasuring Model Parameters

• Alternative use for SUSY observables (invariant mass end-points, thresholds etc.).

• Here assume mSUGRA/CMSSM model and perform global fit of model parameters to observables– So far mostly private codes but e.g. SFITTER, FITTINO now on the market;

– c.f. global EW fits at LEP, ZFITTER, TOPAZ0 etc.

Point m0 m1/2 A0 tan() sign() LHC Point 5 100 300 300 2 +1 SPS1a 100 250 -100 10 +1

Parameter Expected precision (300 fb-1) m0 2% m1/2 0.6% tan() 9% A0 16%


SUSY Dark MatterSUSY Dark Matter

Baer et al. hep-ph/0305191

LEP 2 No REWSB

LHC Point 5: >5 error (300 fb-1)

p=10-11 pb

p=10-10 pb

p=10-9 pb

• Can use parameter measurements for many purposes, e.g. estimate LSP Dark Matter properties (e.g. for 300 fb-1, SPS1a)– h2 = 0.1921 0.0053– log10(p/pb) = -8.17 0.04 SPS1a: >5

error (300 fb-1)

Micromegas 1.1 (Belanger et al.)+ ISASUGRA 7.69

ATLAS

300 fb-1

p

ATLAS

300 fb-1

DarkSUSY 3.14.02 (Gondolo et al.)+ ISASUGRA 7.69

h2


ComplementarityComplementarity

psi

h2

ZEPLIN-MAX

GENIUSXENON

ZEPLIN-4

ZEPLIN-2

EDELWEISS 2

CRESST-II

ZEPLIN-I

EDELWEISS

CDMSDAMA

• Aim to test compatibility of e.g. SUSY signal with DM hypothesis (data from astroparticle experiments)– Fit SUSY model parameters to

ATLAS measurements – Use to calculate DM parameters

h2, m, psi

etc.

• Measurements at ATLAS complementary to direct and indirect astroparticle searches / measurements– Uncorrelated systematics

– Measures different parameters

p

si (

cm

2)

log10 (psi / 1pb)h2

No

. M

C E

xpe

rim

ents

DM particle mass m (GeV)

D.R. Tovey et al., JHEP 0405 (2004) 071

CDMS-II

Provides strongest possible test of Dark Matter model

psi

m


SUSY Dark MatterSUSY Dark Matter• SUSY (e.g. mSUGRA) parameter space strongly constrained by

cosmology (e.g. WMAP satellite) data. mSUGRA A0=0, tan() = 10, >0

'Bulk' region: t-channel slepton exchange - LSP mostly Bino. 'Bread and Butter' region for LHC Expts.

'Focus point' region: significant h component to LSP enhances annihilation to gauge bosons

~

Ellis et al. hep-ph/0303043

01

01

l

llR

~

~~

01

1

/Z/h1

~

~~

Disfavoured by BR (b s) = (3.2 0.5) 10-4 (CLEO, BELLE)

0.094 h2 0.129 (WMAP)

Slepton Co-annihilation region: LSP ~ pure Bino. Small slepton-LSP mass difference makes measurements difficult.

Also 'rapid annihilation funnel' at Higgs pole at high tan(), stop co-annihilation region at large A0


Coannihilation SignaturesCoannihilation Signatures

• ETmiss>300 GeV

• 2 OSSF leptons PT>10 GeV• >1 jet with PT>150 GeV• OSSF-OSOF subtraction applied

• ETmiss>300 GeV

• 1 tau PT>40 GeV;1 tau PT<25 GeV• >1 jet with PT>100 GeV• SS tau subtraction

• Small slepton-neutralino mass difference gives soft leptons – Low electron/muon/tau energy thresholds

crucial.

• Study point chosen within region:– m0=70 GeV; m1/2=350 GeV; A0=0; tanß=10 ;

μ>0;– Same model used for DC2 study.

• Decays of 02 to both lL and lR

kinematically allowed.– Double dilepton invariant mass edge

structure;– Edges expected at 57 / 101 GeV

• Stau channels enhanced (tan)– Soft tau signatures;– Edge expected at 79 GeV;– Less clear due to poor tau visible energy

resolution.

ATLAS

ATLAS

Preliminary

Preliminary

100 fb-1

100 fb-1

~~~



• Two edges from:

01

02

~ llll R

01

02

~ llll L Each slepton is close in mass to one of the neutralinos – one of the leptons is soft

20.6 fb-1

No cuts20.6 fb-1

No cuts

MC Truth, lL

MC Truth, lR

MC Data

Hard lepton

Soft leptonATLAS

Preliminary

ATLAS

Preliminary

•m0 = 70 GeV•m1/2 = 350 GeV•A0 = 0•tan() = 10•sgn() = +1


Focus Point ModelsFocus Point Models• Large m0 sfermions are heavy

• Most useful signatures from heavy neutralino decay• Study point chosen within focus point region :

– m0=3550 GeV; m1/2=300 GeV; A0=0; tanß=10 ; μ>0

• Direct three-body decays 0n → 0

1 ll

• Edges give m(0n)-m(0

1) : flavour subtraction applied

~ ~~ ~

Z0 → ll

300 fb-1

→

ll~ ~

→ ll

~ ~

Preliminary

ATLAS

Parameter Without cuts Exp. value

M 68±92 103.35

M2-M1 57.7±1.0 57.03

M3-M1 77.6±1.0 76.41

M = mA+mB mA-mB


SUSY Spin MeasurementSUSY Spin Measurement• Q: How do we know that a SUSY signal is really due to SUSY?

– Other models (e.g. UED) can mimic SUSY mass spectrum

• A: Measure spin of new particles.• One proposal – use ‘standard’ two-body slepton decay chain

– charge asymmetry of lq pairs measures spin of 02

– relies on valence quark contribution to pdf of proton (C asymmetry)

– shape of dilepton invariant mass spectrum measures slepton spin

~

150 fb -1

spin-0=flat

Point 5

ATLAS

ll

llA

mlq

Straightline distn

(phase-space)

Point 5

Spin-½, mostly winoSpin-0

Spin-½

Spin-0

Spin-½, mostly bino

Polarise

MeasureAngle

ATLAS

150 fb -1


Preparations for 1st PhysicsPreparations for 1st Physics• Preparations needed to ensure efficient/reliable searches

for/measurements of SUSY particles in timely manner:– Initial calibrations (energy scales, resolutions, efficiencies etc.);– Minimisation of poorly estimated SM backgrounds;– Estimation of remaining SM backgrounds;– Development of useful tools.– Definition of prescale (calibration) trigger strategy

• Different situation to Run II (no previous measurements at same s)• Will need convincing bckgrnd. estimate with little data as possible.• Background estimation techniques will change depending on

integrated lumi. • Ditto optimum search channels & cuts.• Aim to use combination of

– Fast-sim;– Full-sim;– Estimations from data.

• Use comparison between different techniques to validate estimates and build confidence in (blind) analysis.


Background EstimationBackground Estimation• Inclusive signature: jets + n leptons + ET

miss

• Main backgrounds:– Z + n jets– W + n jets – QCD– ttbar

• Greatest discrimination power from ETmiss

(R-Parity conserving models)• Generic approach to background

estimation:– Select low ET

miss background calibration samples;

– Extrapolate into high ETmiss signal region.

• Used by CDF / D0• Extrapolation is non-trivial.

– Must find variables uncorrelated with ETmiss

• Several approaches being developed.

Jets + ETmiss + 0 leptons

ATLAS

10 fb-1


W/Z+Jets BackgroundW/Z+Jets Background• Z + n jets, W l + n jets, W +

(n-1) jets ( fakes jet)• Estimate from Z l+l- + n jets (e or )• Tag leptonic Z and use to validate MC

/ estimate ETmiss from pT(Z) & pT(l)

• Alternatively tag W l + n jets and replace lepton with 0l): – higher stats

– biased by presence of SUSY

ATLAS

Preliminary

ATLAS

Preliminary

(Zll)

(Wl)


QCD BackgroundQCD Background• Two main sources:

– fake ETmiss (gaps in acceptance, dead/hot

cells, non-gaussian tails etc.)– real ET

miss (neutrinos from b/c quark decays)

• Much harder : simulations require detailed understanding of detector performance (not easy with little data).

• Strategy:1) Initially choose channels which minimise

contribution until well understood (e.g. jets + ET

miss + 1l)

2) Reject events where fake ETmiss likely:

beam-gas and machine background, bad primary vertex, hot cells, CR muons, ET

miss phi correlations (jet fluctuations), jets pointing at regions of poor response …

3) Cut hard to minimise contribution to background (at expense of stats).

4) Estimate background using data

Pythia dijets

SUSY SU3


QCD BackgroundQCD Background• Work in Progress: several ideas under study• Step 1: Measure jet smearing function from data

– Select events: ETmiss > 100 GeV, (ET

miss, jet) < 0.1

– Estimate pT of jet closest to EtMiss as

pTtrue-est = pT

jet + ETMiss

• Step 2: Smear low ETmiss multijet events with

measured smearing function

MET

jets

fluctuatingjet

Njets >= 4, p

T(j1,j2) >100GeV,

pT(j3,j4) > 50GeV

ETmiss

NB: error bars expected errors on background.

ATLAS

Preliminary

ATLAS

Preliminary


Top BackgroundTop Background• Jets + ET

miss + 1 lepton: cut hard on mT(ET

miss, l)

• Rejects bulk of ttbar (semi-leptonic decays), W l + n jets

• Remaining background mostly fully leptonic top decays (ttbbllttbblttbb) with one top missed (e.g. hadronic tau)

Edge at W Mass

Type of Event Number of Events

Electron-Electron 67

Muon-Muon 80

Tau-Tau 14

Electron-Muon 152

Electron-Tau 100

Muon-Tau 109

Electron-Hadron 77

Muon-Hadron 174

Tau-Hadron 7

ATLAS

Preliminary


Decay ResimulationDecay Resimulation• Work in progress: several approaches under study• Goal is to model complex background events using samples of

tagged SM events.• Initially we will know:

– a lot about decays of SM particles (e.g. W, top)– a reasonable amount about the (gaussian) performance of the detector.– rather little about PDFs, the hard process and UE.

• Philosophy– Tag ‘seed’ events containing W/top– Reconstruct 4-momentum of W/top (x2 if e.g. ttbar)– Decay/hadronise with e.g. Pythia– Simulate decay products with atlfast(here) or fullsim– Remove original decay products from seed event– Merge new decay products with seed event (inc. ET

miss)– Perform standard SUSY analysis on merged event

• Technique applicable to wide range of channels (not just SUSY)• Involves measurement of (top) pT: useful for exotic-top searches?


Decay ResimulationDecay Resimulation• Select seed events:

– 2 leptons with pT>25GeV

– 2 jets with pT>50GeV

– ETmiss < <pT(l1),pT(l2)> (SUSY)

– Both m(lb) < 155GeV

• Solve 6 constraints for p()• Resimulate + merge with seed• Apply 1l SUSY cuts (but Njet >=2)

m(lb)

cut

Pull (top pT)

Sample 5201,pT(top) > 200 GeVreco 11.0.4205,450 pb-1

GeV

Z

Estimate5201

ETmiss

Estimate requires >=2 jets, Normalisation under study

ATLAS

Preliminary

ATLAS

Preliminary

ATLAS

Preliminary


SupersummarySupersummary• The LHC will be THE PLACE to search for, and hopefully

study, SUSY from 2007 onwards at colliders (at least until ILC).

• Complementary to direct/indirect astroparticle dark matter searches– ATLAS: evidence / limits on SUSY– Astroparticle: evidence / limits on DM

• SUSY searches will commence on Day 1 of LHC operation.

• Many studies of exclusive channels already performed.• Lots of input from both theorists (new ideas) and

experimentalists (new techniques).• Big challenge for discovery will be understanding

systematics.• Big effort ramping up now to understand how to exploit

first data in timely fashion


SupersymmetrySupersymmetry


BACK-UP SLIDES


• Reconstruct top mass in semi-leptonic top events

• Select events in peak and examine ETmiss distribution.

• Subtract combinatorial background with appropriately weighted (from MC) top sideband subtraction.

Histogram – 1 lepton SUSY selection (no b-tag)Data points – background estimate

ttbar

ATLAS

Preliminary

SUSY

ATLAS

Preliminary

• Good agreement with top background distribution in SUSY selection.

• With this tuning does not select SUSY events (as required)

With btag

With btag

Top Background to SUSYTop Background to SUSY


Top Background to SUSYTop Background to SUSY• Fully simulated data.• No btag used• Background estimates altered by increased sideband

• Large excess at ETmiss > 500 GeV

• With 10 fb-1 obtain following no. events passing ETmiss cut:

– No SUSY: Observed: 50 ± 7 (stat), Estimate: 55 ± 17(stat)

– With SUSY: Observed: 503 ± 22 (est. stat), Estimate: 7 ± 35 (est. stat)

ATLASPreliminary

T2 + SU3

EstimateSUSY selection (top)SUSY selection (total)

ATLASPreliminary

T2

EstimateSUSY selection

Work in progress


qll edge

qll threshold

ql(min) edge

ql(max) edge

20.6 fb-1

No cuts

ATLAS

Preliminary

ATLAS

Preliminary

ATLAS

Preliminary

ATLAS

Preliminary


•m0 = 70 GeV•m1/2 = 350 GeV•A0 = 0•tan() = 10•sgn() = +1


Full Simulation: Mass FitFull Simulation: Mass Fit• Fit to all distributions. Expected shape, acceptance corrections,

detector resolution to be accounted for.• Need to understand cuts to reject SM background • Assume 1% error on lepton+jets endpoints

20.6 fb-1

No cuts

M(01) (GeV) M(0

1) (GeV)

M(

02)

(GeV

)

ATLAS

Preliminary

ATLAS

Preliminary


llq Edgellq Edge

• Hardest jets in each event produced by RH or LH squark decays.

• Select smaller of two llq invariant masses from two hardest jets– Mass must be < edge position.

• Edge sensitive to LH squark mass.

~~

~

l ll

qL

q

~

• Dilepton edges provide starting point for other measurements.

• Use dilepton signature to tag presence of 02 in event, then work back

up decay chain constructing invariant mass distributions of combinations of leptons and jets.

ATLAS

PhysicsTDR

~

e.g. LHC Point 5

1% error(100 fb-1)

Point 5


lq Edgelq Edge• Complex decay chain at LHC Point 5 gives

additional constraints on masses.• Use lepton-jet combinations in addition to

lepton-lepton combinations.• Select events with only one dilepton-jet

pairing consistent with slepton hypothesis Require one llq mass above edge and one

below (reduces combinatorics).

• Construct distribution of invariant masses of 'slepton' jet with each lepton.

• 'Right' edge sensitive to slepton, squark and 0

2

masses ('wrong' edge not visible).

~

ATLAS

ATLAS

PhysicsTDR

PhysicsTDR

1% error(100 fb-1)

Point 5

Point 5


hq edgehq edge• If tan() not too large can also observe two body decay of 0

2 to higgs and 0

1.

• Reconstruct higgs mass (2 b-jets) and combine with hard jet.• Gives additional mass constraint.

~

~

bh

qL

q

~

b

PhysicsTDR

Point 5

~~

ATLAS

1% error(100 fb-1)


llq Thresholdllq Threshold• Two body kinematics of slepton-

mediated decay chain also provides still further information (Point 5).

• Consider case where 01 produced

near rest in 02 frame.

Dilepton mass near maximal. p(ll) determined by p(0

2).

• Distribution of llq invariant masses distribution has maximum and minimum (when quark and dilepton parallel).

• llq threshold important as contains new dependence on mass of lightest neutralino.

~~

~

ATLAS

ATLAS

PhysicsTDR

PhysicsTDR

2% error(100 fb-1)

Point 5

Point 5


Mass ReconstructionMass Reconstruction

• Combine measurements from edges from different jet/lepton combinations.

• Gives sensitivity to masses (rather than combinations).


High Mass mSUGRAHigh Mass mSUGRA• ATLAS study of sensitivity to models

with high mass scales• E.g. CLIC Point K Potentially

observable … but hard!

• Characteristic double peak in signal Meff distribution (Point K).

• Squark and gluino production cross-section reduced due to high mass.

• Gaugino production significant

ATLAS

1000 fb -1


AMSBAMSB• Examined RPC model with

tan() = 10, m3/2=36 TeV, m0=500 GeV, sign() = +1.

• +/-1 near degenerate with 0

1.

• Search for +/-1 +/-0

1

(m = 631 MeV soft pions).

• Also displaced vertex due to phase space (c=360 microns).

• Measure mass difference between chargino and neutralino using mT2 variable (from mSUGRA analysis).

~ ~

~~


GMSBGMSB• Kinematic edges also useful for GMSB models when neutral

LSP or very long-lived NLSP escapes detector.• Kinematic techniques using invariant masses of

combinations of leptons, jets and photons similar.• Interpretation different though.• E.g. LHC Point G1a (neutralino NLSP with prompt decay to

gravitino) with decay chain:

~~

~

ll l

~

G


GMSBGMSB• Use dilepton edge as before (but different position in chain).• Use also l, ll edges (c.f. lq and llq edges in mSUGRA).• Get two edges (bonus!) in l as can now see edge from 'wrong' lepton

(from 02 decay). Not possible at LHCC Pt5 due to masses.

• Interpretation easier as can assume gravitino massless:


R-Parity ViolationR-Parity Violation

• Use modified effective mass variable taking into account pT of leptons and jets in event

• Missing ET for events at SUGRA point 5 with and without R-parity violation

• RPV removes the classic SUSY missing ET signature

mT ,cent pTjet ,lepton

2



• Baryon-Parity violating case hardest to identify (no leptons).

– Worst case: "212 - no heavy-quark jets

• Test model studied with decay chain:

• Lightest neutralino decays via BPV coupling:

• Reconstruct neutralino mass from 3-jet combinations (but large combinatorics : require > 8 jets!)

Phase space sample 8j +2l

˜ 10 cds

˜ q L ˜ 20q ˜ l Rlq ˜ 1

0llq



˜ 20

˜ 10R

~l

l l

qq

q

q

Lq~ Test point

Decay via allowed where m( ) > m( ) ˜ 2

0

R

~lR

~l

• Use extra information from leptons to decrease background.

• Sequential decay of to through and producing Opposite Sign, Same Family (OSSF) leptons

˜ 10

˜ 20

Lq~

R

~l



Gaussian fit: = 118.9 3 GeV, (116.7 GeV) = 218.5 3 GeV (211.9 GeV)m( ˜ 1

0 )m( ˜ 2

0 )

• Jet energy scale uncertainty 3% 3 GeV systematic

No peak in phase space sample

• Perform simultaneous (2D) fit to 3jet and 3jet + 2lepton combination (measures mass of 0

2).

• Can also measure squark and slepton masses.

~


R-Parity ViolationR-Parity Violation• Different ”ijk RPV couplings

cause LSP decays to different quarks:

• Identifying the dominant ” gives insight into flavour structure of model.

• Use vertexing and non-isolated muons to statistically separate c- and b- from light quark jets.

• Remaining ambiguity from d s• Dominant coupling could be

identified at > 3.5

˜ 10 q1q2q3

Combined2 / df P / % 2 / df P / %

uds udb 59.1/1 - 28.7/1 - 9.4usb 73.0/1 - 31.7/1 - 10.2cds 30.5/1 - 4.0/1 4 5.9cdb 106.9/1 - 47.2/1 - 12.4csb 113.4/1 - 49.2/1 - 12.8

udb usb 1.6/2 44 0.4/1 54 1.4cds 10.3/2 1 13.0/1 - 4.8cdb 18.3/2 - 6.8/2 3 5csb 16.3/2 - 5.1/2 8 4.6

usb cds 17.5/2 - 17.2/1 - 5.9cdb 12.1/2 - 5.1/1 2 4.2csb 9.9/2 1 3.1/1 8 3.6

cds cdb 56.1/2 - 37.4/1 - 9.7csb 55.8/2 - 35.3/1 - 9.5

cdb csb 0.6/2 72 1.3/2 51 1.4

Distinguishing Vertexing Muons " ijk from " lmn


Mass Relation MethodMass Relation Method• New(ish) idea for reconstructing SUSY masses!• ‘Impossible to measure mass of each sparticle using one channel

alone’ (Slide 10).– Should have added caveat: Only if done event-by-event!

• Assume in each decay chain 5 inv. mass constraints for 6 unknowns (4 0

1 momenta + gluino mass + sbottom mass).

• Remove ambiguities by combining different events analytically ‘mass relation method’ (Nojiri et al.).

• Also allows all events to be used, not just those passing hard cuts (useful if background small, buts stats limited – e.g. high scale SUSY).

Preliminary

ATLAS

AT

LA

S

SPS1a

~


DC1 SUSY ChallengeDC1 SUSY Challenge

• First attempt at large-scale simulation of SUSY signals in ATLAS (100 000 events: ~5 fb-1) in early 2003.

• Tested Geant3 simulation and ATHENA (C++) reconstruction software framework thoroughly.

No b-tagWith b-tag

llj endpoint

ATLAS

ATLAS

ATLAS

ATLAS

SUSY Mass Scale

Dijet mT2 distribution

ll endpoint

Modified LHCC Point 5: m0=100 GeV; m1/2=300 GeV; A0=-300 GeV; tanß=6 ; μ>0

Preliminary

Preliminary

Preliminary

Preliminary


DC2 SUSY ChallengeDC2 SUSY Challenge• DC2 tested new G4 simulation and

reconstruction.• Points studied:

– DC1 bulk region point (test G4)

– Stau coannihilation point (rich in signatures - test reconstruction)

• 400k events fully simulated

ATLAS

ll endpoint

Preliminary

+- endpoint

ATLAS

Preliminary

ATLAS

Preliminary

ll endpoint

+- endpoint

DC1 Point

Coannihilation Point

DC1 Point

Coannihilation Point


Rome SUSY StudiesRome SUSY Studies• Concentrate on signatures from WMAP-compatible points

accessible with 10 fb-1 for Rome Physics Workshop (last week).• Several points studied with full simulation (~100k events/point):

– SU1: Stau Coannihilation Region (m0=70 GeV, m1/2 = 350 GeV, A0 = 0, tan() = 10, >0)

– SU2: Focus Point Region (m0=3550 GeV, m1/2 = 300 GeV, A0 = 0, tan() = 10, >0)

– SU3: Bulk Region (m0=100 GeV, m1/2 = 300 GeV, A0 = -300 GeV, tan() = 10, >0)

– SU4: Low mass region (m0=200 GeV, m1/2 = 160 GeV, A0 = -400 GeV, tan() = 10, >0)

– SU5.x: Scan (mostly m1/2 = 600 GeV): Inclusive studies

– SU6: Funnel Region (m0=320 GeV, m1/2 = 375 GeV, A0 = 0, tan() = 50, >0)

• Results hot off the press (1 week old)• More to be shown by I Aracena at SUSY2005 next month!


MethodologyMethodology• Possible approach:

– Select semi-leptonic top candidates (standard cuts – what btag available?);

– Fully reconstruct top and W momenta using ETmiss and W mass constraint

reject (SUSY) background via reconstructed m(top);

– Estimate background at high ETmiss with events in top mass peak

– Normalise to data at low ETmiss.

• Study with Rome background (T1, T2, A7) and SUSY signal (SU3) samples.

• Used SusyPlot and AODs throughout (negative weights included).• Simplified 1 lepton SUSY selection (softer cuts to accept events / no

mT cut applied):

– ETmiss > 20 GeV (we will extrapolate to large ET

miss later)

– At least 4 jets with pT > 40 GeV

– Exactly 1 lepton with pT > 20 GeV

• Similar to top selection cuts but without b-tag requirement (assume not available initially).

• Note: nothing has been tuned excessively (no time!) almost ‘blind’.


• Cannot use hadronic top as W reconstruction efficiency poor at high pT(W) (jets become collinear)

• Reconstruct semi-leptonic top mass from lepton + ETmiss and W

mass constraint – quadratic ambiguity

– select solution nearest to top mass (only place m(top) used)

• Reconstruct top from combination of W with one hard jet – Reduce combinatorics by selecting highest pT top candidate.

Top Mass ReconstructionTop Mass Reconstruction

ATLASPreliminary


ETmiss EstimationETmiss Estimation

• M(top) reasonably uncorrelated with ET

miss.

• Select events at low ETmiss

and examine top mass distribution

• Define – signal band (140 – 200

GeV)

– sideband (200-260 GeV)

• Scale sideband by 1.57 (from W+jets MC distribution combinatorics + use of m(top))

• Assume this factor will be known with negligible stat. error from MC in practice (not the case here 16% err.)

• Assumes same factor for ‘bad’ top / SUSY

ATLASPreliminary

ATLASPreliminary

T2

A7

T1

ETmiss:

100 GeV - 200 GeV

ETmiss:

100 GeV - 200 GeV


Normalising the EstimateNormalising the Estimate• Will use signal band – sideband (scaled) ET

miss distribution as background estimate.

• Still need to normalise to SUSY selection (accounts for relative efficiency of top selection).

• Normalise to low ETmiss region (100 GeV – 200 GeV) where SUSY

signal expected to be small– Use T1 sample and assume low stats (0.5 fb-1)

• Scaling factor 4.13 ± 0.27

ATLASPreliminary

T1



• Use T2 sample (pT>500 GeV) to estimate precision.

• Count events with ETmiss > 500 GeV

in SUSY selection and background estimate.

• With 10 fb-1 obtain:– SUSY: 50 ± 7 (stat)

– Estimate: 55 ± 17(stat) 30%

Background EstimatesBackground Estimates

• With 44 fb-1 (full T2 sample) obtain:– SUSY: 174 ± 13 (stat)

– Estimate: 198 ± 38(stat) 20%

• Errors in estimates mainly from sideband stats (poor top selection efficiency)

• Contribution from statistical error in normalistion (even with 0.5 fb-1) negligible.

ATLASPreliminary

ATLASPreliminary

T2

T2




SUSYSUSY

• So what happens when a SUSY signal is present?

• Study by adding in SUSY sample SU3 (appropriately normalised).

• Repeat previous steps.

• Normalisation procedure OK for SU3 and 100 – 200 GeV window

• Normalisation factor 4.16 ± 0.27• Sideband subtraction appears to

work

ATLASPreliminary

ATLASPreliminary

T1 + SU3

SU3




• Background estimates affected by increased sideband

• Large excess at ETmiss > 500 GeV

• With 10 fb-1 obtain:– SUSY: 503 ± 22 (stat - estimate)

– Estimate: 7 ± 35 (stat - estimate)

• Actual stat error on estimate: ± 49• Clear excess (~13 )

Estimates with SignalEstimates with Signal

• With 44 fb-1 (full T2 sample) obtain:– SUSY: 2156 ± 46 (stat - estimate)

– Estimate: -15 ± 75(stat - estimate)

• Actual staterror on estimate: ± 204• Expected errors on estimate

increase due to dilution of top in signal band by SUSY (high ET

miss).

ATLASPreliminary

ATLASPreliminary

T1 + SU3

T1 + SU3


Documents

Supersymmetry at ATLAS