Early SUSY searches at the LHC - Imperial College London · 2010. 3. 24. · Institute of Physics...

Early SUSY searches at the LHCAlex Tapper

Searching for SUSY at the LHC and interplay with astroparticle physics

Institute of Physics Workshop, 24th March 2010, Imperial College London

Institute of Physics Workshop, 24th March 2010, Imperial College London.

Many different SUSY scenarios investigated by LHC expts

My brief is to describe plans for early SUSY searches

What we plan to do with the 2010 and 2011 data

Covering discovery not really mass, spin etc. determination

2

Introduction


The Large Hadron Collider

3

p

7 TeV

p

7 TeV



4

p

3.5 TeV

p

3.5 TeV



5

p

3.5 TeV

p

3.5 TeV

M. Fero-Luzzi

100 pb-1 by end of 2010

1 fb-1 by end of 2011


The ATLAS detector

6


The ATLAS detector

7


The CMS detector

8

Superconducting Solenoid (4 Tesla)

Muon Detectors

Plastic-Brass Hadronic Calorimeter

Lead TungstateElectromagnetic Calorimeter

Silicon Microstrip Tracker Silicon Pixel Detector

Iron Yoke


The CMS detector

9


Search strategy

Be as model independent as possible But the MSSM has > 100 parameters Need more constrained models Choose a set of benchmark points that

are representative of a range of topologies and areas of phase space

Range of models• MSUGRA (high and low masses)• GMSB• Split SUSY

In this talk MSUGRA at low masses, just above the Tevatron

SU4 for ATLAS

10

Full details of benchmark points in backup slides


0

200

400

600

800

1000

1200

1400

0 200 400 600 800 1000 1200 1400 1600 1800 20000

200

400

600

800

1000

1200

1400

0

200

400

600

800

1000

1200

1400

0 200 400 600 800 1000 1200 1400 1600 1800 2000

0 200 400 600 800 1000 1200 1400 1600 1800 2000

m0 (GeV)

m1/

2 (G

eV)

MSUGRA, tan! = 10, A0 = 0, µ > 0

" ~

1 LSP

NO EWSB

m(e ~ L

)<m

(#20 )

m(u

~ L) >

m(g

~ )

m(t~

1) < m(g

~ )

Tevatron

mh = 114 GeV

mh = 120 GeV

mh = 122 GeV

m# = 103 GeVl~

Br( #~

20!h0#

~10) > 0.5

Br( #~

20!Z0#

~10) > 0.5

Br(

#~ 20 !l ~ l)

> 0

.15

LM1

LM2

LM3LM4

LM5LM6

LM7

LM8

LM9

LM10

HM1HM2 HM3

HM4

4 56

7

8

9

Search strategy

Be as model independent as possible But the MSSM has > 100 parameters Need more constrained models Choose a set of benchmark points that

are representative of a range of topologies and areas of phase space

Range of models• MSUGRA (high and low masses)• GMSB• Split SUSY

In this talk MSUGRA at low masses, just above the Tevatron

LM0 and LM1 for CMS

11

Full details of benchmark points in backup slides

J. Phys. G: Nucl. Part. Phys. 34 (2006)


Search strategy

Production Squark and gluino expected to dominate Strong production so high cross section Cross section depends only on masses Approx. independent of SUSY model

12


Search strategy

Production Squark and gluino expected to dominate Strong production so high cross section Cross section depends only on masses Approx. independent of SUSY model

13

Decay Details of decay chain depend on SUSY model (mass spectra, branching ratios, etc.) Assume RP conserved ➔ decay to lightest SUSY particle (LSP) Assume squarks and gluinos are heavy ➔ long decay chains

Signatures MET from LSPs, high-ET jets and leptons from long decay chain

Focus on robust and simple signatures Common to wide variety of models Let Standard Model background and detector performance define searches not models


Searches

How might such a generic search look? Simple selection ➔ categorise events by numbers of leptons and jets

Good S/B for most channels (200 pb-1 @ 10 TeV COM) but... Backgrounds straight from Monte Carlo

Measuring backgrounds is the key ➔

14

• Jet ET > 100 (40) GeV• ΔΦ(jeti,MET) > 0.2 rad• Lepton ET > 20 (10) GeV • MET > 80 GeV• Meff = ΣETjet + ΣETlep + MET• MET > 0.2-0.3 x Meff

• ST>0.2• MT > 100 GeV

ATL-PHYS-PUB-2009-084


Backgrounds

Physics Standard Model processes that give the same signatures as SUSY Cannot rely on Monte Carlo predictions ➔ measure in data

Detector effects Detector noise, mis-measurements etc. that generate MET or extra jets Commissioning and calibration ➔ results from 2009 pilot run

Beam related Beam-halo muons (and cosmic-ray muons), beam-gas events Data and simulation already ➔ measure in situ too

15


Commissioning and calibration Universal astonishment at how well the simulation describes the data Detector noise under study ➔ promising start in understanding

16

[GeV]missTE

0 2 4 6 8 10 12 14 16 18 20

Nor

mal

ized

eve

nts

[/0.2

5 G

eV]

1

10

210

310

410 dataSingle Gaussian noise model

Cells, |E|>2

dataSingle Gaussian noise model

Topoclusters 4/2/0

ATLAS Preliminary

[GeV]TECalo 0 10 20 30 40 50 60 70 80

Num

ber o

f Eve

nts

/ GeV

1

10

210

310

410

[GeV]TECalo 0 10 20 30 40 50 60 70 80

Num

ber o

f Eve

nts

/ GeV

1

10

210

310

410

No cleaning

After HF noise cleaning

After HF+ECAL noise cleaning

=900 GeVsCMS Preliminary 2009


Commissioning and calibration

17

[GeV]x,yEpf-50 -40 -30 -20 -10 0 10 20 30 40 50

Num

ber o

f eve

nts

/ GeV

!2

-110

1

10

210

310

410

510Data

Simulation

CMS Preliminary 2009=900 GeVs

(GeV)T E0 10 20 30 40 50 60 70

MET

x, M

ETy

Res

olut

ion

(GeV

)

00.5

11.5

22.5

33.5

44.5

5

=900 GeVsData

=2.36 TeVsData

=900 GeVsNon-diffractive MC MinBias

T EData: fit 0.38

T EMC_minbias: fit 0.37

T EData: fit 0.37

ATLAS Preliminary

Universal astonishment at how well the simulation describes the data Resolutions well described by simulation


Commissioning and calibration

18

(GeV)T

jet p10 20 30 40 50 60

Jets

/ G

eV

-110

1

10

210

310CMS preliminary

Inclusive Jets (R = 0.5) PFJetsTanti-k

(jet) > 10 GeVT

p

(jet) | < 3.0η|

= 900 GeVs

data

MC

Universal astonishment at how well the simulation describes the data Higher-level objects looking good ➔ jet pT spectrum


Physics backgrounds Data-driven background estimates are the key challenge in early

SUSY searches

General idea is find a control region where SM is dominant and use this to predict SM background in signal region

Two approaches pursued: Matrix (ABCD) methods ➔ playing variables off against each other Replacement methods ➔ modify SM with same topology as signal to predict signal

In both cases need to identify clean SM control region Difficult to avoid using Monte Carlo in some way

Will discuss searches giving examples of data-driven methods ➔

19


All-hadronic search

All-hadronic search highly sensitive to SUSY But suffers from many backgrounds Nice examples of backgrounds both from detector effects

and from Standard Model physics

20



All-hadronic search

Mis-measurement of a jet leads to MET along the jet axis Remove with ΔΦ(jeti,MET) > 0.2 rad

Several methods developed to predict MET tail from QCD events Matrix methods to estimate from control regions Smearing method ➔

21

arXiv:0901.0512 (2009)


All-hadronic search

• Derive Gaussian part of smearing function from γ + jet control sample

• Derive non-Gaussian part from Mercedes events, requiring that the MET is co-linear with one of the jets

• Combine smearing functions, normalising with di-jet sample

• Apply smearing function to low MET events to predict the tail in the high MET signal region

22

arXiv:0901.0512 (2009)


All-hadronic search

A novel approach combining angular and energy measurements No dependence on MET ➔ robust for early LHC running Originally proposed for di-jet events ➔ generalised up to 6 jets Perfectly balanced events have αT=0.5 (cut at αT>0.55) Mis-measurement of either jet leads to lower values

23

LSPLSP

jet jetjet

jet

€

αT=ET j2

MT j1 j 2

=ET j2 /ET j1

2(1− cosΔϕ)

PRL101:221803 (2008) & CMS-PAS-SUS-09-001

Barr and Gwenlan arXiv:0907.2713


Data-driven background estimates Find a control region in phase space where SM

background dominates Use measurements in this region to infer SM

background in signal region Example Z → νν + jets ➔ irreducible background Replacement technique

Background estimates

24

Z

µ µ

W

µ ν

γ

Z

νν

MET

Z → ll + jetsStrength: very cleanWeakness: low statistics

W → lν + jetsStrength: larger statisticsWeakness: background from SM and SUSY

γ + jetsStrength: large statistics and clean at high ET

Weakness: background at low ET, theoretical errors



Select γ + ≥3 jets with Eγ>150 GeV Clean sample S/B>20 Remove photon from the event Recalculate MET Normalise with σ(Z+jets)/σ(γ+jets) from

MC or measurements

25

100 pb-1 @ 14 TeV COM

CMS-PAS-SUS-08-002


Single-lepton search

Requiring one lepton (e or µ) suppresses QCD background powerfully Highly sensitive to SUSY Backgrounds come from Standard Model processes with neutrinos ➔ real MET In particular top and W decays

26




27

Data-driven background estimates Find a control region in phase space where SM background dominates Use measurements in this region to infer SM background in signal region Example W, top backgrounds to single-lepton search Playing two discriminate quantities off against each other

Well known matrix (MT) method Use low MT control region Predict MET spectrum Weaknesses

• Non-independence of variables• Signal contamination

More sophisticated methods ➔



Background estimates “Tiles” method

Use the Monte Carlo prediction for the shapes of SM backgrounds Assume independence of variables for signal

Can express Nevts in each region in terms of fSM and fSUSY

Take fSM from MC for each region and solve the system of linear equations

Predicts the number of SM background and SUSY signal events in each region Background prediction not biassed by signal contamination

28

[GeV]effM200 400 600 800 1000 1200 1400 1600 1800 2000

[GeV

]T

M

0

100

200

300

400

500

600

SM

A

B

C

D

ATLAS Preliminary

[GeV]effM200 400 600 800 1000 1200 1400 1600 1800 2000

[GeV

]T

M

0

100

200

300

400

500

600

SUSY (SU3)

A

B

C

D

ATLAS Preliminary

Figure 2: Transverse mass (MT ) versus effective mass (Meff) distributions for simulated SM backgroundevents (left) and SUSY SU3 events (right). Indicated by the capital letters are the 2!2 tiles determinedby the cross borders along Meff = 800 GeV and MT = 100 GeV. The correlation coefficients are 6.6%(SM) and 10.7% (SU3).

hypothesis is excluded, the signal events must be distributed differently from the SM background, oth-erwise their discrimination from background would not be possible. On the other hand, if no significantsignal is present, a distribution of signal events among the tiles cannot be determined so that also the sig-nal abundance itself is undetermined. The no-signal case is therefore not detected by a vanishing signalyield (which can be anything), but by a solution of the tiles method (either analytical, or via a fit) that isapproximately independent of the signal yield that is assumed.3) The no-signal case is effectively equalto the case where signal and background distributions are indistinguishable. Both cases would exhibitanticorrelations close to unity between the signal and background yields returned by the method, whosesum must be equal to the number of observed events.

With the above assumptions, the Tiles method has remarkable features as outlined below.

• The overall SM background event yields for one or several inclusive background components, andthe overall inclusive beyond-SM event yield are fully derived by the method.

• No assumption is made about the distribution of signal events among the tiles, thereby excludingany prejudice about background domination in particular tiles.

• The Tiles method also determines the signal event fraction in each tile, thus providing a signalshape estimate within the chosen granularity of the tiles.

• If the model consists of more than 2!2 tiles, the unknowns are overconstrained and the assump-tions can be tested via a log-likelihood test statistics.

• If the model consists of more than 2!2 tiles, parts of the model assumptions can be relaxed toimprove the goodness of the model.

We first discuss the minimum 2!2 tiles setup, before generalising the approach to n!n tiles. Variousconfigurations are studied using toy experiments. Systematic uncertainties are evaluated by varying theMC composition and shape. Within the context of this note, we limit ourselves to the one-lepton searchchannel and choose the event variables MT and Meff to segment the data into tiles. The Tiles method isapplied to preselected data samples including a minimum EmissT requirement (cf. Section 2.3 without theMT requirement).3) In other words, the !lnL difference between free signal yield and signal yield fixed to zero is insignificant (cf. Section 3.3).

6



Di-lepton searches

29

Low yields but very interesting properties Same sign searches

Very low Standard Model background rate Backgrounds from charge mis-identified top events (QCD in τ channel)

Opposite sign Use opposite-sign, opposite-flavour sample to subtract SM background



Di-lepton searches

Fit ee, µµ and eµ distributions simultaneously Resolution function and efficiencies from data 200 pb-1 @ 10 TeV Di-leptonic end-point mll,max=51.3 ± 1.5 (stat.) ± 0.9 (syst.) GeV [52.7 GeV]

Nice example of what could be done with modest dataset

30

CMS-PAS-SUS-09-002


Discovery reach @ 10 TeV

Scan Meff cut for best sensitivity (50% error on backgrounds) All-hadronic and single-lepton searches vie for highest sensitivity Clear discovery potential beyond the Tevatron with 200 pb-1 @ 10 TeV

31



Discovery reach @ 7 TeV

Discovery reach for single-lepton + jets + MET channel

Need to get above the 400 GeV line to be competitive

Possible with ~100 pb-1 @ 7 TeV

32Marie-Hélène Genest EPS 2009 !"#$%$&

'())*+*,-

.*,-+*/0)/12334*,*+5(*36

'(3.07*+8

3*,3(-(7(-8

!033(9:*4

)0+4;<;=>4(,42,41;<?@A 10B*:4C(-D

*EF2:

12334

3EF2+G3

2,B45:F(,03"46

HD*4:*!-0,4!:F34I*-34!:F341(33(,5/JH4.D2,,*:

(3

*1!:08*B>4234-D(3

3D0F:B

9* F,B*+3-00B

10+*4+2!(B:8

-D2,

-D*43-2-(3-(.2::8

10+*4!0C*+)F:

(,.:F3(7*4

I*-3K1(33(,5/J-4.D2,,*:"46

HD*4.F++*,-

!F9:(3D*B

H*72-+0,

:(1(-3

(,4-D(3

10B*:42+*42+0F,B

LMM4?*N

Chamonix 2009


Summary Early searches based on robust generic signatures

Sensitive as possible to a variety of new physics models

Detectors in great shape after 2009 LHC pilot run Commissioning progressing well

A wide range of data-driven techniques developed to measure efficiencies and backgrounds Redundancy builds confidence

Eagerly awaiting start of 7 TeV LHC collisions next week! LHC should be at the search frontier before the end of 2010

33


Backup: Links

ATLAS latest results https://twiki.cern.ch/twiki/bin/view/Atlas/AtlasResults

ATLAS Physics TDR http://cdsweb.cern.ch/record/1125884?ln=en

CMS latest results https://twiki.cern.ch/twiki/bin/view/CMS/PhysicsResults

CMS Physics TDR http://cmsdoc.cern.ch/cms/cpt/tdr/

34


Backup: Benchmark points

35

Early SUSY searches at the LHC - Imperial College London · 2010. 3. 24. · Institute of Physics...

Documents

Transcript of Early SUSY searches at the LHC - Imperial College London · 2010. 3. 24. · Institute of Physics...