search for dijet resonance in pp collisions at cmssertac/sertac_work.pdf · sertac ozturk search...

Search for Dijet Resonance in pp Collisions

at CMS

1

Sertac Ozturk Cukurova University / LPC

Sertac Ozturk

About Myself

✓ I am graduate student at Cukurova University, in Turkey.

✓ I have been at LPC for 2 years and my Ph.D. thesis research has been being done at LPC.

‣ Search for new physics with jets with Dr. Robert M. Harris

‣ HCAL with Dr. Shuichi Kunori

✓ I plan to graduate in December.

2

Sertac Ozturk

Search for Dijet Resonance

A MC based analysis was done for 10 TeV collisions and a CMS reviewed paper (PAS QCD 2009-006) and an analysis note (CMS AN-2009/070) were written

The results based on 120 nb-1 (PAS EXO 2010-001) were approved for ICHEP 2010

The results based on 836 nb-1 data (PAS EXO 2010-010) were approved for HCP 2010

The results based on 2.88 pb-1 data will be submitted to PRL (hopefully this Friday) and it will cover my Ph.D. thesis.

My talk will be based on the latest data.

3

Sertac Ozturk

Dijet in Standard Model• What is a Dijet?

✓ Dijet results from simple 2→2 scattering of “partons”

✓ Dijets are events which primarily consist of two jets in the final state.

• We search for the new particles in “Dijet Mass” spectrum

✓ If a resonance exist, It can show up as a bump in Dijet Mass spectrum

• Dijet Mass from final state

Dijets in Standard Model• What is a dijet?

• Parton Level

✓ Dijet results from simple 2→2 scattering of “partons”

✓ quarks, anti-quarks and gluons

• Particles Level

✓ Partons come from colliding protons

✓ The final state partons become jets of observable particles via the following chain of events

‣ The partons radiate gluons.

‣ Gluons splits into quarks and antiquarks

‣ All colored object “hadronize” into color neutral particles

‣ Jet made of π, k, p, n, etc

• Dijets are events which primarily consist of two jets in the final state.

Jet

Jet

Particle Level

6

m =�

(E1 + E2)2 − (�p1 + �p2)2

4

61

D Event Displays and Table of High Mass Dijet Events725

Figure 38: Lego (left) and ρ− φ (right) displays of the 1st to 3rd Highest Masss Dijet Events

Sertac Ozturk

Resonance Models

• The models which are considered in this analysis are listed.

✓ Produced in “s-channel”

✓ Parton-Parton Resonances

‣ Observed as dijet resonances.

• Search for model with narrow width Γ.

3

Dijet Resonances

! New particles that decay to dijets" Produced in “s-channel” " Parton - Parton Resonances

# Observed as dijet resonances.

" Many models have small width !

X

q, q, g

q, q, g

q, q, g

q, q, g

Time

Spa

ce

Mass

Rat

e

Breit-Wigner

!!!!

M

1"1"2"1"1"1+!+0+J P

qq0.01SingletZ ‘Heavy Z

q1q20.01SingletW ‘Heavy W

qq,gg0.01SingletGR S Graviton

qq,gg0.01Octet#T8Octet Technirho

qq0.05OctetCColoron

qq0.05OctetAAxigluon

qg0.02Triplet q*Excited Quark

ud0.004Triplet DE6 Diquark

Chan! / (2M)Color XModel Name

Sertac Ozturk, University of Cukurova3

Dijet Resonances

! New particles that decay to dijets" Produced in “s-channel” " Parton - Parton Resonances

# Observed as dijet resonances.

" Many models have small width !

X

q, q, g

q, q, g

q, q, g

q, q, g

Time

Spa

ce

Mass

Rat

e

Breit-Wigner

!!!!

M

1"1"2"1"1"1+!+0+J P

qq0.01SingletZ ‘Heavy Z

q1q20.01SingletW ‘Heavy W

qq,gg0.01SingletGR S Graviton

qq,gg0.01Octet#T8Octet Technirho

qq0.05OctetCColoron

qq0.05OctetAAxigluon

qg0.02Triplet q*Excited Quark

ud0.004Triplet DE6 Diquark

Chan! / (2M)Color XModel Name

Sertac Ozturk, University of Cukurova

5

1.3 Summary of Experimental Technique 3

Sundrum gravitons [10], with coupling k/MPL = 0.1, from a model of large extra dimensions65

are produced from gluons or quark-antiquark pairs in the initial state (qq̄, gg → G). Heavy W66

bosons [11] inspired by left-right symmetric grand unified models have electroweak couplings67

and require antiquarks for their production(q1q̄2 → W �), giving small cross sections. Heavy68

Z bosons [11] inspired by grand-unified models are widely anticipated by theorists, but they69

are electroweakly produced, and require an antiquark in the initial state(qq̄ → Z�), so their70

production cross section is around the lowest of the models considered. The model with the71

largest cross section is a recent model of string resonances, Regge excitations of the quarks and72

gluons in open string theory, which includes resonances in all parton-parton channels (qq̄, qq,73

gg and qg) with multiple spin states and quantum numbers [12, 13]. Table 1 summarizes some74

properties of these models.75

Model Name X Color JP Γ/(2M) ChanExcited Quark q* Triplet 1/2+ 0.02 qg

E6 Diquark D Triplet 0+ 0.004 qqAxigluon A Octet 1+ 0.05 qq̄Coloron C Octet 1− 0.05 qq̄

RS Graviton G Singlet 2+ 0.01 qq̄ , ggHeavy W W’ Singlet 1− 0.01 qq̄Heavy Z Z’ Singlet 1− 0.01 qq̄

String S mixed mixed 0.003− 0.037 qq̄, qq, gg and qgTable 1: Properties of Some Resonance Models

Published lower limits [14] on the mass of these models in the dijet channel are listed in table 2.76

q∗ A or C D ρT8 W � Z� G S0.87 1.25 0.63 1.1 0.84 0.74 - -

Table 2: Published lower limits in dijet channel in TeV on the mass of new particles consideredin this analysis. These 95% confidence level exclusions are from the Tevatron [14].

77

1.3 Summary of Experimental Technique78

Our experimental technique starts with a measurement of the inclusive process pp → jet + jet +79

anything. Inclusive means we measure processes containing at least two jets in the final state,80

but the events are allowed to contain additional jets, or anything else. The dijet in the event81

is simply the two highest pt jets, the leading jets. Within the standard model our dataset is82

expected to be overwhelming dominated by the 2 → 2 process of hard parton scatters, with83

additional radiation off the initial and final state partons naturally giving additional jets. We84

do not cut away events that contain this radiation, which would reduce signals that also have85

similar amounts of radiation, and un-necessarily restrict signals to a narrow topology. The86

events can also contain additional particles, such as leptons or photons, but this will occur very87

rarely in the standard model. Finally, even more rarely within the standard model, the two88

leading CaloJets in the event can result from electrons, photons or taus producing energy in the89

calorimeter, and we do not exclude these insignificant contributions to our sample either. Our90

dijet selection is then open to many signals of new physics including high pt jets, leptons and91

photons. However, our selection is optimized for signals in the 2 → 2 parton scattering process,92

and is overwhelmingly dominated by the signal background of dijets from QCD within the93

standard model.94

Sertac Ozturk

Experimental Technique

• Measurement of dijet mass spectrum

• Comparison to PYTHIA QCD Monte Carlo prediction

• Fit of the measured dijet mass spectrum with a smooth function and search for resonance signal (bump)

• If no evidence, calculate cross section upper limit and compare with model cross section.

6

Sertac Ozturk

Data Sample• Dataset

✓ (135059-135735) - /MinimumBias/Commissioning10-SD_JetMETTau-Jun14thSkim_v1/RECO

✓ (136066-137028) - /JetMETTau/Run2010A-Jun14thReReco_v2/RECO

✓ (137437-139558) - /JetMETTau/Run2010A-PromptReco-v4/RECO

✓ (139779-140159) - /JetMETTau/Run2010A-Jul16thReReco-v1/RECO

✓ (140160-141899) - /JetMETTau/Run2010A-PromptReco-v4/RECO

✓ (141900-142664) - /JetMET/Run2010A-PromptReco-v4/RECO

• /QCDDijet_PtXXtoYY/Spring10-START3X_V26_S09-v1/GEN-SIM-RECO

• Official JSON Files

• Estimated Integrated Luminosity: 2.875 pb-1 (with 11% uncertanity)

7

Sertac Ozturk

Event Selection

8

• Trigger

✓ Technical Bit TT0 (for beam crossing)

✓ HLT_Jet50U (un-prescaled)

• Event Selection

✓ Good primary vertex

✓ At least two reconstructed jets

‣ AK7caloJets

‣ JEC: L2+L3, "Summer10" + Residual Data-Driven

✓ Require both |Jet η|< 2.5 and |Δη|<1.3

‣ Suppress QCD process significantly.

✓ Require both leading jets passing the "loose" jet id & Mjj > 220 GeV (corrected)

5. MEASUREMENT OF DIJET MASS SPECTRUM Sertac Ozturk

Dijet Mass (GeV)50 100 150 200 250 300 350 400 450

Trig

ger E

ffici

ency

0.2

0.4

0.6

0.8

1

DijetMass

= 220 GeVjjM

HLT_Jet50U

= 7 TeVs

| < 1.3! "| < 2.5 & |2!,

1!|

(GeV)T

Corrected Leading Jet P60 80 100 120 140 160

Trig

ger E

ffici

ency

0.2

0.4

0.6

0.8

1

pt50U

HLT_Jet50U = 7 TeVs

| < 1.3! "| < 2.5 & |2!,

1!|

= 105 GeVTJet P

Figure 5.2 HLT Jet50U trigger efficiency as a function of dijet mass (left) and as a function of

corrected pT of leading jet is measured in data.

5.1.2 Data Quality

The number of events in the analysis after the basic cuts are shown for each cut in

Table 5.1.

Events after pre-selection cut 6126910 100%

Events after vertex cut 6125930 99.98%

Events after dijet eta cuts 2088922 34.09%

Events after dijet mass cut 414645 6.78%

Events after jet id cut 414131 6.76%

Table 5.1 Cuts and Events

Only 514 events which are mostly HPD noise are rejected by JetID cut and the fraction

of events removed by JetID cut is very small. Because the reqirement kinematic cuts

(|η| < 2.5 and |∆η| < 1.3) and dijet mass cut (M j j > 220 GeV) gives higher the jet purity.

The distributions of the loose JetID variable are shown in Fig.5.3. Electromagnetic

fraction of jet energy, Jet EMF, doesn’t habe a peak near zero or one which indicate a

problem from ECAL and HCAL, such as hot channel. The fraction of jet energy in the

27

Events Fraction

Sertac Ozturk

Trigger Efficiency• Start analysis of Dijet Mass distribution at 220 GeV.

✓ 220 GeV chosen for full trigger efficiency

9

Dijet Mass (GeV)50 100 150 200 250 300 350 400 450

Trig

ger E

ffici

ency

0.2

0.4

0.6

0.8

1

DijetMass

= 220 GeVjjM

HLT_Jet50U

= 7 TeVs

| < 1.3η Δ| < 2.5 & |2η,

1η|

(GeV)T

Corrected Leading Jet P60 80 100 120 140 160

Trig

ger E

ffici

ency

0.2

0.4

0.6

0.8

1

pt50U

HLT_Jet50U = 7 TeVs

| < 1.3η Δ| < 2.5 & |2η,

1η|

= 105 GeVTJet P

Sertac Ozturk

Dijet Data Quality

10

η-3 -2 -1 0 1 2 3

φ

-3-2-1

012

3

Two

Lead

ing

Jets

0

50

100

150

200

)-1

CMS Data (2.875 pb

)-1CMS Data (2.875 pb

1φ

-3 -2 -1 0 1 2 3

2φ

-3

-2

-1

0

1

2

3

0

100

200

300

400

500

600

700Phi_corr

Jet EMF0 0.2 0.4 0.6 0.8 1

Two

Lead

ing

Jets

0

20000

40000

60000

80000


QCD PYTHIA

TMET/Sum E0 0.2 0.4 0.6 0.8 1

Even

ts

1

10

210

310

410

510)-1CMS Data (2.875 pb

QCD PYTHIA

Sertac Ozturk

Dijet Data Stability

• Average dijet mass and average pt of two leading jets for each runs are shown.

• There is a good stability.

11

hDijetMassRunbyRun

Entries 148Mean 74.03RMS 43.57

Run Number

1351

4913

5528

1355

7513

7028

1387

4613

8747

1387

5013

8919

1389

2113

9020

1390

9813

9100

1391

0313

9195

1392

3913

9347

1393

6513

9368

1393

7013

9372

1394

0713

9411

1394

5813

9781

1399

6713

9969

1399

7113

9973

1400

5814

0059

1401

2314

0124

1401

2614

0158

1401

5914

0160

1403

3114

0361

1403

6214

0379

1403

8214

0383

1403

8514

0387

1404

0114

1956

1419

5714

1958

1419

5914

1960

1419

6114

2035

1420

3614

2038

1420

4014

2076

1421

2914

2130

1421

3214

2135

1421

3614

2137

1421

8714

2188

1421

8914

2191

1422

6414

2265

1423

0314

2305

1423

0814

2309

1423

1114

2312

1423

1314

2317

1423

1814

2319

1424

1314

2414

1424

1514

2418

1424

1914

2420

1424

2214

2513

1425

1414

2524

1425

2514

2528

1425

3014

2535

1425

3714

2557

1425

5814

2659

1426

6014

2661

1426

6214

2663

1426

6414

2928

1429

3314

2935

1429

5314

2954

1429

7014

2971

1430

0414

3005

1430

0614

3007

1430

0814

3179

1431

8114

3187

1431

9114

3192

1431

9314

3318

1433

1914

3320

1433

2114

3322

1433

2614

3327

1433

2814

3657

1436

6514

3727

1437

3114

3827

1438

3314

3835

1439

5314

3954

1439

5514

3956

1439

5714

3959

1439

6014

3961

1439

6214

4011

1440

8614

4089

1441

1214

4114

Aver

age

Dije

t Mas

s (G

eV)

220

240

260

280

300

320

340

hDijetMassRunbyRun


DijetMassRunbyRun

HLT_Jet50U

|<1.3η Δ|<2.5 & |2η,

1η|

>220 GeVjjM

-1Run with L>1 nb

> = 279.7 GeVjj<M

hcorPt1RunbyRun


Run Number

1351

4913

5528

1355

7513

7028

1387

4613

8747

1387

5013

8919

1389

2113

9020

1390

9813

9100

1391

0313

9195

1392

3913

9347

1393

6513

9368

1393

7013

9372

1394

0713

9411

1394

5813

9781

1399

6713

9969

1399

7113

9973

1400

5814

0059

1401

2314

0124

1401

2614

0158

1401

5914

0160

1403

3114

0361

1403

6214

0379

1403

8214

0383

1403

8514

0387

1404

0114

1956

1419

5714

1958

1419

5914

1960

1419

6114

2035

1420

3614

2038

1420

4014

2076

1421

2914

2130

1421

3214

2135

1421

3614

2137

1421

8714

2188

1421

8914

2191

1422

6414

2265

1423

0314

2305

1423

0814

2309

1423

1114

2312

1423

1314

2317

1423

1814

2319

1424

1314

2414

1424

1514

2418

1424

1914

2420

1424

2214

2513

1425

1414

2524

1425

2514

2528

1425

3014

2535

1425

3714

2557

1425

5814

2659

1426

6014

2661

1426

6214

2663

1426

6414

2928

1429

3314

2935

1429

5314

2954

1429

7014

2971

1430

0414

3005

1430

0614

3007

1430

0814

3179

1431

8114

3187

1431

9114

3192

1431

9314

3318

1433

1914

3320

1433

2114

3322

1433

2614

3327

1433

2814

3657

1436

6514

3727

1437

3114

3827

1438

3314

3835

1439

5314

3954

1439

5514

3956

1439

5714

3959

1439

6014

3961

1439

6214

4011

1440

8614

4089

1441

1214

4114

(GeV

)T

Aver

age

Jet P

40

60

80

100

120

140

160

180

200

220

hcorPt1RunbyRun


corpt1RunbyRun

hcorPt2RunbyRun


HLT_Jet50U|<1.3η Δ|<2.5 & |

2η,

1η|

>220 GeVjjM

> = 143.0 GeVLeading JetT<P

> = 114.8 GeVSecond JetT<P

-1Runs with L>1 nb

Sertac Ozturk

Dijet Mass and QCD

12

• The data is in good agreement with the full CMS simulation of QCD from PYTHIA.

Dijet Mass (GeV)500 1000 1500 2000

/dm

(pb/

GeV

)σd

-310

-210

-110

1

10

210

310

410

)-1CMS Data (2.875 pbQCD Pythia + CMS Simulation10% JES Uncertainty

= 7 TeVs | < 1.3η Δ| < 2.5 & |2η, 1η|

>220 GeVjjM

Anti-kt R=0.7 CaloJets

Dijet Mass (GeV)500 1000 1500 2000

Dat

a / P

YTH

IA

1

2

3

4

5

6

1

2

3

4

5

6


10% JES Uncertainty

= 7 TeVs | < 1.3 | < 2.5 & |2, 1|

>220 GeVjjM


Sertac Ozturk

Dijet Mass and Fit• We fit the data to a function containing 4 parameters used by

CDF Run 11

13

Pulls

Dijet Mass (GeV)500 1000 1500 2000

/dm

(pb/

GeV

)σd

-310

-210

-110

1

10

210

310

410 / ndf 2χ 32.33 / 31

Prob 0.4011p0 5.398e-08± 2.609e-06 p1 0.1737± 5.077 p2 0.006248± 6.994 p3 0.001658± 0.2658

/ ndf 2χ 32.33 / 31Prob 0.4011p0 5.398e-08± 2.609e-06 p1 0.1737± 5.077 p2 0.006248± 6.994 p3 0.001658± 0.2658


Fit

= 7 TeVs | < 1.3η Δ| < 2.5 & |

2η,

1η|

>220 GeVjjM


Graph

Dijet Mass (GeV)500 1000 1500 2000

(Dat

a-Fi

t)/Fi

t

-1

0

1

2

Graph


= 7 TeVs

Dijet Mass (GeV)500 1000 1500 2000

(Dat

a-Fi

t)/Er

ror

-3

-2

-1

0

1

2

3 )-1CMS Data (2.875 pb

= 7 TeVs

5. Measurament of Dijet Mass Spectrum Sertac Ozturk

QCD MC prediction is in good agreement with the data.

Dijet Mass (GeV)500 1000 1500 2000

Dat

a / P

YTH

IA

1

2

3

4

5

6

1

2

3

4

5

6


10% JES Uncertainty

= 7 TeVs | < 1.3 | < 2.5 & |2, 1|

>220 GeVjjM


Figure 5.11 The dijet mass spectrum data (points) divided by the QCD PYTHIA prediction.

The band shows the sensitivity to a 10% systematic uncertainty on the jet energy scale.

The data points and corresponding uncertainty are listed in Table 5.X.

5.2.1 Dijet Mass Spectrum and Fit

Dijet mass spectrum is compared to a fit in Fig.5.X. The parametrization of smooth fit

function is

dσdm

= p0(1−X)p1

X p2+p3 ln(X) (5.2)

where x = m j j/√

s and p0,1,2,3 are free parameters. The (1−X) term is motivated by

the parton distribution fall of with fractional momentum. The X−p3 ln(x) factor describes

34

5. Measurament of Dijet Mass Spectrum Sertac Ozturk

QCD MC prediction is in good agreement with the data.

Dijet Mass (GeV)500 1000 1500 2000

Dat

a / P

YTH

IA

1

2

3

4

5

6

1

2

3

4

5

6


10% JES Uncertainty

= 7 TeVs | < 1.3 | < 2.5 & |2, 1|

>220 GeVjjM


Figure 5.11 The dijet mass spectrum data (points) divided by the QCD PYTHIA prediction.

The band shows the sensitivity to a 10% systematic uncertainty on the jet energy scale.

The data points and corresponding uncertainty are listed in Table 5.X.

5.2.1 Dijet Mass Spectrum and Fit

Dijet mass spectrum is compared to a fit in Fig.5.X. The parametrization of smooth fit

function is

dσdm

= p0(1−X)p1

X p2+p3 ln(X) (5.2)

where x = m j j/√

s and p0,1,2,3 are free parameters. The (1−X) term is motivated by

the parton distribution fall of with fractional momentum. The X−p3 ln(x) factor describes

34

Sertac Ozturk

Another Fit Parametrization

• In addition to the default fit, 2 alternate functional forms are considered.

• Default 4 parameters fit gives best result.

14

Dijet Mass (GeV)500 1000 1500 2000

/dm

(pb/

GeV

)σd

-310

-210

-110

1

10

210

310

410

/ ndf 2χ 32.33 / 31

p0 5.398e-08± 2.609e-06

p1 0.1737± 5.077

p2 0.006248± 6.994

p3 0.001658± 0.2658

/ ndf 2χ 32.33 / 31

p0 5.398e-08± 2.609e-06

p1 0.1737± 5.077

p2 0.006248± 6.994

p3 0.001658± 0.2658

)-1Data (2.875 pbDefault Fit (4 Par.)

Alternate Fit A (4 Par.)

Alternate Fit B (3 Par.)

= 7 TeVs | < 1.3η Δ| < 2.5 & |

2η,

1η|

>220 GeVjjM


Graph

χ2/NDF for Default Fit: 32.3 / 31 χ2/NDF for Fit A: 36.8 / 31χ2/NDF for Fit B: 39.3 / 32

2.8 Dijet Mass Spectrum and Fit 27

The parameterizations are listed in equation 3.371

dσ

dm=

P0 · (1−m/√

s )P1

mP2, (Default Fit with 3-parameters)

=P0

(P1 + m)P2, (Alternate Fit A with 3-parameters)

=P0 ·

�1−m/

√s + P3 · (m/

√s)2

�P1

mP2(Alternate Fit B with 4-parameters).

=P0 · (1−m

√s)p1

(m/√

s)p2+p3ln(m√

s)(Alternate Fit C with 4-parameters).

(3)

The default three parameter fit is motivated by QCD. It includes a power law fall off with mass372

in the denominator, motivated by the QCD matrix element. It also has a term in the numerator373

motivated by the parton distribution fall off with fractional momentum (1− m/√

s)P1 (where374 √s = 7000 GeV is the center-of-mass energy). This three parameter function was used by CDF375

in run IA. We find that the default fit gives a good χ2/DF of 17.1/18 (probability 52%), and this376

is the best fit we can find of our data.377

We have also explored three alternate parameterizations. All parameterizations have a power378

law in them, because without a power law we cannot get a good fit with only 2, 3 or 4 pa-379

rameters. A 2-parameter fit with just a power law and a constant, p0/mp1 , gives a reasonable380

fit χ2/DF = 19.3/19 (probabilty 44%), but we have been advised to only consider parame-381

terizations with the same number of parameters as our default fit or greater, in order to have382

reasonable flexibility in the fit parameterization. The 2-parameter fit has only one shape pa-383

rameter. Alternate fit A is a 3-parameter fit with a modified power law, obtained by simply384

adding an offset to the mass, and we get a good fit with χ2/DF = 17.9/18 (probability 46%).385

Alternate fit B is a 4-parameter fit very much like our default fit, but we have added a term386

quadratic in m/√

s to the term in the numberator to give the fit a little more flexibility to de-387

scribe data at high mass tails. This 4 parameter function was used by CDF in run IB [16]. We388

find that this function gives a good fit to our data, with χ2/DF of 16.8/17 (probability 47%).389

Alternate fit C is another 4 parameter function which again has our characteristic numerator390

and denominator but includes another term in the power of the power law, again just to give391

the fit more flexibiliity. This 4 parameter function was used by CDF in run II [14]. Again we392

find this function ives a good fit to our data, with χ2/DF of 16.8/17 (probability 47%).393

Figure 18 shows the fractional differences between data and the fit function, (data-fit)/fit, and394

the pulls, (data-fit)/error, for all four fits.395

Notice from both Fig. 17 and 18 that the largest difference from the default 3-parameter fit396

occurs when using the alternate fit A with 3 parameters. We will use this alternate 3-parameter397

function from fit A to find our systematic uncertainty on the background due to the fit parame-398

terization. Notice that there is very little difference between the default 3-parameter fit and the399

alternate 4-parameter fits which were introduced to give the 3-parameter fit more flexibility.400

From this we conclude that no more flexibility is needed to fit this data, and we have found the401

best possible smooth fit with a few parameters. When using these parameterizations to find402

systematic uncertainties on the background we do not find as large a systematic as with the403

alternate 3-parameter function.404



dσ

dm=

P0 · (1−m/√

s )P1


=P0


=P0 ·

�1−m/

√s + P3 · (m/

√s)2

�P1


=P0 · (1−m

√s)p1

(m/√

s)p2+p3ln(m√


(3)
















quadratic in m/√





















dσ

dm=

P0 · (1−m/√

s )P1


=P0


=P0 ·

�1−m/

√s + P3 · (m/

√s)2

�P1


=P0 · (1−m

√s)p1

(m/√

s)p2+p3ln(m√


(3)
















quadratic in m/√



















Default

Fit A

Fit B

Sertac Ozturk

Resonance Shapes• We have simulated dijet resonances using CMS simulation +

PYTHIA.

• qq, qg and gg resonances have different shape mainly due to FSR.

✓ The width of dijet resonance increases with number of gluons because gluons emit more radiation than quarks.

• We search for these three basic types of narrow dijet resonance in our data.

15Resonance Mass (GeV)500 1000 1500 2000 2500 3000 3500

/Mea

n)R

esol

utio

n (

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16 gluon-gluonquark-gluonquark-quark

[0]+[1]/x

Dijet Mass (GeV)400 600 800 1000 1200 1400 1600

Prob

abilit

y

0.05

0.1

0.15

0.2

0.25

gluon-gluonquark-gluonquark-quark

q* Resonance Shape

CMS Simulation

= 1.2 TeVResM

| < 1.3| < 2.5 & ||

Dijet Mass (GeV)500 1000 1500 2000 2500 3000

Prob

abilit

y

0

0.05

0.1

0.15

0.2

0.25q* Resonance Shape

0.5 TeV1 TeV

1.2TeV1.5 TeV

2 TeV

CMS Simulation qg)(q*

Paper

Sertac Ozturk

Dijet Mass (GeV)500 1000 1500 2000

/dm

(pb/

GeV

)d

-410

-310

-210

-110

1

10

210

310

410

/ ndf 2

32.33 / 31

Prob 0.4011

p0 5.398e-08± 2.609e-06

p1 0.1737± 5.077

p2 0.006248± 6.994

p3 0.001658± 0.2658

/ ndf 2

32.33 / 31

Prob 0.4011

p0 5.398e-08± 2.609e-06

p1 0.1737± 5.077

p2 0.006248± 6.994

p3 0.001658± 0.2658

)-1CMS Data (2.9 pbFit10% JES UncertaintyQCD Pythia + CMS SimulationExcited QuarkString = 7 TeVs

| < 1.3 | < 2.5 & ||

q* (0.5 TeV)

S (1 TeV)

q* (1.5 TeV)

S (2 TeV)

Fit and Signal• We search for dijet resonance signal in our data.

• Excited quark signals are shown at 0.5 TeV and 1.5 TeV.

• String resonance is shown at 1 TeV and 2 TeV.

16

Dijet Mass (GeV)500 1000 1500 2000

Dat

a / F

it1

10

Graph

q* (0.5 TeV)

S (1 TeV)

q* (1.5 TeV)

S (2 TeV)

)-1CMS Data (2.9 pb

= 7 TeVs

Paper Pap

er

Sertac Ozturk

The Largest Fluctuation in Data

• Upward fluctuations around 600 GeV and 900 GeV

• Best fit resonance is at 622 GeV with local significance 1.86 sigma from log likelihood ratio.

• There is no evidence for dijet resonance.

17

Dijet Mass (GeV)200 400 600 800 1000 1200 1400

(Dat

a-Fi

t)/Fi

t

-0.1

0

0.1

0.2

0.3

Graph


= 7 TeVs

q* (622 GeV)

(Background + Signal) Fit

Dijet Mass (GeV)500 1000 1500 2000

(Dat

a-Fi

t)/Er

ror

-3

-2

-1

0

1

2

3 )-1CMS Data (2.875 pb

= 7 TeVs

Sertac Ozturk

Setting Limits• For setting upper limit on the resonance production cross

section, a Bayesian formalism with a uniform prior is used.

L =�

i

µnii e−µi

ni!µi = αNi(S) + Ni(B).

• The signal comes from our dijet resonance shapes.

• The background comes from Background+Signal fit.

Measured # of events in data

# of event from signal

Expected # of event from background

18

Sertac Ozturk

Early Limits with Stat. Error Only

19

• 95% CL Upper limit with Stat. Error. Only compared to cross section for various model.

✓ Show quark-quark and quark-gluon and gluon-gluon resonances separately.

✓ gluon-gluon resonance has the lowest response and is the widest and gives worst limit.

Resonance Mass (GeV)500 1000 1500 2000 2500

Cro

ss S

ectio

n*BR

*Acc

. (pb

)

-110

1

10

210

310

410

510


= 7 TeVs | < 1.3 | < 2.5 & |

2,

1|

StringExcited quarkAxigluon

Diquark6EW’Z’RS Graviton

Gluon-GluonQuark-GluonQuark-Quark

Graph

95% CL Upper Limit

STAT. ERROR ONLYResonance Mass (GeV)500 1000 1500 2000 2500

A (p

b)×

BR

×

-110

1

10

210

310

410

510StringExcited quarkAxigluon / Coloron

Diquark6E


= 7 TeVs | < 1.3 | < 2.5 & ||

Smooth Limit with Stat. Error OnlySmooth Limit with 10% JES Uncertainty

Graph

(quark-gluon resonances)

Sertac Ozturk

Systematics

• We found the uncertainty in dijet resonance cross section from following sources.

✓ Jet Energy Scale (JES)

✓ Jet Energy Resolution (JER)

✓ Choice of Background Parametrization

✓ Luminosity

20

Sertac Ozturk

Jet Energy Scale (JES)• JetMET guidance is 10% uncertainty in jet energy scale.

✓ Shifting the resonance 10% lower in dijet mass gives more QCD background

✓ Increases the limit between 14% and 42% depending on resonance mass and type.

21

Resonance Mass (GeV)500 1000 1500 2000 2500 3000

Frac

t. C

hang

e on

Lim

it w

ith J

ES S

ys.

0.1

0.2

0.3

0.4

0.5

0.6

gluon-gluon

quark-gluon

quark-quark


= 7 TeVs

| < 1.3 | < 2.5 & |2

, 1

Jet |

Graph


A (p

b)×

BR

×

-110

1

10

210

310

410

510StringExcited quarkAxigluon / Coloron

Diquark6E


= 7 TeVs | < 1.3 | < 2.5 & ||

Smooth Limit with Stat. Error OnlySmooth Limit with 10% JES Uncertainty

Graph


Sertac Ozturk

Jet Energy Resolution (JER)• JetMET guidance is 10% uncertainty in jet energy resolution.

• We smear our resonance shapes with a gaussian designed to increase the core width by 10%.

✓ σGaus = √{(1.1)2-1} σRes

• This increases our limit between 7% and 22% depending on resonance mass and type.

22

Dijet Mass (GeV)200 400 600 800 1000 1200 1400 1600

Prob

abilit

y

0

0.02

0.04

0.06

0.08

0.1

0.12 Orginal

Convoluted

q* Resonance Shape

=1.2 TeVResM

gg)(G


Frac

t. C

hang

e on

Lim

it w

ith J

ER S

ys.

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

gluon-gluon

quark-gluon

quark-quark


= 7 TeVs

| < 1.3 | < 2.5 & |2

, 1

Jet |

Graph

Sertac Ozturk

Background Parametrization Systematics

• We have varied the choice of background parametrization

• We use the 4 parameter fit as a systematic on our background shape.

• This increases our limit between 8% and 19% depending on resonance mass and type.

23

Dijet Mass (GeV)500 1000 1500 2000

/dm

(pb/

GeV

)σd

-310

-210

-110

1

10

210

310

410

/ ndf 2χ 32.33 / 31

p0 5.398e-08± 2.609e-06

p1 0.1737± 5.077

p2 0.006248± 6.994

p3 0.001658± 0.2658

/ ndf 2χ 32.33 / 31

p0 5.398e-08± 2.609e-06

p1 0.1737± 5.077

p2 0.006248± 6.994

p3 0.001658± 0.2658

)-1Data (2.875 pbDefault Fit (4 Par.)

Alternate Fit A (4 Par.)

Alternate Fit B (3 Par.)

= 7 TeVs | < 1.3η Δ| < 2.5 & |

2η,

1η|

>220 GeVjjM


Graph


Frac

t. C

hang

e on

Lim

it w

ith B

gk S

ys.

0

0.05

0.1

0.15

0.2

0.25

0.3

gluon-gluon

quark-gluon

quark-quark


= 7 TeVs

| < 1.3 | < 2.5 & |2

, 1

Jet |

Graph

Sertac Ozturk

Total Systematic Uncertainties• We add all mentioned systematic uncertainties in quadrature, also 11% for

luminosity.

• JEC is dominant systematic uncertainty.

• Total systematic uncertainty varies from 24% to 48% depending on resonance mass and type.

24


Frac

tiona

l Unc

erta

inty

0.1

0.2

0.3

0.4

0.5

0.6

0.7 TotalJESResolutionBackgroundLuminosity

)-1CMS Preliminary (2.875 pb

= 7 TeVs

| < 1.3 | < 2.5 & |2

, 1

|

Graph

qg)(q*


Frac

tiona

l Unc

erta

inty

0.1

0.2

0.3

0.4

0.5

0.6

0.7gluon-gluon

quark-gluon

quark-quark


= 7 TeVs

| < 1.3 | < 2.5 & |2

, 1

Jet |

Graph

Sertac Ozturk

Incorporating Systematic• We convolute posterior PDF with Gaussian systematics

uncertainties.

✓ Posterior PDF including systematics is broader and gives higher upper limit.

25

30 4 Systematic Uncertainties

4.3 Background Parameterization398

We considered two others functional forms with 2 and 4 parameters to parametrize the QCD399

background as discussed in section 2.6.1 and shown in Equation 3. Fig. 25 show comparison400

of fits with the data points. We find that the 2 parameter form, which is a marginal fit to our401

data, gives the largest fractional change over the vast majority of resonance masses, and we402

conservatively use it for our background parametrization systematic at this time.403

Dijet Mass (GeV)200 400 600 800 1000 1200 1400

/dm

(pb/

GeV

)!d

-110

1

10

210

310

410

/ ndf 2" 11.25 / 16

Prob 0.7937

p0 0.0003714± 0.001181

p1 1.885± 23.43

p2 0.07701± 4.489

/ ndf 2" 11.25 / 16

Prob 0.7937

p0 0.0003714± 0.001181

p1 1.885± 23.43

p2 0.07701± 4.489

)-1Data (7.2 nb

Default Fit with 3 Par.

Fit with 4 Par.

Fit with 2 Par.

= 7 TeVs|<1.3#|Jet

>156 GeVjjM

Graph

)-1CMS Data (7.2 nb

Default Fit with 3 Par.Fit with 4 Par.Fit with 2 Par.

400 600 800 1000 1200 1400 1600

Frac

t. C

hang

e on

Lim

it w

ith B

gk S

ys.

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4


Graph

)-1CMS Data (7.2nb=7 TeVs

|<1.3!|Jet

Figure 25: Left) The data and the default 3 parameter fit and the 2 and 4 parameter fits use toevaluate the systematics. Right) Fractional absolute change in the limit when using th 2 and 4parameter fits for the background.

4.4 Total Uncertainty404

We determine 1σ change for each systematic uncertainty in signal that we can discovery or405

exclude. In addition to the sources already mentioned, we include an uncertainty of 10% on406

the integrated luminosity.407

To find total total systematics, we add the these 1σ changes as quadrature. The individual and408

total systematic uncertainties as a function of resonance mass are illustrated in Fig. 26. Absolute409

uncertainty in each resonance mass is calculated as total systematics uncertainty multiply by410

upper cross section limit.411

4.5 Incorporating Systematics in the Limit412

We convolute the posterior probability density with a Gaussian for each resonance mass. The413

equation of convolution is414

L(σ) =� ∞

0L(σ�)G(σ, σ�)dσ� (7)

Where L(σ�) is the posterior probability density at signal cross section σ�, and G(σ, σ�) is the415

Gaussian probability from systematics to observe σ if σ� is expected. The width of the Gaussian416

is taken as the absolute uncertainty in each resonance mass, equal to the fractional uncertainty417

times the limit on the cross section. This procedure, identical to what was done at CDF, con-418

servatively assigns the same width to the Gaussian in units of pb at each point in the posterior419

G: Gaussian distribution withRMS width equal to systematic

uncertainty in cross section

Cross Section (pb)0 10 20 30 40 50 60 70 80

Post

erio

r PD

F

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

Stat. Error Only

Including Systematic

Likelihood vs sigma

= 1 TeVq*M

Sertac Ozturk

Effect of Systematics on Limit

26

• 95% CL Upper limit with Stat. Error. Only and Including Sys. Uncertainties are shown separately.

• The mass limits are reduced by 0.1 TeV for both string resonance and excited quark.


Cro

ss S

ectio

n*BR

*Acc

. (pb

)

-110

1

10

210

310

410

510

610


= 7 TeVs

| < 1.3 | < 2.5 & |2

, 1

|

String

Excited quarkIncluding SystematicsStat. Error Only

Graph

95% CL Upper Limit

qg)(q*


Frac

t. C

hang

e In

clud

ing

Sys.

0.1

0.2

0.3

0.4

0.5

0.6

0.7



= 7 TeVs | < 1.3 | < 2.5 & |2, 1|

Graph

Sertac Ozturk

Results

27

• We excluded the mass limits as following:

• String

✓ 0.50<M(S)<2.50 TeV

‣ M(S)<1.40 from CDF

• Excited Quark

✓ 0.50<M(q*)<1.58 TeV

‣ 0.40<M(q*)<1.26 from ATLAS

• Axigluon/Coloron

✓ 0.50<M(A)<1.17 TeV & 1.47<M(A)<1.52 TeV

‣ 0.12<M(A)<1.25 TeV from CDF

• E6 Diquark

✓ 0.50<M(D)<0.58 TeV & 0.97<M(D)<1.08 TeV & 1.45<M(D)<1.60 TeV

‣ 0.29<M(D)<0.63 TeV from CDF


A (p

b)×

BR

×

-210

-110

1

10

210

310

410

510 )-1CMS Data (2.9 pb

= 7 TeVs | < 1.3 | < 2.5 & ||

StringExcited QuarkAxigluon/Coloron

Diquark6EW’Z’RS Graviton

Gluon-GluonQuark-GluonQuark-Quark

Graph

Resonance Models

95% CL Upper Limit

Paper

Sertac Ozturk

Expected Limit• Due to downward fluctuation around 1.2 TeV, measured limit

is 250 GeV grater than expected limit for excited quark.

• Due to upward fluctuation around 600 GeV, we lost sensitivity to a E6 diquark at low resonance mass.

28


Cro

ss S

ectio

n*BR

*Acc

. (pb

)

1

10

210

310

410


= 7 TeVs

| < 1.3 | < 2.5 & |2

, 1

|

Axigluon/Coloron

Diquark6EObserved

Expected

Graph

95% CL Upper Limit

(quark-quark resonances)


Cro

ss S

ectio

n*BR

*Acc

. (pb

)

-110

1

10

210

310

410

510

610


= 7 TeVs

| < 1.3 | < 2.5 & |2

, 1

|

String

Excited quarkObserved

Expected

Graph

95% CL Upper Limit


Sertac Ozturk

Conclusion• We have a dijet mass spectrum that extends to

2 TeV GeV with ~2.9 pb-1 data.

• The dijet mass data is in good agreement with a full CMS simulation of QCD from PYTHIA.

• There is no evidence for dijet resonances

• We extended excluded mass limit for dijet resonance models beyond Tevatron and ATLAS.

• First CMS research paper based on these result will be submitted to PRL in this week.

29

Sertac Ozturk

Back-up

30

Sertac Ozturk

Beam Splash 2009

• I spent some effort to determine HF calibration constant in 2009 Beam splash data.

31

Sertac Ozturk

Estimation of IFB Signal

32

6/21/10 Latife Vergili, Shuichi Kunori

HPD Ion feedback & pulse shape (ROC PFG Report)

8

w/o ion feedback w/ ion feedback

0.8

Mean 0.88

Low energy

High energy Ion feedback rate: 2 e-4 per p.e. 3 e-3 per GeV ~1% additional energy (in CMSSW 360 pre3 w/o time structure)

2ts/10ts

Pulse Shape without/with ion feedback

time

time time

time

hLedPulse0HBPall

Entries 6122030

Mean 3.868

RMS 1.738

TS0 1 2 3 4 5 6 7 8 9

0

0.5

1

1.5

2

2.5

hLedPulse0HBPall

Entries 6122030

Mean 3.868

RMS 1.738

h_ifbEntries 10Mean 4.537RMS 1.489

LED HBP pulse 0

Charge<40 fCCharge>60 fc

Estimated Ion FeedbackhLed_plus_HB_iphi_15


Charge (fC)0 20 40 60 80 100

1

10

210

310

410

510hLed_plus_HB_iphi_15


LED fcsum

P1 P2 P3

=15i=0,...,16i

IFB Signal = Signal in P3 - Signal in P1

• Ion feedback signal starts 1 TS later.

Sertac Ozturk

PFG task N.9 - Investigations on HBP14/4 operating at 40V

33

• Compare Pulse Shape in 80 BV, 40 BV and 30 BV

• HBP14_RM4

✓ iphi:54, depth:1, zside>0 and ieta:1,2,3, ..., 16

HBHE uses 4 TS for reconstruction.Containment becomes worse as we lower the BV.

Sertac Ozturk

Jet Commissioning with First Data

• I looked at JetID cuts criteria in 2009 MinimumBias Collision Data. (CMS AN-2010/030 & PAS JME 10-001)

34

4 Survival fraction for loose jet ID selection criteria in dijet eventsEvents having dijet topology are greatly enriched in real jets. For this reason, the same cuts on jet variables passa much higher fraction of reconstructed hadronic jets in these events than they do in the super-sample of inclusivejets. Here we present the fractional survival fractions for the two leading jets in data and in Monte Carlo.

Within this sample set, we define a dijet topology as one in which the two jets leading in transverse momentummeet a loose cut of being back-to-back in azimuth to within 1 radian. This is done to extract the set of events mostlikely to contain the parton- parton scattering events. Some such di-jet events include more than two jets, but theyare ignored in this section to focus on the di-jet itself.

For the leading two jets comprising a dijet the following four figures show the fractional survival fractions as afunction of corrected jet pT to pass cuts of Fig. 13 n90hits > 1, Fig. 14 EMF > 0.01, and Fig. 15 fHPD < 0.98,separately.

The same survival fraction measure for all cuts simultaneously is given in Fig. 16.

In all four cases:

Left: For the numerator we count the number of jets for which both jets in the dijet event pass the cut, and for thethe denominator we count the number of jets in all dijets before cuts.

Right: For the numerator we count the number of jets which pass the cut, and for the denominator we count thenumber of jets in all dijets before cuts.

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MC

CMS Preliminary

n90hits>1

=900 GeVs|<3!|Jet

n90hist Cut Efficieny

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MC

CMS Preliminary

n90hits>1

=900 GeVs|<3!|Jet

n90hist Cut Efficieny

Figure 13: Dijet N90Hits> 1 cut survival fraction as a function of corrected jet pT . Left: The cut is required forboth leading jets. Right: The cut is required jet-by-jet for the two leading jets..

Comparing the survival fractions for jets comprising dijets to those of inclusive jets we see that in the dijet case,survival fractions are much improved across the whole pT range for which statistics are available.

16

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MC

CMS Preliminary

|>2.6!EMF>0.01 or |Jet

=900 GeVs|<3!|Jet

Minimal EMF Cut Efficieny

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MCCMS Preliminary

|>2.6!EMF>0.01 or |Jet

=900 GeVs|<3!|Jet


Figure 14: Dijet Minimal EMF cut survival fraction as a function of corrected jet pT . Left: The cut is required forboth leading jets. Right: The cut is required jet-by-jet for the two leading jets.

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MC

CMS Preliminary

fHPD<0.98

=900 GeVs|<3!|Jet

fRBX<0.98 & fHPD<0.98 Cut Efficieny

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MCCMS Preliminary

fHPD<0.98

=900 GeVs|<3!|Jet


Figure 15: Dijet fHPD cut survival fraction as a function of corrected jet pT . Left: The cut is required for bothleading jets. Right: The cut is required jet-by-jet for the two leading jets.

17

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MC

CMS Preliminary

|>2.6!EMF>0.01 or |Jet

=900 GeVs|<3!|Jet


(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MCCMS Preliminary

|>2.6!EMF>0.01 or |Jet

=900 GeVs|<3!|Jet


Figure 14: Dijet Minimal EMF cut survival fraction as a function of corrected jet pT . Left: The cut is required forboth leading jets. Right: The cut is required jet-by-jet for the two leading jets.

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MC

CMS Preliminary

fHPD<0.98

=900 GeVs|<3!|Jet


(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MCCMS Preliminary

fHPD<0.98

=900 GeVs|<3!|Jet


Figure 15: Dijet fHPD cut survival fraction as a function of corrected jet pT . Left: The cut is required for bothleading jets. Right: The cut is required jet-by-jet for the two leading jets.

17

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MC

CMS Preliminary

Loose Cut

=900 GeVs|<3!|Jet

All Loose Cuts Efficieny

(GeV)tJet P10 20 30 40 50 60

Frac

tion

of J

ets

Surv

ivin

g C

ut

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

Data

MC

CMS Preliminary

Loose Cut

=900 GeVs|<3!|Jet

All Loose Cuts Efficieny

Figure 16: The loose cut survival fraction as a function of corrected jet pT . Left: The cut is required for bothleading jets. Right: The cut is required jet-by-jet for the two leading jets.

(GeV)tJet P10 20 30 40 50 60

Jets

10

210

310

Without Cut

The Loose Cut

=900 GeVs|<3!|Jet

Jet Pt without Cut

(GeV)tJet P10 20 30 40 50 60

Jets

10

210

310

Without Cut

The Loose Cut

=900 GeVs|<3!|Jet

Jet Pt without Cut

Figure 17: The pT spectrum among jets comprising dijets where the loose cut is required Left: on both jets togetherand Right: on each jet independently.

18

search for dijet resonance in pp collisions at cmssertac/sertac_work.pdf · sertac ozturk search...

Documents