Combined ATLAS Standard Model Higgs Search

8/6/2019 Combined ATLAS Standard Model Higgs Search

1/21

Kyle Cranmer (NYU) EPS-HEP 2011 - ATLAS Higgs Combination

Center for

Cosmology and

Particle Physics

Kyle Cranmer, New York Universityon behalf of the ATLAS Collaboration

Combined ATLAS Standard Model Higgs Search

with 1 fb-1 of Data at 7 TeV

1


2/21

Kyle Cranmer (NYU)

Center for

Cosmology and

Particle Physics

EPS-HEP 2011 - ATLAS Higgs Combination

vacuum expectation

fluctuations

Goldstone modes(become longitudinalpolarizations of massive W,Z)

V() = 2||2 + ||4

Introduction

Understanding of electroweak symmetry breaking is amajor goal of the LHC physics program

Initial focus: search for the Standard Model Higgs

drove the design of both the ATLAS and CMS detectors

stresses every major sub-system

ATLAS and LHC are running great!

Analyses here are based on up to 1.2 fb-1

High pile-up environment

2


3/21

Kyle Cranmer (NYU)

Center for

Cosmology and

Particle Physics


[GeV]HM100 200 300 400 500 1000

H+X)[pb]

!

(pp

"

-210

-110

1

10= 7 TeVs

LHCHIGGSXSWG2010

H(NNLO+NNLLQCD+NLOEW)

!pp

qqH(NNLOQCD+NLOEW)

!pp

WH(NNLOQCD+NLOEW

)

!pp

ZH(NNLOQCD+NLOEW)

!pp

ttH(NLOQCD)

!pp

.

]

]

-1

S

S

010

)

)

(

(

)

pp

tt(

)

pp

.

[GeV]HM

100 120 140 160 180 200

Branchingratios

-310

-2

10

-110

1

bb

!!

cc

gg

"" "Z

WW

ZZLHCHIGGSXSWG2010

- .

[

- .

Channels Included in the Combination

Enhance sensitivity by combining allavailable searches channels in the context

of the SM Higgs hypothesis

some channels are composed of sub-channels and include control samples

3

Cha nelbtag

(veto)Jets

MET

(GeV)Shape

Mass Range

(GeV/c2)Main backgrounds

Mgg 110-150 (from sidebands)WH 2 Mbb 110-130 Top (3j - high Mbb) and W+jets (low Mbb)

Z 2 Mbb 110-130 Z+jets (low Mbb)

WW 0-jet 0 >30 110-240 WW (control region Mll)

(ll) 1-jet veto 1 >30 110-240 Top (from reverse btag) and WW (Mll CR)

WW 0-jet 0 >30 MWW 200-600 W+jets (sidebands)

(lqq) 1-jet veto 1 >30 MWW 200-600 W+jets (sidebands)

ZZ (ll) >30 MT 200-600 VV(from MC) and top (MC and checks)

ZZ ( lqq) 2


4/21Kyle Cranmer (NYU)

Center for

Cosmology and

Particle Physics


Detailed presentations on the channels

4


5/21


6/21


7/21



Center for

Cosmology and

Particle Physics


[GeV]4lm

100 200 300 400 500 600

Events/5GeV

0

1

2

3

4

5

6

7

8

100 200 300 400 500 6000

1

2

3

4

5

6

7

8-1

Ldt = 1.1 fb= 7 TeV,sData 2011,ATLASPreliminary Data

=210 GeV, 1xSMHm

Total background

4l(*)

ZZH

The H ZZ 4l Channel (110 < MH< 600)

8

[GeV]Hm150 200 250 300 350 400 450 500 550 600

SM

/

95%

C

.L.

limiton

1

10

210

sObserved CL

sExpected CL

1

2

PreliminaryATLAS

4l(*)

ZZH-1

Ldt = 1.1 fb=7 TeVs

sObserved CL

sExpected CL

1

2

sObserved CL

sExpected CL

1

2



Center for

Cosmology and

Particle Physics


[GeV]Hm

100 200 300 400 500 600

SM

/

95%C

Lli

miton

1

10

HllWWH

bbW,ZH

HllWWH

bbW,ZH

llllZZHllqqZZHllZZH

qqlWWH

llllZZHllqqZZHllZZH

qqlWWH

Exp. Obs. Exp. Obs.

=7 TeVs,-1

L dt ~ 1.0-1.2 fb CLs limitsATLAS Preliminary

Summary of individual channels limits

The situation prior to the combination

Two individual channels are able to exclude by themselves

9

f

,

[GeV]Hm

100 200 300 400 500 600

SM

!/!

95%C

LLimiton

1

10

210

Tevatron exclusion

LEP exclusion

Observed, PCLExpected, PCL

, PCL!1

, PCL!+ 2S

Observed, CL

SExpected, CL

=7 TeVs-1

L dt ~ 35 pb"

ATLAS

, ,, ,

,

[GeV]Hm

100 200 300 400 500 600

SM

!/!

95%

CLlimiton

1

10

210

""#H

$l$l#WW#H

qq$l#WW#H

""#H

$l$l#WW#H

qq$l#WW#H

llll#ZZ#Hllqq#ZZ#H

$$ll#ZZ#H

llll#ZZ#Hllqq#ZZ#H

$$ll#ZZ#H

Expected Observed Expected Observed

=7 TeVs,-1

L dt ~ 35 pb%

PCL limits

ATLAS

, ,

,

,

. . .

ATLAS w/ 35 pb-1

[arXiv:1106.2748]



Center for

Cosmology and

Particle Physics


Systematics

The channels utilize three main strategies for addressing systematic uncertainties:

data-driven techniques: eliminates dominant impact of uncertainty leaving

uncertainties associated with extrapolation from control to signal region

implicit parametrization: parametrized functions flexible enough to describeeffect of uncertainty on distribution (eg. exponential + Crystal ball in H)

explicit parametrization: variational histograms obtained from modifying

simulated samples according to variations the source of uncertainty (eg. HZZ)

Uncertainties in normalization described by log-normal distributions

The combination requires relations between rates of different channels, thus it is

subject to theoretical uncertainties. Prescription agreed upon by LHC-HCG & Higgscross-section working group.

10

[GeV]HM

100 200 300 400 500 1000

H+X)[pb]

!

(pp

"

-210

-110

1

10= 7 TeVs

LHCHIGGSXSWG2010

H(NNLO+NNLLQCD+NLOEW)

!pp

qqH(NNLOQCD+NLOEW)

!pp

WH(NNLOQCD+NLOEW

)

!

ppZH(NNLOQCD+NLOEW

)

!

pp

ttH(NLOQCD)

!pp

H H bbHWW() HZZ()

qq qq

Luminosity 3.7 3.7 3.7 3.7 3.7 3.7 3.7

e/ efficiency +11.610.4 2.3 1.4

+0.90.8 1.9 1.2 1.1

e/ energy scale -+1.51.6

+0.10.4 - -

+0.81.1

e/ resolution - +2.11.5

+0.00.5 - - -

efficiency - +1.12.0

+0.60.6 0.3 1.2

+0.80.7 0.6

resolution - 5.8 +4.24.5 - - -

Jet/MET energy scale - +2117

+4.67.9

+1518 -

+5.94.0

+3.710.4

Jet resolution - 2.5 - +8.29.0 - -

+2.10.0

MET - +5.56.1 - - -

+6.64.2 -

b-tag efficiency - +3733 - - -

+4.34.4 -

Theory +15.020

.

05 +15.0

20.

0

+15.0

20.

0

+15.0

20.

0

+15.0

20.

0

+15.0

20.

0



Center for

Cosmology and

Particle Physics


Statistical Procedure

The full complexity of individual channels likelihood functions are packaged usingRooFit/RooStats workspaces, and a combined probability model is formed by

identifying nuisance parameters associated to common systematic effects

the common parameter of interest is a cross-section scale factor: = /SM

The profile likelihood ratio is used as a test statistic:

nuisance parameters are profiled based on the data

one-sided variants of the test statistic are used forupper-limits and discovery

The distribution of the test statistic is obtained in two ways:

Ensemble tests with toy Monte Carlo using a fully frequentist procedure

randomize auxiliary measurements instead of randomizing nuisance parameters

Using asymptotic distribution of likelihood ratio

used for primary result

Primary result based CLs, conservatism introduced to protect against downward fluctuations

results based on power-constrained CLs+b (PCL) in backup

Additional comparisons with Bayesian procedure with a uniform prior on

11

Cowan, Cranmer, Gross, Vitells

Eur.Phys.J. C71 (2011)

() = Ls+b(, )/Ls+b(, )

C
http://inspirebeta.net/author/Cowan%2C%20Glen?recid=860907&ln=enhttp://inspirebeta.net/author/Cranmer%2C%20Kyle?recid=860907&ln=enhttp://inspirebeta.net/author/Gross%2C%20Eilam?recid=860907&ln=enhttp://inspirebeta.net/author/Vitells%2C%20Ofer?recid=860907&ln=enhttp://inspirebeta.net/author/Vitells%2C%20Ofer?recid=860907&ln=enhttp://inspirebeta.net/author/Vitells%2C%20Ofer?recid=860907&ln=enhttp://inspirebeta.net/author/Gross%2C%20Eilam?recid=860907&ln=enhttp://inspirebeta.net/author/Gross%2C%20Eilam?recid=860907&ln=enhttp://inspirebeta.net/author/Cranmer%2C%20Kyle?recid=860907&ln=enhttp://inspirebeta.net/author/Cranmer%2C%20Kyle?recid=860907&ln=enhttp://inspirebeta.net/author/Cowan%2C%20Glen?recid=860907&ln=enhttp://inspirebeta.net/author/Cowan%2C%20Glen?recid=860907&ln=en



Center for

Cosmology and

Particle Physics


[GeV]Hm

110 120 130 140 150 160 170 180 190 200

SM

/

95%C

LLimito

n

1

10

Observed CLs

Expected

1

2

ATLAS Preliminary

-1Ldt = 1.0-1.2 fb= 7 TeVs

Limits in the low mass range

12

155



Center for

Cosmology and

Particle Physics


[GeV]Hm

110 120 130 140 150 160 170 180 190 200

SM

/

95%C

LLimito

n

1

10

Observed CLs

Expected

1

2

ATLAS Preliminary

-1Ldt = 1.0-1.2 fb= 7 TeVs

Limits in the low mass range

13

130 140 150 160 170 180 190 200m

H(GeV/c

2)

Tevatron Run II Preliminary, L ! 8.2 fb-1

ExpectedObserved1" Expected2" Expected

TevatronExclusion

SM=1

March 7, 2011

155


14/21

Kyle Cranmer (NYU)

Center for

Cosmology and

Particle Physics


[GeV]Hm120 140 160 180 200

0

p

-610

-510

-410

-310

-210

-110

1

ObservedExpected23

-1Ldt = 1.0-1.2 fb

= 7 TeVs

ATLAS Preliminary

Background-only p-values at low-mass

Broad WW excess is modulated by local fluctuations in and 4l

local significance, no look-elsewhere effect correction applied

14

Largest excess hasapproximately 2.8significance

Test statistic definedsuch that downwardfluctuation gives p0=50%

Center for

B k d l l t l


15/21

Kyle Cranmer (NYU)

Center for

Cosmology and

Particle Physics


[GeV]Hm120 140 160 180 200

0

p

-610

-510

-410

-310

-210

-110

1

ObservedExpected23

-1Ldt = 1.0-1.2 fb

= 7 TeVs

ATLAS Preliminary

Background-only p-values at low-mass

Broad WW excess is modulated by local fluctuations in and 4l

local significance, no look-elsewhere effect correction applied

14

4l candidate Broad WW llexcessexcess

deficit

Largest excess hasapproximately 2.8significance

Test statistic definedsuch that downwardfluctuation gives p0=50%

Center for

Li it f ll


16/21

Kyle Cranmer (NYU)

Cosmology and

Particle Physics


[GeV]H

m200 300 400 500 60

SM

/

9

5%C

LLimiton

1

10

Observed CLs

Expected

1

2

ATLAS Preliminary

-1Ldt = 1.0-1.2 fb= 7 TeVs

155


17/21

Kyle Cranmer (NYU)

Cosmology and

Particle Physics


[GeV]Hm100 200 300 400 500 600

0

p

-510

-410

-310

-2

10

-110

1

Obs. Asymp.Expected2

3Obs. Ensemble

-1Ldt = 1.0-1.2 fb

= 7 TeVs

ATLAS Preliminary

Approximately 8% chance of background-only fluctuation this largeanywhere in range

p-values full mass range

16

HZZ 4lcandidates

No combined excessbeyond 3 is observed

Asymptoticapproximation ingood agreementwith ensemble tests

Center for

Cross checks


18/21

Kyle Cranmer (NYU)

Cosmology and

Particle Physics


[GeV]H

m120 140 160 180 200

S

M

/

95%C

LLimiton

0

2

4

6

8

10Observed CLs

Expected CLs

Observed Bayesian

Expected Bayesian

Observed Ensemble Tests

Expected Ensemble Tests

ATLAS Preliminary

-1Ldt = 1.0-1.2 fb

= 7 TeVs

[GeV]H

m100 200 300 400 500 600

S

M

/

95%C

LLimiton

0

2

4

6

8

10Observed CLs

Expected CLs

Observed Bayesian

Expected Bayesian

Observed Ensemble Tests

Expected Ensemble Tests

ATLAS Preliminary

-1Ldt = 1.0-1.2 fb

= 7 TeVs

Cross-checks

The full limit procedure was performed with toys to confirm theasymptotic distributions of the profile likelihood ratio

Toy Monte Carlo is significantly more computationally intensive andsensitive to fit failures etc. Asymptotic results are robust

In addition, a Bayesian procedure, which is known to reproduce the CLslimit in simple problems, also yielded consistent results.

17

Center for

Cos o ogConclusions


19/21

Kyle Cranmer (NYU)

Cosmology and

Particle Physics


Conclusions

Thanks to the excellent LHC operations, ATLAS has collected more than

1 fb-1 of 7 TeV data leading to substantial gains in sensitivity to theStandard Model Higgs

In the low-mass range (120 140 GeV) an excess of events with asignificance of approximately 2.8 is observed.

ATLAS has extended the 95% CL excluded region around 2MW to

155


20/21

Kyle Cranmer (NYU)

Cosmology and

Particle Physics


Backup

19

Center for

Cosmology andResults with PCL


21/21

Kyle Cranmer (NYU)

Cosmology and

Particle Physics


Results with PCL

20

While CLs is well a established technique in our field, it is considered a non-

standard procedure by statistician mixing notions of power and coverage

it intentionally over-covers to protect against setting limits beyond the

experiments sensitivity due to downward fluctuations

An alternative approach (PCL) is based on purely frequentist CLs+b together

with a power-constraint at the experiments sensitivity achieves the same

protection without mixing the notions of coverage and power

[GeV]Hm200 300 400 500 600

SM

/

95%C

LLimiton

-110

1

10

Observed PCLObservedExpected

1+ 2

ATLASPreliminary

-1Ldt = 1.0-1.2 fb

= 7 TeVs

Alternate Test

[GeV]Hm110 120 130 140 150 160 170 180 190 200

SM

/

9

5%C

LLimiton

-110

1

10

Observed PCLObservedExpected

1+ 2

ATLASPreliminary

-1Ldt = 1.0-1.2 fb

= 7 TeVs

Alternate Test

Documents

Combined ATLAS Standard Model Higgs Search