Search for Higgs bosons decaying to aa in the final state ... · USLUA, November 2015 B. Kaplan...

Search for Higgs bosons decaying to aa in theµµττ final state in pp collisions at

√s = 8 TeV

with the ATLAS experimentPhys. Rev. D 92, 05002

B. Kaplan (NYU)

November , 2015

USLUA, November 2015 B. Kaplan (NYU) [email protected] 1

Physics Motivation

Exotic Decay of SM-like Higgs, h

Coupling measurements allow for

non-SM decays up to BR ∼ O(40%),

at 95% CL

Consider a new pseudo-scalar Higgs boson, a

• Focus on, i.e. optimize selection for, 2 ·mτ < ma < 2 ·mb

Leptons from a decay will be highly collimated

We extend analysis above 2 ·mb in case couplings to quarks are suppressed

• Coupling related to mass of decay products ⇒ large BR(a→ ττ) at low mass

• a→ µµ gives a narrow resonance ⇒ clean experimental signature

• One previous search for this signature from D∅, which had no sensitivity

The Next-to-Minimal-Supersymmetric Standard Model

• The NMSSM is a well-motivated (and quite popular) SUSY modelD. Curtin et. al., Phys. Rev. D 90, no. 7, 075004 (2014)

• Naturally achieves mh = 125 GeV

• Notable in that it predicts an additional light pseudoscalar Higgs bosons (a)• Also predicts another scalar Higgs, H, which can decay H → aa


Analysis Overview

1. Pre-select µ±µ∓ events

2. Signal Selection: µ±µ∓+`±trk∓

3. Validation: µ±µ∓+`±trk±

4. Control: µ±µ∓+(b-)jets

Strategy: Use a→ µµ resonance to perform a bump-hunt

• We use single and dimuon triggers

• Require 2 isolated∗ muons, with pT >16 GeV∗For lower a masses, the muons will be collimated. We correct the isolation variable toaccount for this

• They must be OS and pT (µµ) >40 GeV


Analysis Overview

1. Pre-select µ±µ∓ events2. Signal Selection: µ±µ∓+`±trk∓



Detector Signature:

Optimize SR for selecting a→ ττ• Best sensitivity found for at least one τ → `

• Other τ decay inclusive⇒ Require 1, 2, or 3 nearby tracks

• Lead track carries charge of τ⇒ ` and leading track are OS

• Two SRs based on ` flavor


Analysis Overview




Detector Signature:

Test methods in validation regions• Invert OS requirement on ` and leading track

• Comparable backgrounds, but no signal

• Two VRs based on ` flavor


Analysis Overview




Detector Signature:

Measure backgrounds in control regions• Replace ττ selection with jets

• Two CRs for light- and heavy- flavorbackgrounds


Fitting Strategy

Perform a bump hunt in mµµ from 3.7 to 50 GeV in SRs

• Background and signal models measured in the data!

• Use simulation to study shapes

• Need a robust background model for entire mass range

1. Non-resonant background (mainly DY)

2. tt̄ background

3. SM (J/ψ, Υ, Z )

• The J/ψ is used to constrain the ψ′

• The low-end Z tail can be significant above 40 GeV

• All resonances assumed to have a narrow width

• Use a common shape for all narrow µµ resonances


The Narrow µµ Resonances

[GeV]µµm3 3.5 4 4.5 5 5.5

% R

esid

ua

l

100

0

100~~

Event / 50 M

eV

1

10

210

Eve

nt

/ 5

0 M

eV

50

100

150

200

250

300

350

Signal MC

Best Fit

Syst.σ 1±Fit

= 125 GeVh = mH

= 5 GeV, ma

, mµSR

ATLAS Simulation

= 8 TeVs

X → µµ Double-Sided Crystal Ball• Gaussian core, low- and high-end power laws• 3 free parameters, µCB , σCB , αCB

measured in data• Mean (µCB), width (σCB) linearly depend on mX

• Tested on 24 signal samples∗, varying ma and mH

• Fit to SM resonances in data shown below

2.8 3 3.2 3.4 3.6 3.8 4

Events

/ 1

5 M

eV

0.5

1

1.5

2

2.5

3

310×

Data

Background Model

Fit Uncertainty

ATLAS

1 = 8 TeV, 20.3 fbs

CRj

[GeV]µµm2.8 3 3.2 3.4 3.6 3.8 4

% R

esid

ual

20

0

20

9 9.5 10 10.5 11 11.5

Events

/ 3

0 M

eV

0.2

0.3

0.4

0.5

0.6

0.7

310×

Data

Background Model

Fit Uncertainty

ATLAS

1 = 8 TeV, 20.3 fbs

CRj

[GeV]µµm9 9.5 10 10.5 11 11.5

% R

esid

ual

20

0

20

* Details on signal shape (low mass tail and systematics) can be found in the backup


Background MeasurementCRj: ττ(`+ trk)⇒ ≥ 1 jet

Eve

nts

/ G

eV

210

310

410

510

610

710

ATLAS1

= 8 TeV, 20.3 fbsCRjBackground Model

Fit Uncertainty

* ComponentγZ/

Componenttt

Data

* MCγZ/

MCtt

Other

2.8 3 3.2 3.4 3.6 3.8 4

Events

/ 1

5 M

eV

0.5

1

1.5

2

2.5

3

310×

9 9.5 10 10.5 11 11.5E

vents

/ 3

0 M

eV

0.2

0.3

0.4

0.5

0.6

0.7

310×

[GeV]µµm10 20 30 40 50 60 70

% R

esid

ua

l

10

0

10

CRb: ττ(`+ trk)⇒ ≥ 2 b-jets

Eve

nts

/ G

eV

20

40

60

80

100

120

140

ATLAS1

= 8 TeV, 20.3 fbsCRb

Data

Background Model

Fit Uncertainty

* ComponentγZ/

Componenttt

* MCγZ/

MCtt

Other

[GeV]µµm20 30 40 50 60 70

% R

esid

ua

l

50

0

50

• Results used to constrain fit to SRsSome details...

• Resonant backgrounds (J/Ψ, Υ, Z) use a double-sided Crystal Ball

Three CB parameters shared with a→ µµ resonance

• Drell–Yan background modeled by Pγ∗ = eαγ∗ ·mµµ (mµµ)nγ∗

• tt̄ modeled by Rayleigh distribution: Ptt = mµµ × Gaus(mµµ|µtt = 0, σtt)USLUA, November 2015 B. Kaplan (NYU) [email protected] 6

Results: Fit to Signal Region

Eve

nts

/ 200

MeV

[GeV]µµm4 5 6 7 8 910 20 30 40 50 60%

Res

idua

l

-1000

100

5

10

15

ATLAS-1 = 8 TeV, 20.3 fbs

µSR Data

Background Model

Fit Uncertainty

* ComponentγZ/

Componenttt

*γZ/

tt

Other

aa) = 10 %BR(h

Eve

nts

/ 200

MeV

[GeV]µµm4 5 6 7 8 910 20 30 40 50 60%

Res

idua

l-100

0100

5

10

15


SRe Data

Background Model

Fit Uncertainty

* ComponentγZ/

Componenttt

*γZ/

tt

Other

aa) = 10 %BR(h

Constrained from Fit to CR

• Background shape parameters (2 for γ∗, 1 for tt̄, multiple for SM resonances)

• Relative contributions of higher Ψ and Υ spin states

• Relative contribution of Z to total Z/γ∗

Unconstrained:Relative contributions of Ψ, Υ, tt̄ and Z/γ∗

Simulated signal shown in brown for ma = 5, 10 and 20 GeVUSLUA, November 2015 B. Kaplan (NYU) [email protected] 7

Results: Statistical Analysis

[GeV]µµm4 5 6 7 8 9 10 20 30 40 50

pv

alu

e

210

110

1

σ1

σ2

ATLAS1 = 8 TeV, 20.3 fbs

• Observation consistent with SM

• Min p-value = 0.022 for mµµ = 8.65 GeV

• Global p-value > 0.5

• Upper limit on BR(h→ aa) as low as 3.5%for ma = 3.75 GeV, mh = 125 GeV

• Upper limit on BR(H → aa)× σ(gg → H)from 2.33 to 0.72 pb, for ma = 5 GeV

• Compare to D∅ result: BR(h→ aa) > 100%

Scan vs. ma for 125 GeV Higgs:

[GeV]am4 5 6 7 8 9 10 20 30 40

2 )ττ→

BR

(a×

aa)

→

BR

(h×

SM

σ h

)→

(gg

σ

-210

-110

1

10

210 = 125 GeVh = mHmObserved 95% CLMedian Expected 95% CL

σ 1 ± σ 2 ±


Scan vs. mH (new heavy scalar) for 5 GeV a boson:

[GeV]Hm100 150 200 250 300 350 400 450 500

[p

b]

2 )τ

τ→

BR

(a×

aa

) →

BR

(H×

H)

→(g

gσ

110

1

10

210 = 5 GeVam

Observed 95% CL

Median Expected 95% CL

σ 1 ±

σ 2 ±

H→SM gg

ATLAS1 = 8 TeV, 20.3 fbs


Concluding Thoughts

• These are the first limits on this well motivated channel fromthe LHC

• They are the first ever to have sensitivity to BSM physics

• The NMSSM is getting a lot of attention during Run 2, in thischannel and others, so...STAY TUNED!


Backup


‘a’ Mass

ma < 2mτ : a→ µµ(ee)

• Covered by ATLASlepton-jets analysis

2mτ < ma < 2mb: a→ ττ

• 4τ final states are hard!

• We require 1 a→ µµ andtake a 1% hit in BR

• Prime focus of this analysis

2mb < ma: a→ bb

• 4b final states are hard!

• 2µ2τ down by O(0.01)Interesting in case decay to b’s(quarks) are suppressed.

D. Curtin et al., arXiv:1312.4992 [hep-ph]


Signal Model

[GeV]µµm3 3.5 4 4.5 5 5.5

% R

esid

ual

100

0

100~~

Eve

nt

/ 5

0 M

eV

1

10

210

Event / 50 M

eV

50

100

150

200

250

300

350

Signal MC

Best Fit

Syst.σ 1±Fit

= 125 GeVh = mH

= 5 GeV, ma

, mµSR

ATLAS Simulation

= 8 TeVs

a→ µµ Double-Sided Crystal Ball• Gaussian core, low- and high-end power laws• 3 free parameters, µCB , σCB , αCB

measured in data• Width (σCB) enhanced for large mH

determined in simulation

a→ ττ → µµ Gaussian Tail• Mean and width set proportional to CB

determined in simulation• Width was no mH dependence• Fraction of ττ , fττ

determined in simulation

Systematics∗

Shape: Parameters measured in data.Extra uncert. on αCB and fττ to cover tails

Norm.: dom. by theory (11%)∗details in backup


Systematics

Background Model: Use spurious signal method:Perform S+B fit to high stat. bkg-only sample

Signal Model: Parameters measured in data.Extra systematics on αCB and fττ parameters,to ensure coverage of tails

Signal Normalization: Dominant sources:

1. Theory systematic (only applicable to limit vs. ma): 11%

2. Track momentum: vary track pT by 2% ⇒ 5% change in signal acceptance

Sub-Dominant Sources: trigger efficiency, the lepton reconstructionefficiency, the lepton energy scale and resolution, and thecharge of the track


Fits in the Validation RegionBased on SR’s, but require 3rd lepton and track to be SS

Eve

nts

/ 2

00

Me

V

[GeV]µµm4 5 6 7 8 910 20 30 40 50 60

% R

esid

ua

l

100

0

100

5

10

ATLAS1

= 8 TeV, 20.3 fbs

µVR Data

Background Model

Fit Uncertainty

* ComponentγZ/

Componenttt

*γZ/

tt

Other

Eve

nts

/ 2

00

Me

V

[GeV]µµm4 5 6 7 8 910 20 30 40 50 60

% R

esid

ua

l

100

0

100

5

10

ATLAS1

= 8 TeV, 20.3 fbs

VRe Data

Background Model

Fit Uncertainty

* ComponentγZ/

Componenttt

*γZ/

tt

Other

Constrainted from Fit to CR

• Background shape parameters (2 for γ∗, 1 for tt̄, multiple for SM resonances)• Relative contributions of higher Ψ and Υ spin states• Relative contribution of Z to total Z/γ∗

Unconstrained:Relative contributions of Ψ, Υ, tt̄ and Z/γ∗USLUA, November 2015 B. Kaplan (NYU) [email protected] 14

Search for Higgs bosons decaying to aa in the final state ... · USLUA, November 2015 B. Kaplan...

Documents

Transcript of Search for Higgs bosons decaying to aa in the final state ... · USLUA, November 2015 B. Kaplan...