Search for diboson resonances at CMS - DESYphysikseminar.desy.de/sites2009/site_physikseminar/content/e212/e... · Search for diboson resonances at CMS identifying highly energetic

Search for diboson resonances at CMS

identifying highly energetic boson decays and discriminating newphysics signals from the standardmodel background

Clemens Lange (CERN) DESY Physics Seminar

Hamburg/Zeuthen 31st May/1st June 2016

Clemens Lange - Diboson resonances searches at CMS31.05.2016

About diboson resonances

2

heavy resonances?

boson jets?

background estimation?

wasn’t there this diphoton

resonance?( )

[https://ideas.lego.com/projects/94885]

https://ideas.lego.com/projects/94885


>how can we explain the big difference between the weak force and gravity?

>no symmetry in the standard model (SM) protects the Higgs mass

>µ|H2| always a singlet under phase transformations

> „natural“ explanation would be that SM is replaced by another theory at the TeV scale: µ2 ~ (heavier scale)2 ➜ new particles

> these theories could be:

>SUSY: protecting the Higgs mass by a symmetry

>Composite Higgs: the Higgs is not elementary

>Large/warped extra dimensions: gravity is strong at electroweak scale

We have a problem

3

µ2 = �v2 =�

g24M2

W ⇠ 104 GeV2 ⌧ MPl ⇠ 1038 GeV2

H H

t

t

31.05.2016 Clemens Lange - Diboson resonances searches at CMS

Composite Higgs?

> the Higgs could be non-fundamental

> instead: bound state of a new strong interaction

>e.g. size of 10-18 m ~ Fermi scale (100 GeV) ! light Higgs like a pion from a new sector

>solves hierarchy problem, and brings along new heavy particles/states

>heavy partners of SM particles decay to lighter ones (W, Z, H, top, …)

4

[https://www.learner.org/]

H H

t

t

https://www.learner.org/courses/physics/unit/text.html?unit=2&secNum=11


Why extra dimensions? RS1 Model

Original Randall-Sundrum (RS1) model o↵ers a solution to thehierarchy problem by postulating a 5

th space-time bounded by two(3 + 1)-dimensional branes.

Gravity is localized aty = 0, called the UV-

or Planck-brane.

Only gravity canpropagate through

‘bulk’.

SM particles reststrictedto y = ⇡R (IR- or TeV-brane).

Physical masses rescaledby e

�⇡kR: gravity is weak.

The resulting metric is nonfactorizable and depends on the radius yand curvature k

�1 of the extra dimension:

ds

2 = e

�2ky

⌘

µ⌫

dx

µ

dx

⌫ + dy

2; 0 y ⇡R

Therefore the RS warped geometry model proposes a solution to the‘hierarchy problem’ with reasonable values of kR ⇠ 11

Massive excited graviton modes (G⇤) are a defining featureE. Williams (Columbia U.) G⇤ ! WW ! `⌫jj thesis defense July 2nd, 2012 9 / 41

Large extra dimensions?

>another attempt to solve the hierarchy problem

>SM fields are confined to four-dimensional „membrane“, gravity propagates in additional dimensions

>effectively, change power law of gravity from to , where N = number of extra dimensions

> this only applies to particles with - smaller things have more possibilities to move

>„large“, because of size 1 mm to~1/TeV

>proposed by Arkani-Hamed, Dimopoulos, and Dvali (ADD)

5

Eric Williams

1/r2 1/r2+N

r ⌧ N

gravity gravity can propagatethrough bulk

bulk

Planck-brane TeV-brane

http://de.slideshare.net/EricWilliams21/ran-30953217


Warped extra dimensions?

>often referred to as Randall-Sundrum (RS) models

>warping causes energy scale at one end of the extra dimension to be much larger than at the other end

>SM models reside on TeV-brane (in RS1 models)

>bulk graviton models allow SM particles into 5D-bulk

>overlap of 5-D profiles at TeV-brane (and Higgs) determine particle masses

>additionally, if distance between two branes is not fixed, additional fluctuations can occur

6

Eric Williams

Why extra dimensions? Bulk RS Model

Modern RS models (bulk RS) allow SM particles into 5-D bulk

Overlap of 5-D profiles at TeV brane (and the Higgs) determineparticles masses

Suppressed coupling to bosons and lightfermions; negligible rates to �� and ``

Enhanced coupling to heavy particles(t¯t,ZZ and WW ) motivates search in WW

channel!

G*! WW G*! ZZ G*! HH G*! gg G*! tT

E. Williams (Columbia U.) G⇤ ! WW ! `⌫jj thesis defense July 2nd, 2012 10 / 41

bulk

Why extra dimensions? RS1 Model

Original Randall-Sundrum (RS1) model o↵ers a solution to thehierarchy problem by postulating a 5

th space-time bounded by two(3 + 1)-dimensional branes.

Gravity is localized aty = 0, called the UV-

or Planck-brane.

Only gravity canpropagate through

‘bulk’.

SM particles reststrictedto y = ⇡R (IR- or TeV-brane).

Physical masses rescaledby e

�⇡kR: gravity is weak.

The resulting metric is nonfactorizable and depends on the radius yand curvature k

�1 of the extra dimension:

ds

2 = e

�2ky

⌘

µ⌫

dx

µ

dx

⌫ + dy

2; 0 y ⇡R

Therefore the RS warped geometry model proposes a solution to the‘hierarchy problem’ with reasonable values of kR ⇠ 11

Massive excited graviton modes (G⇤) are a defining featureE. Williams (Columbia U.) G⇤ ! WW ! `⌫jj thesis defense July 2nd, 2012 9 / 41

gravity gravity can propagatethrough bulk

bulk

Planck-brane TeV-brane

RS1 model

Bulk model

http://de.slideshare.net/EricWilliams21/ran-30953217


How do we observe these models at the LHC?

>should be able to observe excitations/resonances/fluctuations

>composite Higgs: electroweak composite vector resonances !mostly spin-1 (W’, Z’)

! decay to pairs of W, Z, H

>Randall-Sundrum: Kaluza-Klein excitations of gravitons + radion fluctuations

>gravitons (spin-2): !RS1: decay predominantly to leptons

! bulk: decay to pairs of W, Z

!ADD: broader excess from many narrow-spaced resonances

>radions (spin-0): ! only used for signal modelling here

>focus here on narrow resonances (width < detector resolution), mass ≥ 600 GeV

7

1 Feynman diagrams

W 0W

H

q0

q

b

b

l

⌫

(a)

XW

W

g

g

q

q̄

l

⌫

(b)

g

g

g

q

q̄

c, b

c̄, b̄

(c)

Figure 1.1: Heavy X particle production and decay.

1

1 Feynman diagrams

W 0W

H

q0

q

b

b

l

⌫

(a)

XW

W

g

g

q

q̄

l

⌫

(b)

g

g

g

q

q̄

c, b

c̄, b̄

(c)


1


What about the diphoton resonance?

>neutral resonance could be graviton or radion as in diboson searches

> resonance cannot directly couple to photons ➜ loop of charged particles (e.g. W, top, ?) in decay (and production?)

> there must be more than just a di-photon resonance

>searches presented in this talk constrain what physics models this potential resonance could be

8


Reconstructing heavy resonances

>bosons will be very energetic ➜ collimated decay products

>need to develop dedicated reconstruction methods

>hadronic decays of bosons: ! „boson-tagging“

! exploiting substructure of jets

> leptonic decays: ! special isolation for dileptonic decays

! dedicated reconstruction algorithms for high-pT leptons

! new tau-identification algorithms

9

focus here on Run-2 developments and analyses


Hadronic boson identification

>at CMS use anti-kT jet algorithm with R = 0.4

>already for resonances of 1 TeV a significant fraction of cases where the boson decay is contained in a single jet

> increase jet size to R = 0.8 to contain full decay within „fat“ jet

10

back of the envelope calculation:

for a resonance of mass 1 TeV the bosons from the decay will have pT~0.4 TeV ➜ ΔR ≈ 0.4

�Rqq ⇡ 2MV

pVT

JHEP 12 (2014) 017

5

but instead require that there is at least one AK5 b jet, with an angular distance of DR > 0.8 tothe CA8 jet considered as W jet candidate. To increase the statistical precision of the sample,we select the CA8 jet with the largest mass and with Df between the lepton and the jet greaterthan p/2 as W jet candidate, rather than the highest pT CA8 jet.

5 Algorithms for W jet identification

A jet clustering algorithm with R = 0.8 is used to identify W jets. A large value of R increasesthe efficiency to reconstruct W bosons with small boost as single jets, since the average angulardistance between the W decay products is inversely proportional to the pT of the W. The cho-sen value of R provides a high efficiency for W bosons with small boost and ensures that noefficiency is lost in the transition from classical W reconstruction from two small jets at low WpT and reconstruction from a single large jet at higher W pT (see e.g. Ref. [46]). Another pointto consider when choosing the value of R, is the tt data sample available for validating highlyboosted W jets. If R is chosen too large, the b quark from the t ! Wb decay tends to mergeinto the W jet. The chosen value of R is the result of a compromise between high efficiency forW bosons with small boost and a sufficiently large sample of W jets in tt data for validating theW jet identification algorithms.

(GeV)T

W boson p200 400 600 800 1000 1200

Rec

onst

ruct

ion

effic

ienc

y

0

0.2

0.4

0.6

0.8

1

Merged jet efficiency

single CA R=0.8 jet

R (W,jet) < 0.1∆

Resolved jets efficiency

two AK R=0.5 jets

) < 0.1j

,jeti

R (q∆

8 TeV

CMSSimulation

Figure 1: Efficiency to reconstruct a CA8 jet within DR < 0.1 of a generated W boson, and theefficiency to reconstruct two AK5 jets within DR < 0.1 of the generated quarks from longitudi-nally polarized W bosons, as a function of the pT of the W boson.

Figure 1 shows the pT range of W bosons for which the R = 0.8 algorithm is efficient andcompares this to the efficiency for reconstructing W bosons from two R = 0.5 jets. Above apT of 200 GeV, the CA8 jet algorithm, used to identify W jets, becomes more efficient than thereconstruction of a W boson from two AK5 jets. In this paper we therefore study substructureobservables to identify W jets for an R = 0.8 algorithm. Whether an AK or a CA algorithmis used in such comparison does not affect the overall conclusion. The choice of CA (withR = 0.8) and AK (R = 0.5) is simply due to their wide use in CMS publications, where CAwas introduced in the first top tagging algorithm paper of CMS [47]. Whenever we refer toefficiency (e) in this paper, we refer to the full efficiency to identify a W boson relative to allgenerated W bosons decaying to hadrons.

�R =p

�⌘2 +��2


CMS particle flow reconstruction

> tracking detectors and calorimeters contained in magnetic field

>particle flow algorithm makes use of sub-detectors with best resolution (both spatial and energy)

>actual „particles“ enter jet clustering

11


Jet pruning

>we know the masses of W, Z and Higgs very well ➜ can use them as constraints

>however, large number of particles in jet ➜ rather bad resolution

> jet pruning (generally grooming) removes soft and large angle radiation

>strategy: ! recluster jet using Cambridge-Aachen

(CA) jet algorithm

12

?pruned jet mass

0 50 100 150

arb

itra

ry u

nits

0

0.1

0.2

SM Higgs, m = 600 GeV

ungroomed jet mass

W+Jets, MadGraph+Pythia6

ungroomed jet mass

CMS Simulation

CMS-PAS-HIG-14-008


Reminder: jet clustering algorithms

>kT-algorithms: sequential clustering

>examine four-vector inputs pairwiseand construct jets hierarchically

>anti-kT: preferentially merge constituents with high pT with respect to their nearest neighbours first

>Cambridge-Aachen: no pT-weighting, merge based on spatial separation only ➜ undoing clustering yields subjets

13

JHEP 0804:063,2008


Jet pruning

>we know the masses of W, Z and Higgs very well ➜ can use them as constraints

>however, large number of particles in jet ➜ rather bad resolution

> jet pruning (generally grooming) removes soft and large angle radiation

>strategy: ! recluster jet using Cambridge-Aachen

(CA) jet algorithm

! „soft“:

! „large angle“: , orig = unpruned CA jet

>cut on mass window (~±10 GeV)

14

min(piT , pjT )/p̃T < 1

�Rij > morig/porigTPruned mass [GeV]

0 50 100 150 200 250 300Ev

ents

/ 5.

00 G

eV0

100

200

300

400

500

310×

CMS Data

WW (MADGRAPH))→ (2 TeV) Bulk

W (G

ZZ(MADGRAPH))→ (2 TeV) Bulk

Z (G WZ (MADGRAPH))→W (W' (2 TeV)

QCD PYTHIA8

(13 TeV)-12.6 fb

CMSPreliminary

Pruned mass [GeV]0 50 100 150 200 250 300

MC

Dat

a

0.5

1

1.5

CMS-PAS-EXO-15-002


N-subjettiness

> for boson-tagging: want to quantify how 2-subjetty a jet is

>➜ to what extent is energy flow aligned along 2 momentum directions (N=2)?

15

JHEP 1103:015,2011

(a)

−0.2 0 0.2 0.4 0.6 0.8 1

1

1.2

1.4

1.6

1.8

2

2.2Boosted W Jet, R = 0.6

η

φ

(b)

(c)

−1.2 −1 −0.8 −0.6 −0.4 −0.2

4.6

4.8

5

5.2

5.4

5.6

5.8Boosted QCD Jet, R = 0.6

η

φ

(d)

Figure 1: Left: Schematic of the fully hadronic decay sequences in (a) W+W− and (c) dijet QCDevents. Whereas a W jet is typically composed of two distinct lobes of energy, a QCD jet acquiresinvariant mass through multiple splittings. Right: Typical event displays for (b) W jets and (d)QCD jets with invariant mass near mW . The jets are clustered with the anti-kT jet algorithm [31]using R = 0.6, with the dashed line giving the approximate boundary of the jet. The marker sizefor each calorimeter cell is proportional to the logarithm of the particle energies in the cell. Thecells are colored according to how the exclusive kT algorithm divides the cells into two candidatesubjets. The open square indicates the total jet direction and the open circles indicate the twosubjet directions. The discriminating variable τ2/τ1 measures the relative alignment of the jetenergy along the open circles compared to the open square.

with τN ≈ 0 have all their radiation aligned with the candidate subjet directions and

therefore have N (or fewer) subjets. Jets with τN ≫ 0 have a large fraction of their energy

distributed away from the candidate subjet directions and therefore have at least N + 1

subjets. Plots of τ1 and τ2 comparing W jets and QCD jets are shown in Fig. 2.

Less obvious is how best to use τN for identifying boosted W bosons. While one might

naively expect that an event with small τ2 would be more likely to be a W jet, observe that

QCD jet can also have small τ2, as shown in Fig. 2(b). Similarly, though W jets are likely

– 4 –

(a)

−0.2 0 0.2 0.4 0.6 0.8 1

1

1.2

1.4

1.6

1.8

2


η

φ

(b)

(c)

−1.2 −1 −0.8 −0.6 −0.4 −0.2

4.6

4.8

5

5.2

5.4

5.6


η

φ

(d)









– 4 –

(a)

−0.2 0 0.2 0.4 0.6 0.8 1

1

1.2

1.4

1.6

1.8

2


η

φ

(b)

(c)

−1.2 −1 −0.8 −0.6 −0.4 −0.2

4.6

4.8

5

5.2

5.4

5.6


η

φ

(d)









– 4 –

(a)

−0.2 0 0.2 0.4 0.6 0.8 1

1

1.2

1.4

1.6

1.8

2


η

φ

(b)

(c)

−1.2 −1 −0.8 −0.6 −0.4 −0.2

4.6

4.8

5

5.2

5.4

5.6


η

φ

(d)









– 4 –

W-jet QCD-jet

normalisation sum over particles minimise distance to candidate subjets

⌧N =1

d0

X

k

pT,k min(�R1,k,�R2,k, ...,�RN,k)

1 axis2 axes

low values of τN ➜ compatibility with the hypothesis of N axes


N-subjettiness ratio

>bare τN has very little discrimination power

>take ratio τ2/τ1 instead

>mind: rather complicated variable, difficult to model ➜ need to validate in data

>clean sample of W-jets: top-antitop quark pairs used for calibration

16

21τ0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Even

ts /

0.05

2000

4000

6000

8000

10000

12000

14000CMS Data

WW (MADGRAPH)→ (2 TeV) BulkG

ZZ(MADGRAPH)→ (2 TeV) BulkG WZ (MADGRAPH)→W' (2 TeV)

QCD PYTHIA8

< 105 GeVP65 GeV < M

(13 TeV)-12.6 fb

CMSPreliminary

21τ0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

MC

Dat

a

0.5

1

1.5

CMS-PAS-EXO-15-002

Even

ts /

( 0.0

4 )

0

50

100

150

200

250

300

350 CMS Data tt

Single Top VV

W+jets MC Stat

Wprime1TeV x 50

(13 TeV)-12.2 fb

CMSPreliminary

(13 TeV)-12.2 fb

CMSPreliminary

1τ/2τ0 0.2 0.4 0.6 0.8 1

Dat

a / M

C

0.5

1

1.5

2


Higgs➜bb tagging

>Higgs has higher mass than W/Z bosons ➜ τ2/τ1 less important, exploit b-jet content instead

> two different strategies: ! identify b-subjets

! tag fat jet

>currently, both show comparable performance

>50% lower mis-tagging rate than W-/Z-tagging

>dedicated Higgs-tagger available soon

17

EXO-14-009

(GeV)jJet mass m0 50 100 150 200

Arbi

trary

sca

le

0

0.1

0.2

0.3

0.4q/g MADGRAPH+PYTHIA

q'bq → Wb →t b b→H

4q→ WW* →H q'q →W q q→Z

CA pruned R=0.8

CMSSimulation

8 TeV

CMS DP-2015/038

H➜WW➜qqqq tagging is done using τ4/τ2 ratio

(cf. τ2/τ1 for W/Z)


Higgs➜ττ tagging

>τ-lepton can decay hadronically and leptonically

>need to take into account potential overlap between the two τ-leptons ! remove tracks/particles entering other

isolation cone

>discrimination against q-/g-jets: MVA-based isolation ! sum reconstructed particle energies in

various cones around τ decay products

>neutrinos in decay cannot be reconstructed ➜ missing energy ! ττ-reconstruction using templates from

Monte Carlo simulation (SVfit)

18

CMS-PAS-EXO-15-008

b-tagged untaggedν

!

!

H

μ ν

ν

π

h

CMS Preliminary (8 TeV)-119.7 fb

) [GeV]τ,τM(20 40 60 80 100 120 140 160 180 200 220 240

Even

ts /

10 G

eV

5

10

15

20

25

30

35 Drell-Yan + jetstt

W + jetsSingle topDibosonQCD multijets

=1)RΛ=1.5 TeV, X

30 (M×Signal Observed

currently only 8 TeV results


Lepton (muon + electron) reconstruction

> two isolation issues: ! radiation from highly energetic leptons

spoils isolation

! leptons spoil each other’s isolation

>employ dedicated high-pT algorithms to preserve high efficiency

> loosen selection criteria one of the leptons in Z➜ll decays

> leptonic W-decay: need to recover z-component of neutrino ! use W-mass constraint for reconstruction

19

Effic

ienc

y

0.75

0.8

0.85

0.9

0.95

1

>35 GeVT

Barrel, E

-1 L dt = 19.6 fb∫ = 8 TeV, sCMS Preliminary,

DATA (BG subtracted)

DY (MC)

(GeV)TE40 50 60 70 100 200 300 400 500 1000

Scal

e Fa

ctor

0.90.920.940.960.98

11.021.041.061.08

1.1

High energy electron ID (HEEP)

CMS DP-2013/003


Strategy/event selection for VV analyses

20

VV ➜ qqqq analysis VW ➜ qqlν analysis VZ ➜ qqll analysis

trigger HT trigger (800 GeV)or jet+groomed mass single lepton trigger (e/µ pT > 105/45 GeV)

lepton(s) — HEEP e/high-pT µ special iso/ID for2nd lepton

V-jetanti-kT R=0.8, pT > 200 GeV,exploit substructure τ2/τ1, use groomed mass

V boson candidate(s)

Δη < 1.3 reconstruct leptonic V, pT > 200 GeV use mass window

Xreconstruct X using both reconstructed vector bosons

search for bump in mVV distribution

Models and final states

•  Looking for mass>~600 GeV VV/VH/HH resonances •  Models of extra dimensions

•  KK-Graviton G ! WW/ZZ/HH, Radion R ! WW/ZZ/HH •  Bulk graviton model prefers di-boson decays,

while classical Randall-Sundrum model prefers di-fermion decays •  Models of compositeness, new forces

•  W’ ! WZ/WH, Z’ ! WW/ZH, Technicolor ρTC ! WZ •  Composite Higgs, Little Higgs model prefer di-boson decays,

while sequential standard model prefers di-fermion decays •  Generalized models of heavy vector triplets (HVT) consisting of X± and X0

•  Bosons get high (pT>~200 GeV) boost •  special reconstruction techniques

for jets and taus •  special isolation techniques for leptons

3 4 Nov 2014 Andreas Hinzmann

G

W

W

g

g

q

q

⌫

`

More from Jen, Alexandra

More from Nhan, Dinko, Cesar, Aniello

X

additional cuts on separation of bosons in Δɸ and ΔR


Dijet (VV) analysis

>trigger at ~100% efficiency at mJJ > 1 TeV ! apply cut on reconstructed dijet system

>define different τ2/τ1 regions: ! high purity to suppress background

! low purity to recover signal efficiency at high masses

>split W and Z samples based on pruned jet mass (65-85, 85-105 GeV)

>still dominated by QCD multi-jet events

>difficult to obtain sufficient MC simulation statistics

>need a data-driven approach

>exponentially falling spectrum: use fit function

21

21τ0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Even

ts /

0.05

2000

4000

6000

8000

10000

12000

14000CMS Data

WW (MADGRAPH)→ (2 TeV) BulkG

ZZ(MADGRAPH)→ (2 TeV) BulkG WZ (MADGRAPH)→W' (2 TeV)

QCD PYTHIA8

< 105 GeVP65 GeV < M

(13 TeV)-12.6 fb

CMSPreliminary

21τ0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

MC

Dat

a

0.5

1

1.5

CMS-PAS-EXO-15-002

HP

LP


VV background estimation

>naively, fitting the mJJ spectrum could swallow signal

>also, need to avoid claiming false discovery (in particular in tail)

>fit function:

>number of free parameters determined byF-test: ! check if quality of fit improves by > 10% confidence

level

! if not, stick with current fit function

>extensive bias tests conducted

>combined signal+background fit performed

22

Dijet invariant mass [GeV]1000 1500 2000 2500 3000 3500

Even

ts /

94.7

GeV

-110

1

10

210

310

410CMS data

2 par. background fit

= 0.014 pb)σWW (→G(2 TeV)

WW, low-purity > 200 GeV

T| < 2.4, pη|

| < 1.3jjη∆ > 1 TeV, |jjM

(13 TeV)-12.6 fb

CMSPreliminary

Dijet invariant mass [GeV]1000 1500 2000 2500 3000 3500

σD

ata-

Fit

-3-2-10123

or

2 parameters 3 parameters

CMS-PAS-EXO-15-002

very similar background estimation strategy as for

diphoton resonance search


Intermezzo: Zɣ search overview

> recently published Z➜ll + ɣ search

> inspired by ɣɣ „excess“

>same photon ID as ɣɣ search, dilepton ID as in ZV search, fit background

> limited by statistics, no significant excess

23

CMS-PAS-EXO-16-019

Even

ts /

( 20

GeV

)

1

10

Data

Fit

Uncertainty

) [GeV]γ-e+M(e200 400 600 800 1000 1200 1400

stat

σ(d

ata-

fit)/

2−1−012

(13 TeV)-12.7 fbCMS Preliminary

Even

ts /

( 20

GeV

)

1

10

210Data

Fit

Uncertainty

) [GeV]γ-µ+µM(200 400 600 800 1000 1200 1400

stat

σ(d

ata-

fit)/

2−1−012

(13 TeV)-12.7 fbCMS Preliminary

electron channelmuon channel


VW/VZ analysis overview

>for 2015 data, two separate analyses performed: ! „low mass“: 600-1000 GeV

! „high mass“: 1-4 TeV

!VZ analysis not yet public

>difference low vs. high mass: ! lower boost ➜ can use isolated lepton

triggers with 27 GeV thresholds

>requiring an isolated lepton suppresses QCD multi-jet background significantly

>dominant backgrounds: !Drell-Yan/W+Jets

! top-antitop quark production

>can estimate individual background components from sidebands

24

Even

ts /

( 20

)

0

0.20.4

0.60.8

11.2

1.41.6

1.8

310×CMS Data W+jets

VV tt

Single Top MC Stat

Wprime1TeV x 50

(13 TeV)-12.2 fb

CMSPreliminary

(13 TeV)-12.2 fb

CMSPreliminary

(GeV)AK8pT100 200 300 400 500 600 700

Dat

a / M

C

0.5

1

1.5

2

CMS-PAS-B2G-16-004


VW analysis background estimation

>statistics in MC simulated samples still limited

> furthermore, analysis performed in extreme phase space

>use pruned mass sidebands (40-65 GeV, 135-160 GeV) to exploit correlation between pruned jet mass and resonance mass ! Higgs mass region kept blind

>determine ratio of simulated to data distributions in sideband

>extrapolate to signal region using transfer function (based on simulation)

>method accounts for data-MC differences in shape and normalisation

25

CMS-PAS-B2G-16-004CMS-PAS-EXO-15-002

Pruned jet mass (GeV)40 60 80 100 120 140

data

σD

ata-

Fit

-202

Even

ts /

( 5 G

eV )

50

100

150

200

250

300

350

→ signal region ←

WWenriched

WZenriched

High-Purity

νµCMS Data W+jets

WW/WZ ttSingle Top Uncertainty

(13 TeV)-12.2 fb

CMSPreliminary

(GeV)WVM1 1.5 2 2.5 3 3.5 4

310×

data

σD

ata-

Fit

-202

Even

ts /

( 100

GeV

)-210

-110

1

10

210

310

410 νCMS Data e W+jets

WW/WZ ttSingle Top Uncertainty

100)×=2 TeV (G MBulkGHigh-Purity, WW enriched

(13 TeV)-12.2 fb

CMSPreliminary

Higgs


Intermezzo: ɣɣ vs. Zɣ vs. WW

> limits for narrow resonances ! caveat: slightly different models used

>minimal upward fluctuations?

26

CM

S-PA

S-E

XO

-15-004

CM

S-PA

S-B

2G-16-004

CM

S-PA

S-E

XO

-16-019

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

) [fb

]γZ

→ B

R(A

× σ

95%

CL

UL

on

0

50

100

150

200

250

300

W = 0.014%

Observed

σ 1±Expected

σ 2±Expected

(13 TeV)-1 2.7 fbCMS Preliminary

(GeV)GM600 700 800 900 1000

WW

)(pb)

→ Bu

lk x

BR

(G95

%σ

-310

-210

-110

1

10

210 (13 TeV)-12.3 fb

CMSPreliminary Expected

SAsympt. CL

1 s.d.± Expected S

Asympt. CL 2 s.d.± Expected

SAsympt. CL

, k = 0.5 WW→BulkG BR× THσ

ObservedS

Asympt. CL

ν l→W

ɣɣ

Zɣ

WW


VH analysis overview

>2015 data: H➜bb, leptonic W/Z decays !W➜lν

! Z➜ll

! Z➜νν

>categorise in single and double subjet b-tag categories

>same background estimation method as for VW search

27

CMS-PAS-B2G-16-003

jet pruned mass (GeV)50 100 150 200 250 300

Even

ts /

5.0

GeV

0

5

10

15

20

25

30

35

40

45 (13 TeV)-12.17 fb

CMSPreliminary

)bbνν,ν (ll,l→ Vh →X

0l, 1 b-tag

LSB VV Higgs HSB

DataV+jetsTop, STVV, VHBkg. unc.

jet pruned mass (GeV)50 100 150 200 250 300

Pulls

4−2−024

(GeV)VhTm

1000 1500 2000 2500 3000

Even

ts /

100.

0 G

eV2−10

1−10

1

10

210

(13 TeV)-12.17 fb

CMSPreliminary


0l, 1 b-tag

DataV+jetsTop, STVV, VHBkg. unc.

=2000 GeVXmHVT model B

(GeV)VhTm

1000 1500 2000 2500 3000

Pulls

4−2−024

Z➜νν, 1 b-tag


Signal modelling and uncertainties

>depending on spin of new particles, polarisation of bosons different

>Bulk graviton (spin-2) and W’/Z’ (spin-1) models primarily couple to longitudinal components of W/Z

>analytical description of signal shapes based on fully simulated benchmark mass points ! double-sided Crystal-Ball function

! linearly interpolation between benchmark points

>signal efficiency up to 15% depending on analysis category

> largest uncertainties: ! background estimation

! jet energy and mass scale

! boson-tagging

28

(GeV)Vhm0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

Arbi

trary

uni

ts0

0.1

0.2

0.3

0.4

0.5

0.6

0.7 CMSSimulation Preliminary


channelµ2

13 TeV

=800 GeVXm=1000 GeVXm=1200 GeVXm=1400 GeVXm=1600 GeVXm=1800 GeVXm=2000 GeVXm=2500 GeVXm=3000 GeVXm=3500 GeVXm=4000 GeVXm

CMS-PAS-B2G-16-003

Z➜µµ, ≥1 b-tag


Model interpretation

>several different analyses performed

>advantageous to use common benchmark models for easier interpretation

>need to know individual couplings to bosons ! individual analysis: e.g. σ(gg➜G➜WW)

! combination: e.g. σ(gg➜G)

!mind also production mechanism

29

arXiv:1404.0102

are transverse (see for example [63]). The kk-graviton coupling to bulk weab bosons alsoexperience the dg volume suppression.

Decays

On figure (V.4) we show the kk-graviton branching fractions to SM particles, the leftpart of this figure shows the kk-graviton branching ratios on bulk scenario, assumingthe elementary top scenario. The right part shows kk-graviton branching ratios on RS1scenario.

On this figure we clearly note the couplings of RS1 kk-graviton with SM particlesare democratic with respect to the number of degree of freedom of each field, while thecouplings of bulk kk-graviton with SM particles are predominantly to the fields localizedon TeV brane (massive bosons and higgs field).

The highest branching fraction of a RS1 kk-graviton is to di-jet final states (gg +qq̄).The di-photon channel would also be a good search channel on this scenario, by itscleanness and non negligible branching ratio.

For the bulk kk-graviton case the highest branching fractions are to massive states,heavy di-bosons and top quarks (on the elementary top scenario). On the case of atrans-TeV resonance, each one of the di-bosons products are boosted and consequentlyits sub-products collimated. The development of substructure techniques make from thehadronic di-boson channels golden channels on bulk kk-graviton search.

200 400 600 800 1000 1200 140010- 4

0.001

0.01

0.1

1

MGr H GeVL

GHGe

VL

Bulk

tt

gg

aa

hh

WW

ZZ

200 400 600 800 1000 1200 140010- 4

0.001

0.01

0.1

1

MGr H GeVL

GHGe

VL

RS1

nn

ll

qq

tt

gg

aa

hh

WW

ZZ

Figure V.4: KK-graviton branching fractions. Left: Bulk scenario, assuming elementary topquark. Right RS1 scenario. The dashed line represents the individual RS1 kk-graviton decayrates to a pair of light fermions ff̄ , where f represents both light quarks (u, d, s, c, b) or leptons(e, µ, ⌧). – Branching ratios are independent of k parameter.

Figure (V.5) shows a zoom of kk-graviton branching ratios comparing two distinctscenarios for third family WED embedding: The thick curves considers a fully elementarytop quark [60] while thin dashed ones we ignores kk-graviton coupling with top quark.It is not a surprise to note that kk-graviton branching ratios to bosons pais (W/Z/H)increases when there is not kk-graviton coupling to top quark.

On figure (V.6) we see the total width on RS1 scenario the kk-graviton is around twoorders of magnitude bigger than on bulk scenario, given the same geometric parameters.The experimental resolution on the width of the resonance depends on the investigated

18

are transverse (see for example [63]). The kk-graviton coupling to bulk weab bosons alsoexperience the dg volume suppression.

Decays

On figure (V.4) we show the kk-graviton branching fractions to SM particles, the leftpart of this figure shows the kk-graviton branching ratios on bulk scenario, assumingthe elementary top scenario. The right part shows kk-graviton branching ratios on RS1scenario.

On this figure we clearly note the couplings of RS1 kk-graviton with SM particlesare democratic with respect to the number of degree of freedom of each field, while thecouplings of bulk kk-graviton with SM particles are predominantly to the fields localizedon TeV brane (massive bosons and higgs field).

The highest branching fraction of a RS1 kk-graviton is to di-jet final states (gg +qq̄).The di-photon channel would also be a good search channel on this scenario, by itscleanness and non negligible branching ratio.

For the bulk kk-graviton case the highest branching fractions are to massive states,heavy di-bosons and top quarks (on the elementary top scenario). On the case of atrans-TeV resonance, each one of the di-bosons products are boosted and consequentlyits sub-products collimated. The development of substructure techniques make from thehadronic di-boson channels golden channels on bulk kk-graviton search.

200 400 600 800 1000 1200 140010- 4

0.001

0.01

0.1

1

MGr H GeVL

GHGe

VL

Bulk

tt

gg

aa

hh

WW

ZZ

200 400 600 800 1000 1200 140010- 4

0.001

0.01

0.1

1

MGr H GeVL

GHGe

VL

RS1

nn

ll

qq

tt

gg

aa

hh

WW

ZZ

Figure V.4: KK-graviton branching fractions. Left: Bulk scenario, assuming elementary topquark. Right RS1 scenario. The dashed line represents the individual RS1 kk-graviton decayrates to a pair of light fermions ff̄ , where f represents both light quarks (u, d, s, c, b) or leptons(e, µ, ⌧). – Branching ratios are independent of k parameter.

Figure (V.5) shows a zoom of kk-graviton branching ratios comparing two distinctscenarios for third family WED embedding: The thick curves considers a fully elementarytop quark [60] while thin dashed ones we ignores kk-graviton coupling with top quark.It is not a surprise to note that kk-graviton branching ratios to bosons pais (W/Z/H)increases when there is not kk-graviton coupling to top quark.

On figure (V.6) we see the total width on RS1 scenario the kk-graviton is around twoorders of magnitude bigger than on bulk scenario, given the same geometric parameters.The experimental resolution on the width of the resonance depends on the investigated

18

1 Feynman diagrams

W 0W

H

q0

q

b

b

l

⌫

(a)

XW

W

g

g

q

q̄

l

⌫

(b)

g

g

g

q

q̄

c, b

c̄, b̄

(c)


1


Model tuning

>models described before can be tuned by a handful of parameters

>bulk graviton: !mass of graviton

! coupling constant determining production cross section and width

>heavy vector triplet (HVT): ! phenomenological Lagrangian

! describes production and decay of heavy spin-1 resonances

! 4 parameters for resonance mass, interaction strength, couplings to bosons and fermions

! focus here on „Model B“ with enhanced couplings to bosons

30

500 1000 1500 2000 2500 3000 3500 4000

0.02

0.04

0.06

0.08

0.10

0.12

M0 @GeVD

BRHV0Æ2XL W+W-

Zhuudd

llè

nnbbttè

gV = 1

Model A

500 1000 1500 2000 2500 3000 3500 4000

10-3

10-2

10-1

M0 @GeVD

BRHV0Æ2XL W+W-

Zhuudd

llè

nnbbttè

gV = 3

Model B

500 1000 1500 2000 2500 3000 3500 4000100

101

102

M0 @GeVD

GHV0Æ2XL@G

eVD gV = 1

gV = 2gV = 3

Model A

500 1000 1500 2000 2500 3000 3500 4000

102

103

M0 @GeVD

GHV0Æ2XL@G

eVD

gV = 3gV = 5gV = 8

Model B

Figure 2.1: Upper panel: Branching Ratios for the two body decays of the neutral vector V 0 forthe benchmarks AgV =1 (left) and BgV =3 (right). Lower panel: Total widths corresponding to di↵erentvalues of the coupling gV in the models A (left) and B (right).

Therefore, for model BgV =3

the dominant BRs are into di-bosons and the fermionic decays are

extremely suppressed, of the order of one percent to one per mil. Moreover, the total width

increases with increasing gV since it is dominated by the di-boson width which grows with gVas expected from Eq. (2.33). Finally, in model B we see that a very large coupling gV (the case

of gV = 8 is shown in the Figure) leads to an extremely broad resonance, with �/M � 0.1,

for which the experimental searches for a narrow resonance are no longer motivated. For

this reason we expect, if no further suppression is present in the parameter cH , to be able to

constrain heavy vector models from direct searches only up to gV of the order 6�7. For larger

couplings di↵erent searches become important, like for instance those for four fermion contact

interactions (see for instance Refs. [83, 84]).

16

heavy vector triplet (HVT)

10.1007/JHEP09(2014)060

1 Feynman diagrams

W 0W

H

q0

q

b

b

l

⌫

(a)

XW

W

g

g

q

q̄

l

⌫

(b)

g

g

g

q

q̄

c, b

c̄, b̄

(c)


1


Putting it all together

>currently, larger number of channels has been covered with 8 TeV data>differentiate between final states (number of leptons and jets)>violet analyses interpreted in dark matter scenarios

>several analyses repeated and improved w.r.t. 8 TeV (considering mX > 1TeV)>several more to come this year

31

W ➞ lν Z ➞ ll V ➞ qq Z ➞ νν H ➞ qqqq H ➞ ττ H ➞ bb H ➞ ɣɣ

W ➞ lν 13 TeV 13 TeVZ ➞ ll 13 TeV 13 TeV

V ➞ qq 13 TeV 13 TeV 13 TeV 13 TeV

Z ➞ νν 13 TeV 13 TeV

H ➞ qqqqH ➞ ττH ➞ bb 13 TeV 13 TeV 13 TeV

H ➞ ɣɣ


8 vs. 13 TeV

>While 13 TeV has opened up new energy regime, integrated luminosity recorded in 2015 significantly below the one of 2012 ! 20 fb-1 vs. ~3 fb-1

>LHC is hadron collider - √s ≠ energy available in collision

>need to consider parton luminosities

>exceed 2012 reach with 2015 data already at 1-2 TeV resonance mass

>nevertheless, worthwhile combining results

32

100 10001

10

100

gg Σqq qg

WJS2013

ratios of LHC parton luminosities: 13 TeV / 8 TeV

lum

ino

sity

ra

tio

MX (GeV)

MSTW2008NLO

_

[website]

http://www.hep.ph.ic.ac.uk/~wstirlin/plots/plots.html


Combination of diboson analyses

>example here: V’ combination in HVT model B

>seven 8 TeV analyses, three at 13 TeV

>how to combine upper cross section limits from two different √s?

33

(TeV)V'M1 1.5 2 2.5

(pb)

WV/

VH→

V'

BR

× 95

%σ

-210

-110

1 (8 TeV)-119.7 fbCMS Preliminary

Obs.S

Asympt. CLσ 1± Exp.

SAsympt. CL

σ 2± Exp. S

Asympt. CL=3)

V (gB HVT

lllv qqqqllqq qqbb(4q)lvqq ττqqlvbb

(TeV)V'M1 2 3 4

(pb)

WV/

VH→

V'

BR

× 95

%σ

-210

-110

1

10 (13 TeV)-12.2-2.6 fbCMS Preliminary

Obs.S


SAsympt. CL

σ 2± Exp. S

Asympt. CL=3)

V (gB HVT

lvqqllbb/lvbb/vvbbqqqq

CMS-PAS-B2G-16-007



>convert cross section limits into signal strength limits

>8+13 TeV limits comparable at lower masses, 13 TeV dominates high mass

> lower masses: leptonic analyses, higher masses: hadronic final states

34

(TeV)V'M1 2 3 4

theo

ryσ/

95%

σ

-110

1

10

210

(8 TeV)-1 (13 TeV) + 19.7 fb-12.2-2.6 fbCMS Preliminary

Obs.S


SAsympt. CL

σ 2± Exp. S

Asympt. CL=3)

V (gB HVT

lllv (8 TeV) qqbb(4q) (8 TeV)llqq (8 TeV) (8 TeV)ττqqlvqq (8 TeV) lvqq (13 TeV)lvbb (8 TeV) llbb/lvbb/vvbb (13 TeV)qqqq (8 TeV) qqqq (13 TeV)

(TeV)V'M1 2 3 4

theo

ryσ/

95%

σ

-110

1

10

210 (8 TeV)-1 (13 TeV) + 19.7 fb-12.2-2.6 fbCMS Preliminary

=3)V

(gB HVT

8 TeV

13 TeV

8+13 TeV

Obs.(solid)/ Exp.(dashed)S

Asympt. CL

CMS-PAS-B2G-16-007



>can translate observed limits into exclusion contours in the HVT couplings space

>additional input to theorists for model building

35

CMS-PAS-B2G-16-007

HcV

g-3 -2 -1 0 1 2 3

V/g Fc2 g

-1

-0.5

0

0.5

1

B

Mexpσ

≈ > 7% MthΓ

1.5 TeV2 TeV3 TeV

(8 TeV)-1 (13 TeV) + 19.7 fb-12.2-2.6 fb

CMSPreliminary

coupling to bosons

coup

ling

to fe

rmio

ns


Summary

>discovery of a diboson resonance might solve hierarchy problem

>however, currently no sign of new physics

>13 TeV results already exceed 8 TeV ones

>expect another boost with 2016 data

36


W-tagging calibration

>cutting on τ2/τ1-ratio ➜ need to know efficiency of cut in data and simulation

>select at generator level clean W-events and those that do not match

>perform simultaneous fit

38

pruned jet mass (GeV)40 50 60 70 80 90 100 110 120 130

Even

ts /

( 5 G

eV )

50

100

150

200

250

300

350tt single Top

VV W+jets

data HPν µ e/→W

CMSPreliminary

= 13TeV s at -1 2.2 fb

pruned jet mass (GeV)40 50 60 70 80 90 100 110 120 130

Even

ts /

( 5 G

eV )

20

40

60

80

100

120

140

160

180

200

220

240tt single Top

VV W+jets

data HPν µ e/→W

CMSPreliminary

= 13TeV s at -1 2.2 fb


boson-tagging efficiencies

39

Tagger BR(W/Z/H ➜ xx) efficiency mistag rate (q-/g-jets)

W/Z➜qq 70 % 35 % 1.2 %H➜bb 57 % 35 % 0.5 %H➜WW➜qqqq 10 % 35 % 1.5 %H➜ττ 6 % 35 % 0.03 %


Zɣ search limits

> recently published Z➜ll + ɣ search

40

CMS-PAS-EXO-16-019

5.4% width0.014% width


) [fb

]γZ

→ B

R(A

× σ

95%

CL

UL

on

0

50

100

150

200

250

300

W = 0.014%

Observed

σ 1±Expected

σ 2±Expected



) [fb

]γZ

→ B

R(A

× σ

95%

CL

UL

on

0

100

200

300

400

500

600

W = 5.6%

Observed

σ 1±Expected

σ 2±Expected


Documents

Search for diboson resonances at CMS - DESYphysikseminar.desy.de/sites2009/site_physikseminar/content/e212/e... · Search for diboson resonances at CMS identifying highly energetic