Heavy flavor jet-tagging, W+b,c-jet, and top measurements ... · Introduction Overview • JINST 10...

Heavy flavor jet-tagging, W+b, c-jet, and topmeasurements with LHCb

Philip Iltenon behalf of the LHCb Collaboration

Massachusetts Institute of Technology

DPF2015

Ilten W + b and top with LHCb August 4, 2015 1 / 27

Introduction

Overview

• JINST 10 (2015) P06013: b and c-jetidentification performance

• arXiv:1505.04051: W + udsg, b, c-jetratios

• arXiv:1506.00903: forward topobservation and cross-section

• 1 fb−1 pp,√

s = 7 TeV (2011)• 2 fb−1 pp,

√s = 8 TeV (2012)

• excellent luminosity uncertainty,JINST 9 (2014) P12005

• 1.71% for 7 TeV dataset• 1.16% for 8 TeV dataset

date [day/month]28/02 09/04 19/05 29/06 08/08 18/09 28/10

L[ fb−

1]

0

1

2

3

4

5

6LHCbATLASCMS

√s = 7 TeV

(2011)

date [day/month]28/02 18/04 05/06 23/07 09/09 27/10 14/12

L[ fb−

1]

0

5

10

15

20

25

30LHCbATLASCMS

√s = 8 TeV

(2012)


http://arxiv.org/abs/1504.07670




Introduction

Detector JINST 3 (2008) S08005ĲMPA 30 (2015) 1530022

VELO

RICH1

TT

magnet

OT

IT

T1

T2

T3

RICH2

M1

M2 M3 M4 M5

HCAL

ECALmuon system5 m

5 m 15 mz-axis →

y-ax

is→

SPD/PRS

• fully instrumented between 2 < η < 5• momentum resolution between 0.4% at 5 GeV to 0.6% at 100 GeV• impact parameter resolution of 13− 20 µm for tracks• secondary vertex precision of 0.01− 0.05(0.1− 0.3) mm in xy(z)


http://dx.doi.org/10.1088/1748-0221/3/08/S08005


b, c-jet Tagging

b, c-jet Tagging


b, c-jet Tagging

Jet Reconstruction JHEP 1401 (2014) 033

• standard particle flow algorithm• anti-kT with R = 0.5• flat jet energy resolution (JER)

of ≈ 20%• from Z + 1-j with

∆φ(Z , j) ≈ π• jet reconstruction efficiency of≈ 95%

• jet fiducial definition:• pT(j) > 20 GeV• 2.2 < η(j) < 4.2• reduced from full• uniform tag and

reconstruction

ZT

p / jetT

p0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

Frac

tion

of e

vent

s

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

Data

Simulation

LHCb

[GeV]jetT

p20 40 60 80 100 120

jet

ε

0.70

0.75

0.80

0.85

0.90

0.95

1.00

LHCb simulation

< 4.5jetη2 <


http://arxiv.org/abs/arXiv:1310.8197

b, c-jet Tagging

Secondary Vertex Tagger (1) JINST 10 (2015) P06013

• build 2-body SVs• n-body SVs from linking 2-body

SVs with shared tracks• require vertex flight direction

within jet, ∆R(SV, j) < 0.5• two BDTs

• BDT(bc|udsg): separatesudsg-jet from b, c-jet (BDT0)

• BDT(b|c): separates b-jet fromc-jet (BDT1)

SV

jet

PV



b, c-jet Tagging

Secondary Vertex Tagger (2) JINST 10 (2015) P06013

variable separation variable separationM (SV) udsgc b Mcor(SV) udsgb cmin(FDT(SV)) udsg cb pT(SV)/pT(j) udsg cb∆R(SV, j) udsg cb N (trk) udsgc bN (trk ∈ j) udsgc b |Q(SV)| udsgb clog(χ2

FD(SV)) all log(χ2IP(SV)) all

Mcor(SV) N (trk)

[GeV]corMSV 0 2 4 6 8 10

cand

idat

es

0

5000

10000LHCb

databc

udsg

(tracks)NSV 2 4 6 8 10

cand

idat

es

0

5000

10000

15000

20000 LHCbdatabc

udsg



b, c-jet Tagging

Jet Flavor Determination (1) JINST 10 (2015) P06013

• fit 2-dimensional BDT(bc|udsg) versus BDT(b|c) distributions

)udsg|bcBDT(-1 -0.5 0 0.5 1

)c|bB

DT

(

-1

-0.5

0

0.5

1

LHCb simulation

-jetsudsg

)udsg|bcBDT(-1 -0.5 0 0.5 1

)c|bB

DT

(

-1

-0.5

0

0.5

1

LHCb simulation

-jetsc

)udsg|bcBDT(-1 -0.5 0 0.5 1

)c|bB

DT

(

-1

-0.5

0

0.5

1

LHCb simulation

-jetsb

• validate with four tag+probesamples

• B + j : b-enhanced• D + j : c,b-enhanced• displaced-µ+ j : c,b-enhanced• isolated-µ+ j : udsg-enhanced

SV?

probe-jet

tag,�e.g.�B-decay

tag-jet

PV



b, c-jet Tagging

Jet Flavor Determination (2) JINST 10 (2015) P06013

c,b-enhanced (D + j)

0

200

400

600

800

1000

)udsg|bcBDT(-1 -0.5 0 0.5 1

)c|bB

DT

(

-1

-0.5

0

0.5

1

LHCb data

+jetD0

200

400

600

800

1000

)udsg|bcBDT(-1 -0.5 0 0.5 1

)c|bB

DT

(

-1

-0.5

0

0.5

1

LHCb fit

)udsg|bcBDT(-1 -0.5 0 0.5 1

cand

idat

es

0

2000

4000

6000

8000 LHCbdatabc

udsg

)c|bBDT(-1 -0.5 0 0.5 1

cand

idat

es

0

2000

4000

6000 LHCbdatabc

udsg



b, c-jet Tagging

Efficiencies (1) JINST 10 (2015) P06013

• determine efficiency with: Nx(SV)Nx(χ2

IP) , x ∈ udsg, c, b

c,b-enhanced (D + j) b-enhanced (B + j)

χ2IP of hardest-pT

track, very largeudsg-jet contam-ination prior totagging )

IP2χlog(

-5 0 5 10 15

cand

idat

es

0

50000

100000

150000 LHCbdatabc

udsg

+jetD

)IP2χlog(

-5 0 5 10 15

cand

idat

es

0

10000

20000

LHCbdatabc

udsg

+jetB

χ2IP of hardest-pT

muon, only O(10%)of jets

)IP2χmuon log(

-5 0 5 10 15

cand

idat

es

0

500

1000

1500

2000

LHCbdatabc

udsg

+jetD

)IP2χmuon log(

-5 0 5 10 15

cand

idat

es

0

1000

2000

LHCbdatabc

udsg

+jetB



b, c-jet Tagging

Efficiencies (2) JINST 10 (2015) P06013

light-parton mistag probability0.001 0.002 0.003 0.004 0.005

(b,c

)-je

t tag

eff

icie

ncy

0

0.2

0.4

0.6

0.8

1

-jetb-jetc

LHCb simulation

(jet) < 4.2η2.2 <

(jet) < 100 GeVT

20 < p

(jet) [GeV]T

p20 40 60 80 100

SV-t

ag e

ffic

ienc

y

0

0.2

0.4

0.6

0.8

1LHCb

-jetb

-jetc



b, c-jet Tagging

Systematics JINST 10 (2015) P06013

source δεb−jets [%] δεc−jets [%]

BDT templates∗ ≈ 2 ≈ 2udsg-jet large IP component∗ ≈ 5 ≈ 10− 30IP resolution − −muon mis-ID (hardest-µ only) 5 20out-of-jet (b, c)-hadron decay − −gluon splitting 1 1pile up − −

systematic (combined fit) ≈ 10 ≈ 10∗dependent on jet type and pT



W + b, c-jet Ratios

W + b, c-jet Ratios


W + b, c-jet Ratios

W + j Measurements arXiv:1505.04051

• use W → µν final state• measure ratios and asymmetries

• σ(W c)σ(Wj) , σ(W b)

σ(Wj) , σ(W +j)σ(Zj) , σ(W−j)

σ(Zj)

• A(WX) ≡ σ(W +X)− σ(W−X)σ(W +X) + σ(W−X)

• A(W c), A(W b)

• fiducial definition• pT(µ) > 20 GeV, 2.0 < η(µ) < 4.5• pT(j) > 20 GeV, 2.2 < η(j) < 4.2• ∆R(µ, j) > 0.5• pT(µ+ j) > 20 GeV



W + b, c-jet Ratios

Signals and Backgrounds arXiv:1505.04051

W + udsg-jet W + c-jet W + b-jet

g

qi

qj

W

g

s

c

W

qj

qi

b

b

W

Z+jet di-jet

q

q

g

Z

g

g

q

q

Z → ττ top

Z

q

q

τ

τ

g

qi

b

t

qj



W + b, c-jet Ratios

Analysis Strategy arXiv:1505.04051

• selection:• fiducial requirements except pT(µ+ j)→ pT(jµ + j)• hardest-pT muon candidate, jet containing muon is ju• hardest-pT jet candidate from same primary vertex

• W + j content from isolation fit• BDT(bc|udsg) and BDT(b|c) fit (same BDT0 and BDT1)• W + b-jet: top extrapolated from side-band• W + c-jet: Z → ττ from pT(SV)/pT(j) fit



W + b, c-jet Ratios

W + j Determination arXiv:1505.04051

• isolation defined as pT(µ)/pT(jµ)• fit in bins of

√s and muon charge

• di-jet template from pT-balanced events, pT(jµ + j) < 10 GeV• Z + j yield and template extrapolated from di-muon Z + j data• W + j template from di-muon Z + j data, corrected to W + j with

simulation

0.5 0.6 0.7 0.8 0.9 1

Can

dida

tes/

0.05

20000

40000 Data

W

Z

Jets

= 8 TeVs, +µ

) µj(T

p)/µ(T

p0.5 0.6 0.7 0.8 0.9 1

cand

idat

es

20000

40000 = 8 TeVs, −µ LHCb



W + b, c-jet Ratios

Flavor Determination (1) arXiv:1505.04051

• fit BDT(bc|udsg) versus BDT(b|c) distribution in each bin of√

s,muon charge, and pT(µ)/pT(jµ) (bin of 0.9− 1.0 below)

0

10

20

30

40

50

60

70

80

90

)udsg|bcBDT(-1 -0.5 0 0.5 1

)c|bB

DT

(

-1

-0.5

0

0.5

1

LHCb data

0

10

20

30

40

50

60

70

80

90

)udsg|bcBDT(-1 -0.5 0 0.5 1

)c|bB

DT

(

-1

-0.5

0

0.5

1

LHCb fit

)udsg|bcBDT(-1 -0.5 0 0.5 1

Can

dida

tes/

0.1

0

200

400

600LHCb

Databc

udsg

)c|bBDT(-1 -0.5 0 0.5 1

Can

dida

tes/

0.1

0

200

400

600

LHCbDatabc

udsg



W + b, c-jet Ratios

Flavor Determination (2) arXiv:1505.04051

0.5 0.6 0.7 0.8 0.9 1

Can

dida

tes/

0.1

0

500

1000 = 8 TeVs, +µLHCb

-jetb+µ

) µj(T

p)/µ(T

p0.5 0.6 0.7 0.8 0.9 1

cand

idat

es

0

500

1000 = 8 TeVs, −µ Data

W

Z

Jets

0.5 0.6 0.7 0.8 0.9 1

Can

dida

tes/

0.1

0

500

Data

W

Z

Jets

= 8 TeVs, +µLHCb-jetc+µ

) µj(T

p)/µ(T

p0.5 0.6 0.7 0.8 0.9 1

cand

idat

es

0

500

= 8 TeVs, −µ



W + b, c-jet Ratios

Systematics arXiv:1505.04051

source δσ(Wb)σ(Wj)

[%] δσ(Wc)σ(Wj)

[%] δσ(Wj)σ(Zj)

[%] δA(Wb) δA(Wc)

(b, c)-tag efficiency 10 10 − −isolation templates 10 5 4 0.08 0.03top 13 − − 0.02SV-tag BDT templates 5 5 0.02 0.02Z [ττ ] − 3 − − −jet reconstruction 2 2 − − −jet energy 2 2 1 0.02 0.02trigger and selection 1 1 2 − −W [τν] − − 1 − −other electroweak − − − − −

systematic 20 13 5 0.09 0.04statistical (7 TeV) 20 8 2 0.20 0.08statistical (8 TeV) 10 5 1 0.13 0.05



W + b, c-jet Ratios

Results arXiv:1505.04051 1

00×

(Wj)

σ(W

x)/

σ

0

1

2

3

4

5

6

7 = 7 TeV LHCb datas = 8 TeV LHCb datas

(Zj)

σ(W

j)/σ

5

6

7

8

9

10

11

12 = 7 TeV LHCb datas = 8 TeV LHCb datas

A(W

x)

-0.2

0

0.2

0.4

0.6

0.8 = 7 TeV LHCb datas = 8 TeV LHCb datas points data (total, stat)fills MCFM NLO theory

CT10 (scale + PDF)green W + c-jetred W + b-jetblue W + + jmagenta W− + j



Forward Top Measurement




Analysis Strategy arXiv:1506.00903

• tightened fiducial region• pT(µ) > 25 GeV reduces di-jet background• 50 < pT(b) < 100 GeV reduces W + b

• similar analysis strategy to W + b, c ratios• fit pT(µ+ b) and A distributions to determine significance• use excess from W + b prediction to calculate cross-section

)µ

j(T

p)/µ(T

p0.5 0.6 0.7 0.8 0.9 1

Can

dida

tes

/ 0.0

5

5000

10000

Data

W

Z

Jets

LHCb+jetµ

)µ

j(T

p)/µ(T

p0.5 0.6 0.7 0.8 0.9 1

Can

dida

tes

/ 0.1

100

200

300

Data

W

Z

Jets

LHCb-tag+bµ




Background arXiv:1506.00903

) [GeV]j+µ(T

p

+je

t)W(

N

0

2000

4000

6000

8000 LHCb

Data

SM

20 45 70 95 ∞) [GeV]j+µ(

Tp

Cha

rge

Asy

mm

etry

-0.4

-0.2

0

0.2

0.4

20 45 70 95 ∞

LHCb

Data

SM

• constrain W + b backgroundusing W + j from data andW + b/W + j from theory

• theory and experimentaluncertainties partiallycancel

• validate against W + c ) [GeV]c+µ(T

p

)c+

W(N

0

50

100

150

200

Data

SM

LHCb

20 45 70 95 ∞




Significance arXiv:1506.00903

• profile likelihood used tocompare W + b hypothesis withW + b + top

• uncertainties treated asGaussian nuisanceparameters

• 5.4σ significance observed

source δσ [%]

GEC 2pT(µ)/pT(jµ) templates 5–10jet reconstruction 2SV-tag BDT templates 5b-tag efficiency 10

trigger & µ selection 2jet energy 5W [τ [µνν]ν] 1

luminosity 1–2

) [GeV]b+µ(T

p

)W

+b

(N

0

100

200

Data+topWb

Wb

LHCb

20 45 70 95 ∞) [GeV]b+µ(

Tp

Cha

rge

Asy

mm

etry

-0.4

-0.2

0

0.2

0.4

Data

+topWb

Wb

LHCb

20 45 70 95 ∞




Cross-section arXiv:1506.00903

• cross-section determined from subtracting W + b background fromdata

• total top yield (background subtracted W + b) of 220± 39

) [fb

]t

+ t

+

t(tσ

100

150

200

250

300

350

400

= 7 TeV LHCb datas = 8 TeV LHCb datas

σ(tt̄ + t + t̄)points data (total, stat)fills MCFM NLO theory

CT10 (scale + PDF)




Conclusions and Outlook

• robust inclusive heavy flavor tagging algorithm implemented• cut on BDT(bc|udsg) and BDT(b|c), or fit• 25% c-jet and 65% b-jet tagging efficiencies, 0.3% udsg-jet mistag• efficiency well modeled by simulation• fully data driven method using two techniques• systematic uncertainty ≈ 10%

• unique forward measurement of W + udsg, c, b-jet ratios• results in agreement with theory predictions• methods validated for Run II measurements

• forward top cross-section measurement• observation of 5.4σ• Run II expectations for additional 5 fb−1, LHCb-TALK-2015-113• [`, b]: expect ≈ 8300 tt̄, 5000 t-channel, 600 s-channel, and 180 Wt• [`, `, b, b]: expect ≈ 530 tt̄


https://cds.cern.ch/record/2022847

Appendix


Trigger

Trigger JINST 8 (2013) P04022

• movable L0 thresholds andsoftware HLT

• low pT leptons/photons andtracks (IP > 0.1 mm)

L0

µ: pT > 1.5 GeV

340 kHz

µµ:√

p1Tp2

T > 1.3 GeV

75 kHz

h: ET > 3.5 GeV

405 kHz

e: ET > 2.5 GeV

160 kHz

γ: ET > 2.5 GeV

80 kHz

µ: pT > 4.8(1) GeV

0.7(5) kHz

µµ: m > 2.7(1)GeV

1.2(1.3) kHz

track: pT > 1.7 GeV

33 kHz

e: pT > 10 GeV

. . .

µ: pT > 10(4.8) GeV

µµ: m > 4.8(3) GeV

displaced vertex

inclusive topological

beauty BBDT

inclusive and

exclusive charm

. . .

15 MHz 5 kHz

HLT1 HLT2

1 MHz 50 kHz



Heavy flavor jet-tagging, W+b,c-jet, and top measurements ... · Introduction Overview • JINST 10...

Documents

Transcript of Heavy flavor jet-tagging, W+b,c-jet, and top measurements ... · Introduction Overview • JINST 10...