Heavy flavor jet-tagging, W+b,c-jet, and top measurements ... · Introduction Overview • JINST 10...
Transcript of 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)
Ilten W + b and top with LHCb August 4, 2015 2 / 27
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)
Ilten W + b and top with LHCb August 4, 2015 3 / 27
b, c-jet Tagging
b, c-jet Tagging
Ilten W + b and top with LHCb August 4, 2015 4 / 27
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 <
Ilten W + b and top with LHCb August 4, 2015 5 / 27
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
Ilten W + b and top with LHCb August 4, 2015 6 / 27
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
Ilten W + b and top with LHCb August 4, 2015 7 / 27
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
Ilten W + b and top with LHCb August 4, 2015 8 / 27
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
Ilten W + b and top with LHCb August 4, 2015 9 / 27
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
Ilten W + b and top with LHCb August 4, 2015 10 / 27
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
Ilten W + b and top with LHCb August 4, 2015 11 / 27
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
Ilten W + b and top with LHCb August 4, 2015 12 / 27
W + b, c-jet Ratios
W + b, c-jet Ratios
Ilten W + b and top with LHCb August 4, 2015 13 / 27
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
Ilten W + b and top with LHCb August 4, 2015 14 / 27
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
Ilten W + b and top with LHCb August 4, 2015 15 / 27
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
Ilten W + b and top with LHCb August 4, 2015 16 / 27
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
Ilten W + b and top with LHCb August 4, 2015 17 / 27
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
Ilten W + b and top with LHCb August 4, 2015 18 / 27
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
Ilten W + b and top with LHCb August 4, 2015 19 / 27
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
Ilten W + b and top with LHCb August 4, 2015 20 / 27
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
Ilten W + b and top with LHCb August 4, 2015 21 / 27
Forward Top Measurement
Forward Top Measurement
Ilten W + b and top with LHCb August 4, 2015 22 / 27
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
Ilten W + b and top with LHCb August 4, 2015 23 / 27
Forward Top Measurement
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 ∞
Ilten W + b and top with LHCb August 4, 2015 24 / 27
Forward Top Measurement
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 ∞
Ilten W + b and top with LHCb August 4, 2015 25 / 27
Forward Top Measurement
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)
Ilten W + b and top with LHCb August 4, 2015 26 / 27
Forward Top Measurement
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
Ilten W + b and top with LHCb August 4, 2015 27 / 27
Appendix
Ilten W + b and top with LHCb August 4, 2015 1 / 2
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
Ilten W + b and top with LHCb August 4, 2015 2 / 2