New$method$for$precise$determinaon$ …of$the$analysis t t b b ν l + j j...
Transcript of New$method$for$precise$determinaon$ …of$the$analysis t t b b ν l + j j...
New method for precise determina2on of top quark mass at LHC
Sayaka Kawabata (Tohoku University)
in collabora2on with Y. Shimizu (Tohoku Univ.) Y. Sumino (Tohoku Univ.) H. Yokoya (Toyama Univ.)
1. Introduc2on
2. Weight func2on method
3. Top mass reconstruc2on
4. Summary and future work
Outline
1. Introduc2on
• EW precision tests for SM
• beyond SM
• SM vacuum stability
w Heaviest elementary par2cle mt 〜 173 GeV
Top quark and its mass
Mo2va2on to measure the top mass
w Top mass is an important input parameter to
w One of the fundamental parameters of the SM
Snowmass 2013
EW precision tests for SM
Radia2ve correc2ons
Radia2ve correc2ons are oUen sensi2ve to the top quark mass
It is important to test for devia2ons from SM predic2ons in the EW sector
Beyond SM For example, SUSY
Feng et al., PRL 111,131802(2013)
The predic2ons beyond the SM oUen depend strongly on the value of the top quark mass.
SM vacuum stability
Precision measurements of mt , αs , mH are needed.
Degrassi et al. ’12
Is the SM vacuum stable or metastable ?
Stable Meta-‐stable
Buaazzo et al. ’13
Tevatron Main produc2on process : a pair produc2on
• pp collider • discovered the top quark in 1995 • shut down in 2011
Top quarks at Tevatron and LHC
LHC
• pp collider • ended data taking at √s = 7, 8 TeV • restart in 2015 at √s = 13/14 TeV
σ(a) 〜 8pb
σ(a) 〜 900pb ( √s = 14 TeV)
t
t
b
b
ννννν
W+
W-‐
l + ν l - ν
l + ν j j
j j j j
a decay channel
Cross sec2on S / N
Dilepton Lepton+jets All hadronic
Small Medium Large
Very good Good
Not good
All hadronic
Lepton+jets Dilepton
44%
a branching ra2o
Main method to measure the top mass Template Method : compare some distribu2ons of data with templates from MC
1. Reconstruct invariant top mass (mtreco, mt
reco(2)) and invariant W mass (mjj) in each event
2. Create MC distribu2ons at discrete values of mt and correc2on factor of JES (templates), and interpolate
3. Fit the MC distribu2on to data
t
t
b
b
l ν j j
mtreco mjj ( j : non-‐b-‐tagged jet ) mt
reco(2)
Example
Current status of top mass measurements (direct measurement)
0.4% precision!
Tevatron+LHC mt combina2on arXiv:1403.4427
Measured mass : mtPythia
jets
Signal modeling in MC
w Measurements use jet momenta and fit the data with MC
� Assume some model
� Cannot be treated within perturba2ve QCD
w Hadroniza2on in MC
The measured mass is hadroniza2on-‐model dependent (mtPythia)
mtPythia ≠ mt
pole
Cannot evaluate the difference between mtPythia and mt
pole
PYTHIA
Top quark mass?
Far from a fundamental param.
� mtpole : Top quark has a color, so the physical
on-‐shell quark cannot exist
Known as a good parameter in pert. QCD
� mtMS : Short-‐distance mass
Free from IR contamina2on
� mtPythia : Not a parameter defined in perturba2ve theory
Important to determine mtMS accurately
Measurement of MS mass (and pole mass) ( from a cross sec2on )
MS mass (Tevatron)
The measured MS mass s2ll has a large error
pole mass (CMS)
=
� The sensi2vity of σa to mt is not so strong
� Theore2cal uncertain2es 〜 1.5 -‐ 2 GeV
2. Weight func2on method SK, Y.Shimizu, Y.Sumino, H.Yokoya, PLB 710, 658 (2012) SK, Y.Shimizu, Y.Sumino, H.Yokoya, JHEP 08, 129 (2013)
A new method to measure the mass of parent par2cle.
� Only lepton energy distribu2on is needed Free from the ambiguity of hadroniza2on models
Independent of produc2on process of top quark
� Does not depend on the veloci2es of top quarks
We can determine the MS mass of top quark
t
b
l +
jet
ν
Weight func2ons and the weighted integrals D(El):実験室系でのレプトンエネルギー分布
I (m) ≡!!!
dEl D(El)W (El,m)
W (mtrueH ) ≡
!!!dEl D(El)W (El,m
trueH ) = 0
W (El ,m) =
!!!dE D0(E ;m)
1
EEl×( ρの奇関数 )|eρ=El/E
D0 (E ;m) : ヒッグス(質量m )の静止系でのレプトンエネルギー分布
W exp
W MC
W (m)
W (m) = 0
2
Weight func2on
Lepton energy distribu2on in the lab. frame
There exist an infinite number of weight func2ons which sa2sfy
t
b
l +
ν
I ( m = mttrue ) = 0 for an arbitrary velocity distribu2on of top quarks
160 165 170 175 180 185!4
!2
0
2
4
m !GeV"
106"I
W (m) = 0
D(El) (1)
W (El,m) =!
dE D0(E, m)1
EEl× ( ρの奇関数 )|eρ=El/E (2)
t t + Wt, WW+ jets
t t, Wt, WW+jets
t t + Wt, WW + jets
mtruet (3)
3
large velocity
small velocity
I (m)
!1.0 !0.5 0.0 0.5 1.0!3
!2
!1
0
1
2
3
Ρ
ρ = ln (El / E0 )
Explicit form of weight func2ons
Lepton energy distribu2on
0.0 0.5 1.0 1.5 2.0 2.50.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
El ! E0
The rest frame of X
A boosted frame 1ΓdΓdEl
(X is scalar or unpolarized) For a two-‐body decay : X l + Y
Explicit form of weight func2ons (X is scalar or unpolarized) For a two-‐body decay : X l + Y
ρ = ln (El / E0 )
Lepton energy distribu2on
0.0 0.5 1.0 1.5 2.0 2.50.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
El ! E0
The rest frame of X
A boosted frame 1ΓdΓdEl
!!!dEl D(El)W (El,m
trueH ) = 0
!!!dρ ( ρの偶関数 ) × ( ρの奇関数 ) = 0
D0 = δ(E − E0)
dρ ∝ e−ρdEl
W (El,m)
W (m = mtrueH ) = 0
D(El):実験室系でのレプトンエネルギー分布
W (m) ≡!!!
dEl D(El)W (El,m)
W (mtrueH ) ≡
!!!dEl D(El)W (El,m
trueH ) = 0
1
!!!dEl D(El)W (El,m
trueX ) = 0
!!!dρ ( ρの偶関数 ) × ( ρの奇関数 ) = 0
D0 = δ(E − E0)
dρ ∝ e−ρdEl
W (El,m)
W (m = mtrueH ) = 0
D(El):実験室系でのレプトンエネルギー分布
W (m) ≡!!!
dEl D(El)W (El,m)
W (mtrueH ) ≡
!!!dEl D(El)W (El,m
trueH ) = 0
1
!1.0 !0.5 0.0 0.5 1.0!3
!2
!1
0
1
2
3
Ρ
Can be expressed as a superposi2on of lepton distribu2on for a two-‐body decay
(X is scalar or unpolarized) For a many-‐body decay : X l + anything
0.0 0.5 1.0 1.5 2.0 2.50.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
El ! E0δ -‐func2on
two-‐body decay
A weight func2on would be also a superposi2on of that for a two-‐body decay
Explicit forms of weight func2ons
Lepton energy distribu2on in the rest frame of X
0 20 40 60 80 1000.000
0.005
0.010
0.015
0.020
El !GeV"
n = 2n = 3n = 5n = 15
m = 173GeV
0 20 40 60 80 100 120 140!1.0!0.8!0.6!0.4!0.2
0.00.2
El !GeV"
103 "
W m = 173 GeV
Examples of weight func2ons For a top quark decay : t Wb l νb
Summary of the weight func2on method
2. Use the lepton energy distribu2on measured by experiment as D(El)
1. Construct weight func2ons for the process
Lepton energy dist. in the rest frame of parent par2cle, which can be calculated in pert. QCD
160 170 180 190!1.0
!0.5
0.0
0.5
1.0
m !GeV"
I
I
mtrue
I (m)
I ( m = mtrue ) = 0
3. Obtain the zero of I (m) as mtrue
★ We can use this method if parent par2cle is scalar or unpolarized to determine any parameter which enter
3. Top mass reconstruc2on (Simula2on analysis: LO)
arXiv:1405.2395 [hep-‐ph]
� Weight func2ons used in this analysis
n = 2n = 3n = 5n = 15
m = 173GeV
0 20 40 60 80 100 120 140!1.0!0.8!0.6!0.4!0.2
0.00.2
El !GeV"
103 "
W
Setup of the analysis
t
t
b
b
νl +
j j
a events, Lepton(μ)+jets channel LHC √s = 14 TeV
� Signal
Small weight at El 〜 0
� Background W+jets, other a events
The polariza2on of top quarks at LHC ? Almost unpolarized
SM: B 〜 0.003, ATLAS: compa2ble with zero (ΔB〜0.08)
Weight func2on W(El ,m)
n = 2n = 3n = 5n = 15
m = 173GeV
0 20 40 60 80 100 120 140!1.0!0.8!0.6!0.4!0.2
0.00.2
El !GeV"
103 "
W
In principle, our method works
0 50 100 150 200 250 3000
50000
100000
150000
200000
250000
300000
350000
El !GeV"
Events#5Ge
V
Lepton energy distribu2on at parton level (signal)
Parton level analysis
n = 3n = 5n = 15
n = 2
160 165 170 175 180 185!3
!2
!1
0
1
2
3
m !GeV"10
6"
I
input mt
Δmt ≈ +0.7GeV
Effect of Γt : +0.34GeV MC stat. error : 0.4GeV Consistent with expecta2on
Effect of lepton cuts
In prac2ce, event selec2on cuts and backgrounds deform the lepton distribu2on.
160 165 170 175 180 185!20
0
20
40
60
80
m !GeV"106"I
Before lepton cuts (×10)
AUer lepton cuts
0 50 100 150 200 250 3000
50000
100000
150000
200000
250000
El !GeV"
D
The major effect is from the lepton cuts : pT(l) >20 GeV, |η(l)| < 2.4
0 10 20 30 400
20000
40000
60000
80000
pTHlL HGeVL
Eventsê0.5
8GeV
Compensated MC events
0 50 100 150 200 250 3000
50000
100000
150000
200000
250000
300000
350000
El HGeVL
Eventsê5G
eV
Compensated MC events with mtc
Events aUer lepton cuts with mt
input = 173GeV
We compensate for the loss using MC events. A solu2on to the problem of lepton cuts
� To fix the normaliza2on of the added events, perform a χ2 fit so that pT(l) distribu2ons are connected smoothly
� We need to assume some value for mt of the compensated MC events
A solu2on to the problem of lepton cuts
176 GeV 179 GeV
mt = 167 GeV
173 GeV 170 GeV
160 165 170 175 180 185!3
!2
!1
0
1
2
3
4
m !GeV"
106"
I
c
n=2
The weighted integrals with various mtc
The zeros of I(m) come close to the input mass (mt
input = 173GeV)
166 168 170 172 174 176 178 180
!5
0
5
10
mtMC !GeV"
mtrec!mtMC!GeV
"zero of I(m
) − m
tc
mtc (GeV)
Δmt = +0.5GeV
mtc = mt
input ⇒ zero of I(m) = mtc
mtc ≠ mt
input ⇒ zero of I(m) ≠ mtc (guess)
Event selec2on cuts
Cross sec2on aUer all cuts
t
t
b
b
ννννν
W+
W-‐
l + ν j j
� 1 muon with pT > 20GeV, |η| < 2.4 (lepton cuts)
� At least 4 jets
� At least 1 b-‐tag with the b-‐tag efficiency 0.4 independent of pT and η
� pT(j1) > 55, pT(j2) > 25, pT(j3) > 15, pT(j4) > 8GeV
In addi2on to lepton cuts, we impose the following cuts to events:
We do not use cuts concerning missing momenta.
Signal (mt=173GeV) W+jets BG Other a BG
22.4 pb 1.8 pb 5.7 pb
Top mass reconstruc2on (aUer event selec2on cuts)
176 GeV 179 GeV
mt = 167 GeV
173 GeV 170 GeV
160 165 170 175 180 185
!2
0
2
4
m !GeV"
106"
I
n=2
c
The weighted integrals with various mtc
mtc = mt
input ⇒ zero of I(m) = mtc
mtc ≠ mt
input ⇒ zero of I(m) ≠ mtc (guess) 166 168 170 172 174 176 178 180
!5
0
5
10
mtMC !GeV"
mtrec!mtMC!GeV
"zero of I(m
) − m
tc Δmt = 1.2GeV
mtc (GeV)
Sensi2vity of mass determina2on
Signal stat. error μF scale JES BG stat. error n = 2 0.4 +1.5/-‐1.6 +0.0/-‐0.1 0.4
3 0.4 +1.5/-‐1.5 +0.1/-‐0.3 0.4
5 0.5 +1.4/-‐1.4 +0.2/-‐0.4 0.5
15 0.5 +1.5/-‐1.3 +0.2/-‐0.6 0.6
� Assuming that the error of electron mode is the same as the muon mode
� At 100 �-‐1
� Lepton(e,μ)+jets channel
Uncertain2es [GeV]
Can be improved by including NLO
n = 3n = 5n = 15
n = 2
165 170 175 180160
165
170
175
180
185
Input top mass !GeV"
mtre
c!GeV
" mtrec = mt
input n=2
� The es2mated stat. error is about 0.4GeV with 100�-‐1. Major systema2c uncertain2es are under good control.
� More precise measurements of mt are needed.
� We performed a simula2on analysis of top mass reconstruc2on with lepton+jets channel at LO.
� The problem of the lepton cuts can be solved by compensa2ng lepton distribu2on with MC events.
� We proposed a new method to measure a theore2cally well-‐defined top quark mass at LHC.
4. Summary and future work
Future work
★ NLO, NNLO
� Include NLO, NNLO correc2ons to the top decay process in weight func2ons.
� Include NLO correc2ons to the top produc2on process in MC.
★ Effects of top off-‐shellness
mtpole, mt
MS
μF scale uncertain2es can be reduced