[PPT] - Top-quark mass determination using new NLO+PS generators Silvia PowerPoint Presentation

SLIDE 1

Top-quark mass determination using new NLO+PS generators

Silvia Ferrario Ravasio*

Milan Christmas Meeting, 20th December 2017 *In collaboration with T. Jeˇ zo, P. Nason and C. Oleari [1712.XXXX]

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 1/16

SLIDE 2

Monte Carlo Event generators

Standard methods to infer mt are based on the use of MC event generators to mimic top-pair production process.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 2/16

SLIDE 3

Monte Carlo Event generators

Standard methods to infer mt are based on the use of MC event generators to mimic top-pair production process. Current standard NLO+PS: hard process described with NLO accuracy, further emissions handled by the PS in the soft and collinear limit.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 2/16

SLIDE 4

Monte Carlo Event generators

Standard methods to infer mt are based on the use of MC event generators to mimic top-pair production process. Current standard NLO+PS: hard process described with NLO accuracy, further emissions handled by the PS in the soft and collinear limit. POWHEG BOX is an NLO event generator, based on the POWHEG

method. It generates the hardest emission. The event is then

completed by standard SMC that implements the PS. [arXiv: hep-ph/0409146]

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 2/16

SLIDE 5

Monte Carlo Event generators

Standard methods to infer mt are based on the use of MC event generators to mimic top-pair production process. Current standard NLO+PS: hard process described with NLO accuracy, further emissions handled by the PS in the soft and collinear limit. POWHEG BOX is an NLO event generator, based on the POWHEG

method. It generates the hardest emission. The event is then

completed by standard SMC that implements the PS. [arXiv: hep-ph/0409146] Vetoed shower: emissions harder than the first one are vetoed.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 2/16

SLIDE 6

Monte Carlo Event generators

Standard methods to infer mt are based on the use of MC event generators to mimic top-pair production process. Current standard NLO+PS: hard process described with NLO accuracy, further emissions handled by the PS in the soft and collinear limit. POWHEG BOX is an NLO event generator, based on the POWHEG

method. It generates the hardest emission. The event is then

completed by standard SMC that implements the PS. [arXiv: hep-ph/0409146] Vetoed shower: emissions harder than the first one are vetoed. The SMC Pythia and Herwig offer the possibility to complete events generated with POWHEG BOX (LHIUP).

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 2/16

SLIDE 7

Top pair production in POWHEG BOX

Three current implementation of top pair production in POWHEG BOX

1 hvq [arXiv:0707.3088]

⇒ NLO corrections in production. ⇒ Decay performed at LO using reweighting. ⇒ Approximate spin correlation and offshell effects.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 3/16

SLIDE 8

Top pair production in POWHEG BOX

Three current implementation of top pair production in POWHEG BOX

1 hvq [arXiv:0707.3088]

⇒ NLO corrections in production. ⇒ Decay performed at LO using reweighting. ⇒ Approximate spin correlation and offshell effects.

2 t¯

tdec [arXiv:1412.1828] ⇒ NLO corrections in production and decay using NWA. ⇒ Spin correlation and offshell effects exact at LO. ⇒ Interference with process sharing the same final state at LO.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 3/16

SLIDE 9

Top pair production in POWHEG BOX

Three current implementation of top pair production in POWHEG BOX

1 hvq [arXiv:0707.3088]

⇒ NLO corrections in production. ⇒ Decay performed at LO using reweighting. ⇒ Approximate spin correlation and offshell effects.

2 t¯

tdec [arXiv:1412.1828] ⇒ NLO corrections in production and decay using NWA. ⇒ Spin correlation and offshell effects exact at LO. ⇒ Interference with process sharing the same final state at LO.

3 b¯

b4ℓ [arXiv:1607.04538] ⇒ pp → b¯ bℓ¯ νℓ¯ lνl at NLO. ⇒ Exact spin correlation and offshell effects at NLO ⇒ Interference with process sharing the same final state at NLO. ⇒ Interference of radiation in production and decay.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 3/16

SLIDE 10

Interface between POWHEG BOX RES and SMC

New resonance-aware formalism that generates emissions preserving the virtuality of the intermediate resonances. This new formalism also

ffers the opportunity to generate multiple emissions.

Production (ISR) t ¯ t

dσ = ˜ B dΦb

αISR,αb,α¯

b

∆α(kmin

⊥ ) + ∆α(kα ⊥)Rα(Φb, Φα rad)

B(Φb) dΦrad

.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 4/16

SLIDE 11

Interface between POWHEG BOX RES and SMC

New resonance-aware formalism that generates emissions preserving the virtuality of the intermediate resonances. This new formalism also

ffers the opportunity to generate multiple emissions.

Production (ISR) t ¯ t

dσ = ˜ B dΦb

αISR,αb,α¯

b

∆α(kmin

⊥ ) + ∆α(kα ⊥)Rα(Φb, Φα rad)

B(Φb) dΦrad

.

The SMC programs Pythia8 and Herwig7 veto radiation in production harder than the POWHEG one. Radiation from resonances is left, by default, unrestricted.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 4/16

SLIDE 12

Interface between POWHEG BOX RES and SMC

New resonance-aware formalism that generates emissions preserving the virtuality of the intermediate resonances. This new formalism also

ffers the opportunity to generate multiple emissions.

Production (ISR) t ¯ t

dσ = ˜ B dΦb

αISR,αb,α¯

b

∆α(kmin

⊥ ) + ∆α(kα ⊥)Rα(Φb, Φα rad)

B(Φb) dΦrad

.

The SMC programs Pythia8 and Herwig7 veto radiation in production harder than the POWHEG one. Radiation from resonances is left, by default, unrestricted. We implemented the PowhegHooksBB4L and bb4lShowerVeto classes to perform the veto also in the resonances decay.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 4/16

SLIDE 13

Our strategy

Experimental analyses based on hvq: we want to show it is obsolete and it should be replaced with b¯ b4ℓ (or with t¯ tdec for semileptonic or hadronic top decay). In order to do this, we employed a simplified version of the template method.

1 We generate samples pp → b¯

be+νeµ−¯ νµ for mt =mt,c = 172.5 GeV with the hvq, t¯ tdec and b¯ b4ℓ generators and we shower them with Pythia 8 and Herwig 7.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 5/16

SLIDE 14

Our strategy

Experimental analyses based on hvq: we want to show it is obsolete and it should be replaced with b¯ b4ℓ (or with t¯ tdec for semileptonic or hadronic top decay). In order to do this, we employed a simplified version of the template method.

1 We generate samples pp → b¯

be+νeµ−¯ νµ for mt =mt,c = 172.5 GeV with the hvq, t¯ tdec and b¯ b4ℓ generators and we shower them with Pythia 8 and Herwig 7.

2 We consider a generic observable that can be written as

O = Oc + B(mt − mt,c) + O(mt − mt,c)2. The O value we measure for the sample generated with mt,c is the Oc value associated to that given NLO+PS generator.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 5/16

SLIDE 15

Our strategy

Experimental analyses based on hvq: we want to show it is obsolete and it should be replaced with b¯ b4ℓ (or with t¯ tdec for semileptonic or hadronic top decay). In order to do this, we employed a simplified version of the template method.

1 We generate samples pp → b¯

be+νeµ−¯ νµ for mt =mt,c = 172.5 GeV with the hvq, t¯ tdec and b¯ b4ℓ generators and we shower them with Pythia 8 and Herwig 7.

2 We consider a generic observable that can be written as

O = Oc + B(mt − mt,c) + O(mt − mt,c)2. The O value we measure for the sample generated with mt,c is the Oc value associated to that given NLO+PS generator.

3 We generate samples for several mt values for hvq that we shower with

Pythia 8 in order to extract the B coefficient of a given observable. We choose the value b¯ b4ℓ+Pythia 8 as reference sample, the mass extracted using another generator is given by mt = mt,c − Oc − Oref

c

B

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 5/16

SLIDE 16

Reconstructed top mass

We take mW bj as a proxy for all top-mass sensitive observables that rely upon the mass of the decay products. ⇒ W ± = hardest ℓ± + corresponding hardest (anti-)neutrino; ⇒ B-jet: jet containing the hardest ¯ B (B) hadron; ⇒ We assume to know the b flavour in the B-jet to match it with the W.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 6/16

SLIDE 17

Reconstructed top mass

We take mW bj as a proxy for all top-mass sensitive observables that rely upon the mass of the decay products. Experimental resolution effects are simply represented as a Gaussian smearing (σ =15 GeV) ˜ f(x) = N

dy f(y) exp

−(x − y)2 2σ2

Silvia Ferrario Ravasio — Dec 20th, 2017

mt determination using new NLO+PS generators 6/16

SLIDE 18

Reconstructed top mass

We take mW bj as a proxy for all top-mass sensitive observables that rely upon the mass of the decay products. Experimental resolution effects are simply represented as a Gaussian smearing (σ =15 GeV) ˜ f(x) = N

dy f(y) exp

−(x − y)2 2σ2

We fit the smeared distribution using a skewed Lorentzian

˜ f(mW bj) = b

1 + d
mW bj − a
mW bj − a

2 + b2 + e , mmax

W bj = a +

√ 1 + d2 b2 − 1 2d

1 mmax

W bj is assigned to the bin with highest y value;

2 We set ∆ equal to the FWHM. 3 We find the values of the parameters that minimize the χ2in the

range [mmax

W bj − ∆, mmax W bj + ∆].

4 From the fitted function we extract mmax

W bj

5 If ˜

χ2 < 2 we stop; otherwise ∆ → 0.95 × ∆ and we go to step 3.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 6/16

SLIDE 19

Reconstructed top mass

We take mW bj as a proxy for all top-mass sensitive observables that rely upon the mass of the decay products. Experimental resolution effects are simply represented as a Gaussian smearing (σ =15 GeV) ˜ f(x) = N

dy f(y) exp

−(x − y)2 2σ2

We fit the smeared distribution using a skewed Lorentzian

˜ f(mW bj) = b

1 + d
mW bj − a
mW bj − a

2 + b2 + e , mmax

W bj = a +

√ 1 + d2 b2 − 1 2d We can assume B ≃ 1, thus ∆mt ≃ −∆mmax

W bj

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 6/16

SLIDE 20

Reconstructed top mass: which NLO generator?

Brief look without smearing: Large shape differences with hvq if MEC are off. With MEC, differences among the generators of the order of 10-20 MeV.

0.05 0.1 0.15 0.2 0.25 0.3 168 170 172 174 176 178

dσ/dmWbj [pb/GeV] b¯ b4ℓ mmax

Wbj = 172.805 ± 0.005 GeV

t¯ tdec mmax

Wbj = 172.818 ± 0.003 GeV

hvq mmax

Wbj = 172.741 ± 0.004 GeV

8 TeV No smearing Py8.2 no MEC mWbj [GeV]

b¯ b4ℓ t¯ tdec hvq 0.05 0.1 0.15 0.2 0.25 0.3 168 170 172 174 176 178

dσ/dmWbj [pb/GeV] b¯ b4ℓ mmax

Wbj = 172.793 ± 0.004 GeV

t¯ tdec mmax

Wbj = 172.813 ± 0.003 GeV

hvq mmax

Wbj = 172.803 ± 0.003 GeV

8 TeV No smearing Py8.2+MEC+MECaf mWbj [GeV]

b¯ b4ℓ t¯ tdec hvq Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 7/16

SLIDE 21

Reconstructed top mass: which NLO generator?

0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055 0.06 150 155 160 165 170 175 180 185 190 195

dσ/dmWbj [pb/GeV] b¯ b4ℓ mmax

Wbj = 172.662 ± 0.002 GeV

t¯ tdec mmax

Wbj = 172.882 ± 0.001 GeV

hvq mmax

Wbj = 171.654 ± 0.001 GeV

8 TeV Smearing Py8.2 noMEC mWbj [GeV]

b¯ b4ℓ t¯ tdec hvq 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055 0.06 150 155 160 165 170 175 180 185 190 195

dσ/dmWbj [pb/GeV] b¯ b4ℓ mmax

Wbj = 172.717 ± 0.002 GeV

t¯ tdec mmax

Wbj = 172.857 ± 0.001 GeV

hvq mmax

Wbj = 172.570 ± 0.001 GeV

8 TeV Smearing Py8.2+MEC+MECaf mWbj [GeV]

b¯ b4ℓ t¯ tdec hvq

Scale: envelope of 7 scale choices PDF: rwgt members of PDF4LHC15 nlo 30 pdfas (hvq only) αS: NNPDF30 nlo as115, NNPDF30 nlo as121 % − b¯ b4ℓ (µR, µF) PDF αS b¯ b4ℓ +0 MeV

+86 −53 MeV

±64 MeV

t¯ tdec +140 MeV

+6 −6 MeV

±54 MeV

hvq −147 MeV

+7 −7 MeV

±5 MeV ±9 MeV

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 8/16

SLIDE 22

Reconstructed top mass: which SMC generator? b¯ b4ℓ

0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055 0.06 150 155 160 165 170 175 180 185 190 195

dσ/dmWbj [pb/GeV] Py8.2 mmax

Wbj = 172.717 ± 0.002 GeV

Hw7.1 mmax

Wbj = 171.626 ± 0.002 GeV

8 TeV Smearing σ = 15 GeV b¯ b4ℓ mWbj [GeV]

Py8.2 Hw7.1

1 GeV displacement between Py8.2 and Hw7.1;

hvq

0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055 150 155 160 165 170 175 180 185 190 195

dσ/dmWbj [pb/GeV] Py8.2 mmax

Wbj = 172.570 ± 0.001 GeV

Hw7.1 mmax

Wbj = 172.319 ± 0.001 GeV

8 TeV Smearing hvq mWbj [GeV]

Py8.2 Hw7.1

0.25 GeV displacement between Py8.2 and Hw7.1;

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 9/16

SLIDE 23

B-jet energy peaks

Based on arXiv: 1603.03445.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 10/16

SLIDE 24

B-jet energy peaks

Based on arXiv: 1603.03445. If we do not vary mt too much, we can write Emax

bj

= Oc + B(mt − mt, c) .

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 10/16

SLIDE 25

B-jet energy peaks

Based on arXiv: 1603.03445. If we do not vary mt too much, we can write Emax

bj

= Oc + B(mt − mt, c) . We fit d σ d log Ebj 1 Ebj to a fourth order polynomial.

0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 3.5 4 4.5 5 5.5 6 6.5

dσ/d log(Ebj)/Ebj [pb/GeV]

8 TeV

Emax

bj

= 71.20 ± 0.08 GeV log(Ebj) Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 10/16

SLIDE 26

B-jet energy peaks

Based on arXiv: 1603.03445. If we do not vary mt too much, we can write Emax

bj

= Oc + B(mt − mt, c) . We fit d σ d log Ebj 1 Ebj to a fourth order polynomial.

0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 3.5 4 4.5 5 5.5 6 6.5

dσ/d log(Ebj)/Ebj [pb/GeV]

8 TeV

Emax

bj

= 71.20 ± 0.08 GeV log(Ebj)

We find B ≃ 1 2 ⇒ ∆mt ≃ −2∆Emax

bj

.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 10/16

SLIDE 27

B-jet energy peaks: which NLO generator?

Large differences between b¯ b4ℓ and hvq that does not contain radiative correction in decays and the Wt contribution. (+456 ± 103 MeV) Small differences between b¯ b4ℓ and t¯ tdec that has radiative correction in decays, implemented using NWA, and the Wt at LO. (−161 ± 102 MeV)

0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 3.5 4 4.5 5 5.5

dσ/d log(Ebj)/Ebj [pb/GeV] b¯ b4ℓ Emax

bj

= 71.200 ± 0.081 GeV t¯ tdec Emax

bj

= 71.361 ± 0.062 GeV hvq Emax

bj

= 70.744 ± 0.064 GeV 8 TeV log(Ebj)

b¯ b4ℓ+Py8.2 t¯ tdec+Py8.2 hvq+Py8.2

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 11/16

SLIDE 28

B-jet energy peaks: which SMC generator? b¯ b4ℓ

0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 3.5 4 4.5 5 5.5 6

dσ/d log(Ebj)/Ebj [pb/GeV] Py8.2 Emax

bj

= 71.200 ± 0.081 GeV Hw7.1 Emax

bj

= 69.050 ± 0.081 GeV 8 TeV b¯ b4ℓ log(Ebj)

Py8.2 Hw7.1

2 GeV displacement between Py8.2 and Hw7.1; ∆mt ≃ −4 GeV

hvq

0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 3.5 4 4.5 5 5.5 6

dσ/d log(Ebj)/Ebj [pb/GeV] Py8.2 Emax

bj

= 70.744 ± 0.064 GeV Hw7.1 Emax

bj

= 69.716 ± 0.062 GeV 8 TeV hvq log(Ebj)

Py8.2 Hw7.1

1 GeV displacement between Py8.2 and Hw7.1; ∆mt ≃ −2 GeV

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 12/16

SLIDE 29

Leptonic observables

Based on arXiv:1407.2763.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 13/16

SLIDE 30

Leptonic observables

Based on arXiv:1407.2763. Measure O for Oi =

pj

⊥(ℓ+), pj ⊥(ℓ+ℓ−), mj(ℓ+ℓ−), (E(ℓ+) + E(ℓ−))j, (p⊥(ℓ+) + p⊥(ℓ−))j

with j = 1, 2, 3.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 13/16

SLIDE 31

Leptonic observables

Based on arXiv:1407.2763. Measure O for Oi =

pj

⊥(ℓ+), pj ⊥(ℓ+ℓ−), mj(ℓ+ℓ−), (E(ℓ+) + E(ℓ−))j, (p⊥(ℓ+) + p⊥(ℓ−))j

with j = 1, 2, 3. Assume Oi = Oc,i + Bi

mj

t − mj t, c

, thus the extracted mass

corresponding to the observable i is given by mt,i =

mj

t, c − Oc,i − Oref c,i

Bi 1/j .

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 13/16

SLIDE 32

Leptonic observables

Based on arXiv:1407.2763. Measure O for Oi =

pj

⊥(ℓ+), pj ⊥(ℓ+ℓ−), mj(ℓ+ℓ−), (E(ℓ+) + E(ℓ−))j, (p⊥(ℓ+) + p⊥(ℓ−))j

with j = 1, 2, 3. Assume Oi = Oc,i + Bi

mj

t − mj t, c

, thus the extracted mass

corresponding to the observable i is given by mt,i =

mj

t, c − Oc,i − Oref c,i

Bi 1/j . Obtain Oc,i and its uncertainty due to PDF and scale variations. Combine all the errors in quadrature and mt,i and ∆mt,i.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 13/16

SLIDE 33

Leptonic observables

Based on arXiv:1407.2763. Measure O for Oi =

pj

⊥(ℓ+), pj ⊥(ℓ+ℓ−), mj(ℓ+ℓ−), (E(ℓ+) + E(ℓ−))j, (p⊥(ℓ+) + p⊥(ℓ−))j

with j = 1, 2, 3. Assume Oi = Oc,i + Bi

mj

t − mj t, c

, thus the extracted mass

corresponding to the observable i is given by mt,i =

mj

t, c − Oc,i − Oref c,i

Bi 1/j . Obtain Oc,i and its uncertainty due to PDF and scale variations. Combine all the errors in quadrature and mt,i and ∆mt,i. Average all the measurements using as covariance matrix Vik = ∆mt,i

2δik + (1 − δik) min

∆mt,i

2, ∆mt,k 2, ρik∆mt,i∆mt,k

where ρik is the statistical correlation between Oi and Ok.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 13/16

SLIDE 34

Leptonic observables

168 170 172 174 176 178 180 182 184

pT(ℓ+) pT(ℓ+ℓ−) m(ℓ+ℓ−) E(ℓ+ℓ−) pT(ℓ+)+pT(ℓ−) b¯ b4ℓ, mt = 172.5 GeV Py8.2: mt = 172.500+0.784

−0.766 GeV

Hw7.1: mt = 175.392+1.045

−1.138 GeV

Extracted mt [GeV]

1st Mellin moment 2nd Mellin moment 3rd Mellin moment 168 170 172 174 176 178 180 182 184

pT(ℓ+) pT(ℓ+ℓ−) m(ℓ+ℓ−) E(ℓ+ℓ−) pT(ℓ+)+pT(ℓ−) t¯ tdec mt = 172.5 GeV Py8.2: mt = 171.760+0.751

−0.752 GeV

Hw7.1: mt = 175.473+0.962

−1.104 GeV

Extracted mt [GeV]

1st Mellin moment 2nd Mellin moment 3rd Mellin moment 168 170 172 174 176 178 180 182 184

pT(ℓ+) pT(ℓ+ℓ−) m(ℓ+ℓ−) E(ℓ+ℓ−) pT(ℓ+)+pT(ℓ−) hvq, mt = 172.5 GeV Py8.2: mt = 172.238+0.754

−0.748 GeV

Hw7.1: mt = 174.607+0.961

−1.097 GeV

Extracted mt [GeV]

1st Mellin moment 2nd Mellin moment 3rd Mellin moment

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 14/16

SLIDE 35

Summary and Outlooks

Which Observable?

smeared mW bj: oversimplification; small sensitivity to the production mechanism (small pdf/scale variations); Ebj: small sensitivity to the production mechanism, large shower uncertainties. leptonic observables: sensitivity to the production mechanism, large shower uncertainties.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 15/16

SLIDE 36

Summary and Outlooks

Which Observable?

smeared mW bj: oversimplification; small sensitivity to the production mechanism (small pdf/scale variations); Ebj: small sensitivity to the production mechanism, large shower uncertainties. leptonic observables: sensitivity to the production mechanism, large shower uncertainties.

Which NLO generator (using Pythia8+MEC)?

b¯ b4ℓ generator is the most accurate one and should be preferred if possible. smeared mW bj: hvq and t¯ tdec lead to a systematic uncertainty of roughly 150 MeV. Emax

bj

: t¯ tdec ∆mt ≃ 0.3 ± 0.2 GeV, hvq ∆mt ≃ 0.9 ± 0.2 GeV. leptonic observables: t¯ tdec mt 700 MeV smaller than the nominal value, hvq not accurate for observables depending on spin correlations although better average (mt = 172.2 GeV).

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 15/16

SLIDE 37

Summary and Outlooks

Pythia8 or Herwig7?

hvq must be showered with both showers, the difference leads to a systematic uncertainty of 250 MeV when using mW bj, 2 GeV when using Ebj/leptonic observables. when using b¯ b4ℓ (or t¯ tdec), the difference between Pythia8 and Herwig7 is greater than 1 GeV even for mW bj, 4 GeV Ebj, 3 GeV leptonic observables. matching procedure in Herwig7 introduces new systematic errors and requires further investigation.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 16/16

SLIDE 38

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 17/16

SLIDE 39

“Measurement” of the top-quark mass

Many ways to infer mt, the most precise is the template method

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 18/16

SLIDE 40

“Measurement” of the top-quark mass

Many ways to infer mt, the most precise is the template method

1 Top momentum reconstruction from its decay products.

⇒ B-jet; ⇒ W decay products: → charged lepton + neutrino → two light jets

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 18/16

SLIDE 41

“Measurement” of the top-quark mass

Many ways to infer mt, the most precise is the template method

1 Top momentum reconstruction from its decay products. 2 Given a MC event generator, produce several templates varying

the input mass mt.

160 165 170 175 180 185 190 mW

bj [GeV]

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 d /dmW

bj [pb/GeV]

Reconstructed top mass

mt=169.5 GeV mt=171.0 GeV mt=172.5 GeV mt=174.0 GeV mt=175.5 GeV

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 18/16

SLIDE 42

“Measurement” of the top-quark mass

Many ways to infer mt, the most precise is the template method

1 Top momentum reconstruction from its decay products. 2 Given a MC event generator, produce several templates varying

the input mass mt.

3 Extract the parametric dependence on the input mass mt.

160 170 180 190 200 mWbj [GeV] 0.035 0.040 0.045 0.050 0.055 0.060 dσ/dmWbj [pb/GeV]

Reconstructed top mass fit

mWbj = 172.789 ± 0.001 GeV

170 171 172 173 174 175 m

t input [GeV]

170 171 172 173 174 175 176 m

W−b j f tted peak [GeV]

Reconstructed mass dependence on m

t

Theoretical error barr: scales, PDF and coupling dependence

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 18/16

SLIDE 43

“Measurement” of the top-quark mass

Many ways to infer mt, the most precise is the template method

1 Top momentum reconstruction from its decay products. 2 Given a MC event generator, produce several templates varying

the input mass mt.

3 Extract the parametric dependence on the input mass mt. 4 The mt value that fits the data the best is the extracted mass. 170 171 172 173 174 175 mt input [GeV] 170 171 172 173 174 175 176 mW−bj fitted peak [GeV] 172.65

Example of mt extraction

exp data mt = 172.498+0.153

−0.235 (th) ± 0.003 (stat) GeV

MC generator

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 18/16

SLIDE 44

“Measurement” of the top-quark mass

Many ways to infer mt, the most precise is the template method

1 Top momentum reconstruction from its decay products. 2 Given a MC event generator, produce several templates varying

the input mass mt.

3 Extract the parametric dependence on the input mass mt. 4 The mt value that fits the data the best is the extracted mass. 5 mt can depend on the MC used

⇒ if A is more accurate than B, use A; ⇒ otherwise |mA

t − mB t |

contributes to the sys- tematic uncertanty;

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 18/16

SLIDE 45

Phenomenological setup

Process: pp → b¯ b e+ µ− νe ¯ νµ, dominated by top pair production plus leptonic decay, at √s = 8 TeV. Central PDF: MSTW2008. Dynamic scale choice t¯ t events: µ =

(E2

t − p2 z,t)(E2 ¯ t − p2 z,¯ t)

1/4 Z¯ bb events: µ = p2

Z

2

Scale variations (KF, KR) = (1, 1) , (2, 2) , 1

2, 1 2

, (1, 2) ,
1, 1

2

, (2, 1) ,

1

2, 1

PDF

Rwgt using several sets: PDF4LHC15, NNPDF3.0, CT14nlo, MMHT2014. Rwgt 30 pdf inside the set PDF4LHC15 nlo 30 pdfas, Gaussian symmetric error, for hvq only.

αS: Use NNPDF30 nlo as 0115 and NNPDF30 nlo as 0121; half difference.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 19/16

SLIDE 46

Physics objetcs

B hadrons are considered as stable. Jets reconstructed using anti-k⊥ algorithm for R = 0.5. Impose selection cuts to suppress the Wt background: ⇒ 2 opposite charged leptons with: p⊥(ℓ) > 20 GeV, |η(ℓ)| < 2.4 ⇒ 2 B-jet with opposite b flavour with: p⊥(jB) > 30 GeV, |η(jB)| < 2.5 We assume to know neutrinos momentum. W + reconstructed merging the hardest ℓ+ and the hardest neutrino; W − reconstructed merging the hardest ℓ− and the hardest anti-neutrino. Reconstructed t: W + and jet containing the hardest b-flavoured hadron; reconstructed ¯ t: W − and jet containing the hardest ¯ b-flavoured hadron.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 20/16

SLIDE 47

mWbj: backup material

mW bj extracted peak (with smearing): difference between Pythia8 and Herwig7 for different jet radius values.

0.2 0.4 0.6 0.8 1 1.2 1.4 0.4 0.45 0.5 0.55 0.6

∆mmax

Wbj [GeV]

8 TeV Py8.2 − Hw7.1 R b¯ b4ℓ t¯ tdec hvq

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 21/16

SLIDE 48

Ebj: R dependence

Emax

bj

: difference with b¯ b4ℓ (left) for all generators showered with Pythia8, and difference between Pythia8 and Herwig7 (right) for all generators for several values of the jet radius.

−0.8 −0.6 −0.4 −0.2 0.2 0.4 0.6 0.8 0.4 0.45 0.5 0.55 0.6

∆Emax

bj

[GeV] 8 TeV % − b¯ b4ℓ R

b¯ b4ℓ+Py8.2 t¯ tdec+Py8.2 hvq+Py8.2 0.5 1 1.5 2 2.5 3 0.4 0.45 0.5 0.55 0.6

∆Emax

bj

[GeV] 8 TeV Py8.2 − Hw7.1 R b¯ b4ℓ t¯ tdec hvq

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 22/16

SLIDE 49

Ebj: scale, PDF and αS dependence

Ebj: independent from the production mechanism, indeed small dependence on scale/PDF. % − b¯ b4ℓ (µR, µF) PDF αS stat b¯ b4ℓ +0 MeV

+22 −15 MeV

±35 MeV

±81 MeV t¯ tdec +161 MeV

+22 −24 MeV

±17 MeV

±62 MeV hvq −456 MeV

+32 −47 MeV

±30 MeV ±25 MeV ±64 MeV

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 23/16

SLIDE 50

Leptonic observables

First Mellin moment for mt = mt, c for all generators showered with

Pythia8. The angular coefficients have been obtained by considering

three mt values: 169.5, 172.5, 175.5 GeV.

Obs gen B Oc % − b¯ b4ℓ (µF, µR) PDF αS [GeV] [MeV] [MeV] [MeV] [MeV] b¯ b4ℓ 0.17 ± 0.04 56.653 ± 0.050

+79

−86

±26 (±92)

pT(ℓ+) t¯ tdec 0.19 ± 0.02 56.804 ± 0.033 +151 ± 60

+84 −86

±41 (±23)

hvq 0.19 ± 0.02 56.738 ± 0.032 +85 ± 59

+82 −86

±130 ±49 (±23) b¯ b4ℓ 0.30 ± 0.05 69.759 ± 0.059

+710

−444

±85 (±110)

pT(ℓ+ℓ−) t¯ tdec 0.30 ± 0.02 69.660 ± 0.040 −100 ± 71

+538 −361

±78 (±28)

hvq 0.29 ± 0.02 69.201 ± 0.038 −558 ± 71

+553 −367

±95 ±95 (±27) b¯ b4ℓ 0.31 ± 0.08 108.685 ± 0.099

+234

−341

±57 (±191)

m(ℓ+ℓ−) t¯ tdec 0.31 ± 0.03 108.812 ± 0.065 +127 ± 119

+244 −259

±33 (±46)

hvq 0.33 ± 0.03 109.200 ± 0.064 +515 ± 118

+247 −265

±395 ±68 (±45) b¯ b4ℓ 0.55 ± 0.14 186.803 ± 0.163

+342

−385

±540 (±305)

E(ℓ+ℓ−) t¯ tdec 0.56 ± 0.05 187.005 ± 0.107 +201 ± 195

+448 −434

±474 (±76)

hvq 0.56 ± 0.05 186.809 ± 0.105 +6 ± 194

+441 −427

±1068 ±559 (±74) b¯ b4ℓ 0.38 ± 0.08 113.322 ± 0.095

+165

−184

±93 (±178)

pT(ℓ+) + pT(ℓ−) t¯ tdec 0.39 ± 0.03 113.598 ± 0.063 +276 ± 114

+165 −174

±72 (±44)

hvq 0.39 ± 0.03 113.425 ± 0.062 +104 ± 113

+165 −174

±259 ±101 (±43)

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 24/16

SLIDE 51

Radiation scale in POWHEG BOX

In the POWHEG formalism, the emission probability at a scale µ is given by the Sudakow form factor ∆(µ) = exp

−
dφradθ
k⊥(µ′) − k⊥(µ)

αS(µ′) 2π Rs (k⊥(µ′)) B

,

where k⊥(µ) is the transverse momentum of the emitted particle corresponding to the scale µ. In the Fortran code POWHEG BOX µ = k⊥ and there is no way to change the definition of the scale of the emission. Since αS(µ) = αS (µ; αS(mZ)) =

αS(mZ) 1+β0αS(mZ) log

µ2

m2 Z

instead of

changing µ, is possible to change the reference value of αS(mZ). For an average k⊥=30 GeV, we get: αS(k⊥; 0.118) = 0.1402 αS(2k⊥; 0.118) = 0.1253 αS(0.5k⊥; 0.118) = 0.1590 αS(k⊥; 0.115) = 0.1360 αS(k⊥; 0.121) = 0.1444 αS variations should be enhanced by a factor 4 to get the corresponding uncertainty on the scale of the emission.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 25/16

SLIDE 52

Interface between POWHEG BOX and SMC

The radiation provided by the SMC with transverse momentum larger than scalup = kPOWHEG

⊥

must be vetoed: vetoed showers.

p⊥

ISR: (kPOWHEG

⊥

)2 = p2

⊥

p2 p1

FSR: (kPOWHEG

⊥

)2 = 2p1 · p2

E2 E1

It is desiderable that the SMC employs the POWHEG BOX definition of k⊥ to perform the veto.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 26/16

SLIDE 53

Interface between POWHEG BOX and SMC

The radiation provided by the SMC with transverse momentum larger than scalup = kPOWHEG

⊥

must be vetoed: vetoed showers.

p⊥

ISR: (kPOWHEG

⊥

)2 = p2

⊥

p2 p1

FSR: (kPOWHEG

⊥

)2 = 2p1 · p2

E2 E1

It is desiderable that the SMC employs the POWHEG BOX definition of k⊥ to perform the veto. Problems have been observed e.g. in dijet production, a solution was proposed in Ref. [arXiV:1303.3922]. For FSR, in case of massless emitter, scalup is computed by new with the definition

kPOWHEG

⊥

2 = d12 = 2p1 · p2 E1 E2 (E1 + E2)2

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 26/16

SLIDE 54

Interface between POWHEG BOX and SMC

The radiation provided by the SMC with transverse momentum larger than scalup = kPOWHEG

⊥

must be vetoed: vetoed showers.

p⊥

ISR: (kPOWHEG

⊥

)2 = p2

⊥

p2 p1

FSR: (kPOWHEG

⊥

)2 = 2p1 · p2

E2 E1

It is desiderable that the SMC employs the POWHEG BOX definition of k⊥ to perform the veto. Problems have been observed e.g. in dijet production, a solution was proposed in Ref. [arXiV:1303.3922]. For FSR, in case of massless emitter, scalup is computed by new with the definition

kPOWHEG

⊥

2 = d12 = 2p1 · p2 E1 E2 (E1 + E2)2 In Pythia8, it is possible to veto using this “improved” definition: PowhegHooks.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 26/16

SLIDE 55

Interface between POWHEG BOX and SMC

Pythia8 is a k⊥-ordered shower and the hadronization model employed is the Lund string fragmentation one.

Hardest emission Vetoed shower

⇒ Natural matching with POWHEG radiation.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 27/16

SLIDE 56

Interface between POWHEG BOX and SMC

Pythia8 is a k⊥-ordered shower and the hadronization model employed is the Lund string fragmentation one.

Hardest emission Vetoed shower

⇒ Natural matching with POWHEG radiation. Herwig7 is an angular-ordered shower and it employs the cluster model.

Vetoed-Truncated Shower Hardest Emission Vetoed Shower tI t0 pT, t

⇒ Truncated-vetoed showers often give rise to little contribution; so only a vetoed shower is implemented.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 27/16

SLIDE 57

POWHEG BOX RES

Technical problems of processes containing resonances whose decay products can radiate:

1

NLO computation: we need a subtraction scheme that constructs the counterterms to real diagrams preserving the virtuality of the resonances, in

rder not to spoil the cancellation of the infra-red poles. This simply results

in poor convergence.

2

Hardest emission generation (more severe): in POWHEG formalism, the emission probability is described by R/B. If R contains an onshell resonance, while B does not, the ratio R/B is large, also for high transverse momentum radiation. Moreover it does not approach the Altarelli-Parisi splitting function in the infrared limit, as it is required by the POWHEG method, giving rise to unphysical distortions of the distributions.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 28/16

SLIDE 58

POWHEG BOX RES

Technical problems of processes containing resonances whose decay products can radiate:

1

NLO computation: we need a subtraction scheme that constructs the counterterms to real diagrams preserving the virtuality of the resonances, in

rder not to spoil the cancellation of the infra-red poles. This simply results

in poor convergence.

2

Hardest emission generation (more severe): in POWHEG formalism, the emission probability is described by R/B. If R contains an onshell resonance, while B does not, the ratio R/B is large, also for high transverse momentum radiation. Moreover it does not approach the Altarelli-Parisi splitting function in the infrared limit, as it is required by the POWHEG method, giving rise to unphysical distortions of the distributions. If we can separate the resonances in different singular regions (e.g. pp → t¯ t), we can write dσ = ˜ BdΦb

αISR,αb,α¯

b

∆α(kmin

⊥

) + ∆α(kα

⊥) Rα(Φb, Φα rad)

B(Φb) dΦrad

.

The multi-emission formalism is crucial for process where ISR is much more likely: in this way the first emission is generated by POWHEG BOX RES instead of the PS.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 28/16

SLIDE 59

POWHEG BOX RES and SMC: general algorithm

When a LH event is read we get

1 Production process (ISR): Read scalup from the file. For

remnant we set scalup = √ ˆ s/2.

2 t (ot ¯

t) resonance: If an emission is present, µ2

t = 2pb · pg Eg

Eb in the top frame. Otherwise µ2

t = 0.8 GeV2.*

3 Check that the PS generates emissions off the top decay products

with a k⊥ smaller than µt.

*For hvq and remnant events in b¯

b4ℓ emissions in decay are not generated, thus no veto is performed.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 29/16

SLIDE 60

POWHEG BOX RES and PYHTIA 8 and HERWIG 7

We implemented subroutines to veto radiation in the t resonance: PYTHIA 8: It is possible to use PowhegHooks to veto radiation in

production. We implemented PowhegHooksBB4L for emissions in

decay:

1 FSREmissionVeto (default):

After each emission, we decide if keeping or rejecting it. It employs the POWHEG BOX definition of k⊥.

2 ScaleResonance:

µt is used as starting scale for the shower off the t (¯ t) resonance. The shower scale is the PYHTIA transverse momentum.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 30/16

SLIDE 61

POWHEG BOX RES and PYHTIA 8 and HERWIG 7

We implemented subroutines to veto radiation in the t resonance: PYTHIA 8: It is possible to use PowhegHooks to veto radiation in

production. We implemented PowhegHooksBB4L for emissions in

decay:

1 FSREmissionVeto (default):

After each emission, we decide if keeping or rejecting it. It employs the POWHEG BOX definition of k⊥.

2 ScaleResonance:

µt is used as starting scale for the shower off the t (¯ t) resonance. The shower scale is the PYHTIA transverse momentum.

HERWIG 7: we implemented two alternatives

1 bb4lShowerVeto (default):

After each emission, we decide if keeping or rejecting it. Herwig7 provides us the k⊥ and the momenta of the emitted particles are not known yet.

2 bb4lFullShowerVeto:

before the hadronization phase, we look at the emissions

riginated from the t decay chain, if every emission is softer than

the POWHEG one the event is accepted, otherwise it is reshowered. k⊥ is computed using the “improved” POWHEG BOX definition. Partons have been reshuffled and the k⊥ computed contains ambiguity due to this procedure.

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 30/16

SLIDE 62

Matching procedures

We now compare the results obtained with b¯ b4ℓ+Pythia8 using the different matching procedures. Results are expressed in GeV.

Observable FSREmission FSR+PowhegHooks ScaleResonance mmax

W bj

172.793 ± 0.004 172.828 ± 0.005 172.816 ± 0.004 mmax

W bj (smear)

172.717 ± 0.002 172.794 ± 0.002 172.737 ± 0.002 Emax

bj

71.200 ± 0.081 71.204 ± 0.082 71.179 ± 0.082

We now compare the results obtained with b¯ b4ℓ+Herwig7 using the different matching procedures. Results are expressed in GeV. Observable bb4lShowerVeto bb4lFullShowerVeto mmax

W bj

172.727 ± 0.005 172.776 ± 0.005 mmax

W bj (smear)

171.626 ± 0.002 171.829 ± 0.002 Emax

bj

69.050 ± 0.081 69.190 ± 0.082

Silvia Ferrario Ravasio — Dec 20th, 2017 mt determination using new NLO+PS generators 31/16