NLO+PS off-shell effects in precise
determinations of the mass and width
of the top-quark at the LHC
Eps-Hep 2023
Aug 22nd, 2023
Simone Amoroso, D. Britzger,
S. Kluth, L. Scyboz
1
Top quark mass at hadron colliders
Most accurate top quark mass determinations through direct measurements,
from the reconstruction of the top decay products (lepton, jets, b-jets)
Fits to data using templates obtained from exclusive Monte Carlo predictions
ATLAS+CMS Preliminary
LHCtop WG
World comb. (Mar 2014) [2]
stat
total uncertainty
LHC comb. (Sep 2013)
World comb. (Mar 2014)
LHCtop WG
ATLAS, l+jets
ATLAS, dilepton
ATLAS, all jets
ATLAS, single top
ATLAS, dilepton
ATLAS, all jets
ATLAS, l+jets
Oct 2022
total stat
mtop ± total (stat ± syst)
173.29 ± 0.95 (0.35 ± 0.88)
173.34 ± 0.76 (0.36 ± 0.67)
172.33 ± 1.27 (0.75 ± 1.02)
s
Ref.
7 TeV [1]
1.96-7 TeV [2]
7 TeV [3]
173.79 ± 1.41 (0.54 ± 1.30)
175.1± 1.8 (1.4 ± 1.2)
7 TeV [3]
172.2 ± 2.1 (0.7 ± 2.0)
172.99 ± 0.85 (0.41± 0.74)
173.72 ± 1.15 (0.55 ± 1.01)
172.08 ± 0.91 (0.39 ± 0.82)
172.69 ± 0.48 (0.25 ± 0.41)
8 TeV [5]
ATLAS, leptonic invariant mass
ATLAS, dilepton (*)
CMS, l+jets
CMS, dilepton
CMS, all jets
CMS, l+jets
CMS, dilepton
CMS, all jets
CMS, single top
174.41± 0.81 (0.39 ± 0.66 ± 0.25)
172.63 ± 0.79 (0.20 ± 0.67 ± 0.37)
173.49 ± 1.06 (0.43 ± 0.97)
172.50 ± 1.52 (0.43 ± 1.46)
13 TeV [9]
173.49 ± 1.41 (0.69 ± 1.23)
172.35 ± 0.51 (0.16 ± 0.48)
172.82 ± 1.23 (0.19 ± 1.22)
7 TeV [13]
172.32 ± 0.64 (0.25 ± 0.59)
172.95 ± 1.22 (0.77 ± 0.95)
8 TeV [14]
CMS comb. (Sep 2015)
172.44 ± 0.48 (0.13 ± 0.47)
172.25 ± 0.63 (0.08 ± 0.62)
7+8 TeV [14]
172.33 ±
172.34 ±
172.13 ±
171.77 ±
172.76 ±
13 TeV [17]
ATLAS comb. (Oct 2018)
CMS, l+jets
CMS, dilepton
CMS, all jets
CMS, single top
CMS, l+jets (*)
CMS, boosted (*)
165
0.70 (0.14 ±
0.73 (0.20 ±
0.77 (0.32 ±
0.38
0.81 (0.22 ±
[1] ATLAS-CONF-2013-102
[2] arXiv:1403.4427
[3] EPJC 75 (2015) 330
[4] EPJC 75 (2015) 158
[5] ATLAS-CONF-2014-055
[6] PLB 761 (2016) 350
[7] JHEP 09 (2017) 118
* Preliminary
ATLAS-CONF-2022-058
mtop summary, s = 7-13 TeV
170
175
mtop [GeV]
0.69)
0.70)
0.70)
8 TeV [6]
8 TeV [7]
8 TeV [8]
7+8 TeV [8]
13 TeV [10]
7 TeV [11]
7 TeV [12]
8 TeV [14]
8 TeV [14]
8 TeV [15]
13 TeV [16]
13 TeV [18]
13 TeV [19]
13 TeV [20]
0.78)
[8] EPJC 79 (2019) 290
[9] arXiv:2209.00583
[10] ATLAS-CONF-2022-058
[11] JHEP 12 (2012) 105
[12] EPJC 72 (2012) 2202
[13] EPJC 74 (2014) 2758
[14] PRD 93 (2016) 072004
180
7 TeV [4]
13 TeV [21]
[15] EPJC 77 (2017) 354
[16] EPJC 78 (2018) 891
[17] EPJC 79 (2019) 368
[18] EPJC 79 (2019) 313
[19] arXiv:2108.10407
[20] CMS-PAS-TOP-20-008
[21] CMS-PAS-TOP-21-012
185
TOPLHCWG
2
The Powheg bb4l event generator
All recent direct measurements of mtop use NLO+PS Monte Carlo samples
• NLO in production, LO spin correlations and NLO decay through shower ME corrections
S. Frixione, P. Nason, G. Ridolfi JHEP 0709 (2007) 126
double-resonant
single-resonant
non-resonant
State of the art since a few years the bb4l code in Powheg-Box-RES
• NLO computation of the full 6-fermion process
• non-resonant, off-shell , interference & spin-correlation effects at exact NLO QCD
T. Jezo, P. Nason, JHEP 12 (2015) 065
• Unified treatment of ttbar and Wt single-top
T. Jezo et al, Eur.Phys.J.C 76 (2016) 12, 691
T. Jezo, J.M. Lindert, S. Pozzorini, [2307.15653] [hep-ph]
3
Off-shell effects in top production
Compare bb4l with hvq
including NLO decays of
the tops in the NWA
Distortions in the mWb
lineshape at particle level
Do they impact direct
determinations of mtop?
T. Jezo, J.M. Lindert, S. Pozzorini, [2307.15653] [hep-ph]
4
Off-shell effects in top production
Particle-level study including
smearing to emulate detector effects
Finds effects of the order of O(200 MeV)
on the peak position of the
reconstructed top mass
Some dependence on the specific
observable and parton shower code
Can we learn anything using
existing measurements?
S. Ferrario Ravasio, T. Jezo, P. Nason, C.Oleari Eur.Phys.J.C 78 (2018) 6
5
ATLAS measurement of mlbminimax
ATLAS √s = 13TeV measurement
• 36.1 fb-1 of Run-II data
• Dileptonic channel (ee, eµ, µµ)
• Combine one lepton and one b-jet
using mlbminimax variable
Kinematic endpoint at
2
mtop
−
2
mW
sensitive to the top quark mass
Above endpoint sensitive to tt/Wt
interference and top-quark width effects
ATLAS Collaboration, Phys.Rev.Lett. 121 (2018) 15
6
ATLAS measurement of mlbminimax
ATLAS √s = 13TeV measurement
• 36.1 fb-1 of Run-II data
• Dileptonic channel (ee, eµ, µµ)
• Combine one lepton and one b-jet
using mlbminimax variable
C. Herwig, T. Jezo, B. Nachman, Phys.Rev.Lett. 122 (2019) 23
Kinematic endpoint at
2
mtop
−
2
mW
sensitive to the top quark mass
Above endpoint sensitive to tt/Wt
interference and top-quark width effects
ATLAS measurement used for a
competitive direct extraction of Γtop
Can we use it to also learn
something on mtop?
7
2 ⋅107 bb4l LHE events generated
mtop = 172.5 GeV, Γtop = 1.33 GeV
hdamp = mtop
NNPDF31_nnlo_hessian PDF
!"
10−2
ATLAS Data, 36.1 fb−1
nominal bb4l
Γtop ± 0.66 GeV
mtop ± 2.5 GeV
!"!
!"
!!"
!"
10−3
"!!
Same-flavor lepton decays
included by relabeling LHE events
(see L. Jeppe poster)
Interfaced to Pythia8.307 using
dedicated bb4l UserHook
Templates generated for
• mtop = 170.0, 171.5, 173.5, 175.0 GeV
• Γtop = 0.66, 1.00, 1.66, 2.00 GeV
!
!"
"!
!
!"
!
!"
10−4
!
!"
!
!
!"
!"
"!!
!
!
!"
!"
10−5
MC/Data
•
•
•
•
1
dσ
[1/GeV]
minimax
σ dmbℓ
bb4l predictions
1.4
1.3
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5
!
!
!"
!
!"
!"!
!"
!"
!
!"
!
!
!
!"
!"
0
!"
!
!
!
!"
!"
!
Δmtop
100
150
200
ΔΓtop
!"
250
!
!"
!"
50
!
300
350
mminimax
bℓ
!"
400
[GeV]
8
The “linear template fit”
Goodness of fit for binned distributions with gaussian uncertainties
T −1
2
χ (α) = (d − λ(α)) V (d − λ(α))
λ: theory-prediction
α parameter(s) of interest,
(top mass and/or width)
d the measurement,
V covariance matrix (W=V-1)
D. Britzger, Eur.Phys.J.C 82 (2022) 8, 731
The Linear Template Fit
The parameter dependence λ(α) approximated linearly from 'templates'
→ Valid approximation when templates are close to the best estimator
The fit result ('best estimator') is directly obtained in closed analytic form:
T
−1
T
α̂ = [(YM̃) WYM̃] [(YM̃) W(d − Ym)]
Y: matrix of the templates
M, m: quantities computed from the reference values of the templates
9
Sensitivity
Templates generated for
• mtop {170, 171.5, 172.5, 173.5, 175} GeV
• Γtop {0.66, 1.0, 1.33, 1.66, 2.0 } GeV
Good linearity observed in all bins
• No difference between linear
and quadratic interpolation
Low mlb mostly sensitive to mtop
High mlb mostly sensitive to Γtop
10
Systematic uncertainties
Experimental uncertainties all
encoded in data covariance matrix
Dominated by uncertainties in the
theoretical modeling of tt+Wt
and in the description of
additional b-tagged jets
Could be significantly reduced in the
future by directly using bb4l
predictions in the measurement
11
Systematic uncertainties
• 7-point scale variations and
NNPDF31 eigenvector variations
obtained through bb4l reweighting
dσ
1
[1/GeV]
minimax
σ dmbℓ
For the interpretation, need to consider theoretical uncertainties on bb4l prediction:
• 3-point hdamp variations (0.5, 2.0 mtop)
!"
10−2
nominal bb4l
7-point µ R , µ F
PDF variations
3-point hdamp
3-point shower
ATLAS Data, 36.1 fb−1
!"!
!"
"!!
!"
10−3
"!!
!"
!
!"
10−4
!
!"
!
!"
!"
• 3-point shower scales and
• NB: Work ongoing to evaluate
!
!
!"
!
"!!
!
!
1.4
1.3
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5
!"
!"
10−5
MC/Data
non-singular terms variations
in the splitting kernels
through Pythia8 reweighting
!"
!
!
!"
!"
!
!"!
!"
!"
!
!"
!
!
!
!"
!"
0
!"
!
!
!
!"
!"
!
100
150
200
uncertainties on the matching to Pythia (negligible)
and possible bias due to wrong shower recoils (few ~100 MeV on mtop)
!"
250
!
!"
!"
50
!
300
350
mminimax
bℓ
!"
400
[GeV]
12
GoF dependence on the fit range
χ²/ndf for different mbl fit ranges before and after the fit
Very good agreement of the post-fit predictions with data in the entire mbl range
13
Fit performed between 40 < mlb < 380 GeV
Results
1
dσ
[1/GeV]
σ dmminimax
bℓ
Simultaneous determination of mass and width
#$!"
10−2
nominal bb4l
post-fit bb4l
ATLAS Data, 36.1 fb−1
#$!"!
#$!"
!#$!"
#$!"
10−3
!"
#$
!
!#$!"
!#$"
!#$
!"
!#$
!"
10−4
!
#$
!"
!
!#$
!"
!"#$
#$!
!"
#$!
$#!
Weak correlation (-30%) between mtop and Γtop
Extracted width consistent with
previous determination
MC/Data
Excellent agreement of data and bb4l post-fit
10−5
1.4
1.3
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5
!"
!"
!
!
#$!"
!
#$!"
#$!"!
!"#$
#$!"
#$
!
!"
!
!"
!
#$!
!"
#$!"
0
#$
#$
!
#$
#$!
!
!"
#$
!"
!
!"
!"
50
100
150
200
!"
250
300
#$!
!
#$
350
mminimax
bℓ
!"
400
[GeV]
Theoretical uncertainty dominated
by hdamp and shower variations
14
Simultaneous determination of mass and width
Fit performed between 40 < mlb < 380 GeV
Results
Excellent agreement of data and bb4l post-fit
Weak correlation (-30%) between mtop and Γtop
Extracted width consistent with
previous determination
Theoretical uncertainty dominated
by hdamp and shower variations
15
Top quark mass determination
mtop determination including
the top-mass dependence in Γtop
LO Γtop relation
Result
Moderate reduction of exp. uncertainties
(limited precision of data at large mlb)
Reduced theoretical uncertainties
Competitive with dedicated measurements
16
Summary
Simultaneous determination of mtop and Γtop from a single measurement
• First mtop with full NLO treatment of non-resonant and interference effects
exploiting the bb4l Monte Carlo event generator
•
• Multivariate Linear Template Fit to study sensitivity to mtop and Γtop
Using ATLAS measurement of mlbminimax at √s=13 TeV
Already competitive uncertainties to dedicated measurements
Results consistent with world average and previous determinations
Top-quark mass effects in Γtop provides small additional sensitivity to mtop
with dominant model uncertainties found to be reduced
Promising avenue for future LHC measurements
17
Backup
18