MC for t H and SM backgrounds - Durham University
Transcript of MC for t H and SM backgrounds - Durham University
MC for ttH and SM backgrounds
Stefano Pozzorini
Zurich University
Higgs Couplings 2015Lumley Castle, 13 October 2015
Large backgrounds in ttH searches
Dominant ttH(bb) channel suffers from heavy tt+jets contamination
∼ 3′000 ttH events but only 3–4×σSM exclusion at Run1
main bottleneck is the background theory uncertainty
1 / 30
Theory priorities in ttH searches
Key priority is precision for backgrounds
various multi-particle processes: tt+ jets,ttV+ jets, ttγγ, V V+ jets
NLO automation crucial but 2→ 4 CPUintensive
ttH Analyses at the LHC
H
b
b
t
W q
q
b
t
W l
ν
b• direct probe of ttH Yukawa coupling
• ttH(bb) originally considered best discovery channel for
light Higgs
• complicated bbbbℓνjj final state hampers H → bb peak
reconstruction ⇒ very large QCD backgrounds
Status of ttH analyses in various channels
95% exclusion in σSMttH units H → bb H → V V ∗ H → γγ
ATLAS 4.1 (2.6) 4.7 (5.4)
CMS 5.2 (4.1) 6.6 (2.4) 5.3 (5.4)
• sensitivity already ∼ 100%λSMt and ttH searches still quite active
• expected sensitivity ∼ 10%λSMt with 300 fb−1
NLO matching & merging crucial
various new methods (FxFx, MEPS@NLO, MINLO, UNLOPS,. . . )
various automated tools support NLO precision for signal and most backgrounds:MG5 aMC@NLO, Sherpa+OpenLoops/GoSam, Powheg/Powhel
Theory uncertainty estimates nontrivial
still limited experience in NLO matching+merging framework
sophisticated analyses (profile likelihood, MEM, background reweighting, . . . )
2 / 30
Recent progress in theory precision [slide by M. Zaro]
Marco Zaro, 17-09-2015
2015!
(4FS and 5FS) Demartin et al. arXiv:1504.00611 tHW see poster by F. Demartin
2015!
2015!
2015!
2015!
• t tbb • NLO QCD corrections !!
• Matching to PS • t tV • NLO QCD corrections !!!
• Matching to PS !
• Electro-Weak corrections !• t tVV
• NLO QCD corrections + PS
Higher order predictions for signal and backgrounds
4
Beenakker et al. hep-ph/0107081 & hep-ph/0211352Dawson et al. hep-ph/0211438 & hep-ph/0305087
aMC@NLO: Frederix et al. arXiv:1104.5613 Powhel: Garzelli et al. arXiv:1108.0387
Powheg Box: Hartanto et al. arXiv:1501.04498
Frixione et al. arXiv:1407.0823 & arXiv:1504.03446 Zhang et al. arXiv:1407.1110
Kulesza et al. arXiv:1509.02780
Bredenstein et al. arXiv:0905.0110 & arXiv:1001.4006 Bevilacqua et al. arXiv:0907.4723
Denner et al. arXiv:1506.07448
Kardos et al.1303.6201 Cascioli et al. 1309.5912
t t�� Melnikov et al. arXiv:1102.1967
t tW,t t��*/Z, t t�� Hirschi et al. arXiv:1103.0621 t tZ Lazopoulos et al. arXiv:0804.2220
t tZ Kardos et al. arXiv:1111.0610 t tW Campbell et al. arXiv:1204.5678
t tZ Garzelli et al. arXiv:1111.1444 t tW, t tZ Garzelli et al. arXiv:1208.2665
t t��𝛄 Kardos et al. arXiv:1408.0278 all t tVV Maltoni et al. arXiv:1507.05640
t t��𝛄 van Deurzen et al. arXiv:1509.02077
t tW, t tZ (and t tH) Frixione et al. arXiv:1504.03446(5FS) Farina et al. arXiv:1211.3737
(5FS) Campbell et al. arXiv:1302.3856
2015!
2015!
2015!
• t tH • NLO QCD corrections (30% @ RunII)
!• Matching to PS
!!
• NLO QCD corrections to bbℓ+ℓ-ννH !• Weak and Electro-Weak corrections
(1.5% @ RunII) !
• Soft gluon resummation (2-6% @ RunII) !• tH • NLO QCD corrections
!• Matching to PS
3 / 30
Outline
1 ttH Signal
2 Backgrounds for H → bb
3 Backgrounds for H →WW,ZZ, ττ →multi-leptons
4 Backgrounds for H → γγ
NLO tools for ttH
σ(pp → tt_ H + X) [fb]
√s = 14 TeV
MH
= 120 GeV
µ0 = m
t + M
H/2
NLO
LO
µ/µ0
0.2 0.5 1 2 5
0
200
400
600
800
1000
1200
1400 σttH at NLO [Beenakker et al ’01; Reina, Dawson ’01]
moderate uncertainty (9% scale, 8% PDF+αS)
sufficient for mid-term future
Publicly available NLO+PS tools
MG5 aMC@NLO [Frederix et al. ’15]
Powhel [Garzelli at al. ’11] andPowheg Box [Jaeger et al. ’15]
Sherpa+OpenLoops or GoSam
0
0.0001
0.0002
0.0003
0.0004
0.0005
0.0006
0.0007
0.0008
0.0009
0.001
0 50 100 150 200 250 300 350dm
/dp T
(H) [
pb/G
eV]
POWHEG+PYTHIA 6POWHEG+PYTHIA 8POWHEG+HERWIG
NLO
0.8
1
1.2
0 50 100 150 200 250 300 350
R
pT(H) [GeV]
<∼ 10% uncertainty for inclusive observables (more if sensitive to jet radiation)4 / 30
Spin-correlated Top decays
Marco Zaro, 17-09-2015
Including spin correlations at NLO Frixione, Leanen, Motylinski, Webber, arXiv:hep-ph/0702198
• Generate events (to be showered) for the production process MP
• Before showering, produce a decayed event file starting from the undecayed events
• Exploit the fact that |M2P+D|/|M2P| is bounded from above
• The generation of unweighted decayed events is possible: generate many kinematics configurations until
• In NLO computations use only tree-level matrix elements (with n or n+1 particles)
• Loop effects on spin correlation assumed to be negligible
• Automated in MadSpin (MadGraph5_aMC@NLO) and Decayer (PowHel)
21
qL
qR
t
t
b
b
W+
W�
l+
⌫l
l�
⌫l
|MP+D|2 / |MP |2 > Rand() max
⇣|MP+D|2 / |MP|2
⌘
Artoisenet, Frederix, Mattelaer, Rietkerk, arXiv:1212.3460 Garzelli, Kardos, Trocsanyi, arXiv:1405.5859
Mandatory for signal and all backgrounds
ttγγ : uncorttγγ : corr
tth(γγ) : uncortth(γγ) : corr
Frame 1
cos θℓℓ
1 NdN
dcosθℓℓ
10.60.2-0.2-0.6-1
0.65
0.6
0.55
0.5
0.45
0.4
[Biswas et al ’14]
NLO+PS level automation in Powheg, MG5 aMC@NLO and Sherpa
via full production×decay LO MEs for ttH(+j) assuming same 1-loop correlations
only (NLO production)×(LO decay) accuracy
5 / 30
Soft-gluon NLL resummation for σttH [Kulesza et al. ’15]
Resummation of threshold logarithms of β =√
1− (2mt +mH)2/s
optimal for mH + 2mt →√s
logs are β4-suppressed by ttH phase space ⇒ no realistic estimate of full NNLO
Matched prediction: σttHNLO+NLL = 650+7.9%−5.7% (σttHNLO = 613+6.2%
−9.4%)
moderate +7% correction and modest reduction of scale uncertainty
dominant effects from O (αS) matching coefficient C(1) (formally NNLL)
0.00
200.00
400.00
600.00
800.00
1000.00
0.2 0.5 1 2 5µR/µ0
σ(pp→ Htt+X)[fb]√S = 14 TeV
µ0 = mt +mH/2mH = 125 GeV
LO
NLO
NLO+NLL
NLO+NLL(w C-coef);
0.00
200.00
400.00
600.00
800.00
1000.00
0.2 0.5 1 2 5µF /µ0
σ(pp→ Htt+X)[fb]√S = 14 TeV
µ0 = mt +mH/2mH = 125 GeV
LO
NLO
NLO+NLL
NLO+NLL(w C-coef);
6 / 30
Soft-gluon NNLL resummation for dσ/dΦttH [Broggio et al. ’15]
SCET PIM resummation formalism
resums threshold-type logarithms of 1− z = 1−M2ttH/s
valid for MttH > 2mt +mH and optimal for M2ttH → s
⇒ applicable to differential distributions
nNLO approx. from truncated NLO+NNLL µ0 = mt +mH/2 = 235 GeV
µ = 2.64 µ0 ⇒ +8% nNLO correction & 5% uncertainty (scale+subleading logs)
standard scale µ = µ0 ⇒ smaller nNLO correction
0 1 2 3 4 5
350
400
450
500
550
μ
235GeV
σ[fb
] approx NLO
nNLO
NLO MG5
NLO MG5 no qg
MSTW 2008 PDFs
μ0 = 620 GeV
0 50 100 150 200 250 300 3500
5
10
15
20
25
30
35
PTH [GeV]
crosssectionperbin[fb
]
7 / 30
NLO QCD corrections to ttH + 1 jet
Parton-level NLO with Sherpa+GoSam[Deurzen et al, ’13]
more reliable prediction and uncertainty forextra jet activity
room for significant effects in the hardregion (depending on µ)
NNLO contributions of hard type
Applicable to ttH analysis through ttH + 0, 1 jets NLO merging (automated)
MEPS@NLO merging [Hoche et al. ’12] in Sherpa+OpenLoops/GoSam
FxFx merging [Frederix, Frixione ’12] in MG5 aMC@NLO
⇒ impact on σttH and ttH searches?
8 / 30
ttH (and ttV ) at NLO QCD+EW [Frixione et al. ’14–15; Zhang et al. ’14]
Emerging NLO QCD+EW automation
Recola [1411.0916], OpenLoops [1412.5157],MG5 aMC [1504.03446], GoSam [1507.08579]
applicable to ttH signal and backgrounds(matching/merging automation ongoing)
large NLO EW effects at pT �MW
NLO EW corrections to ttH and ttV
NLO correction to mt = λtv/√
2 relation
impact below QCD scale+PDF uncertainties andmore significant for backgrounds
ttH: −1% for σtot (-8% at pT > 200 GeV)
ttW : −8% for σtot (-12% at pT > 200 GeV)σ
pe
r b
in [
pb
]
tt-H production at the 13 TeV LHC
LO QCD
LO+NLO QCD
LO+NLO QCD+EW
LO+NLO QCD+EW, no γ
10-4
10-3
10-2
10-1
MadGraph5_aMC@NLO
1
1.2
1.4ratio over LO QCD; scale unc.
1
1.2
1.4ratio over LO QCD; PDF unc.
pT(H) [GeV]
relative contributions
NLO QCD
LO+NLO EW
LO+NLO EW, no γ
HBR
-0.2
0
0.2
0.4
0 200 400 600
9 / 30
pp→ ttH → e+νeµ−νµbbH at NLO QCD [Denner and Feger ’15]
Challenging 2→ 5 NLO calculation
heptagons, complex mass scheme for top resonances
Recola+Collier [Denner, Dittmaier, Hofer, Uccirati]
g
g
t
t
H
b
t
W+
e+
νe
t
b
W−
νµ
µ−
g
σ[f
b]
Γt [GeV]
LONLO
2.2
2.3
2.4
2.5
2.6
2.7
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Results for total cross section with tt-type cuts
excellent agreement with NWA (Γt → 0) as for tt
bigger off-shell effects from pp→ tWH if one bunresolved (would require mb > 0)
Distributions with µdyn = (ET,tET,tET,H)1/3
nontrivial NLO shape corrections
due to NLO corrections to production and decay?
⇒ cf. (NLO production)×(LO+PS decay) in MCdσ
dφe+
µ−
[ fb ◦]
Kfa
ctor
φe+µ− [◦]
LONLO
0.005
0.01
0.015
0.02
0.025
0.03
0.81
1.21.4
0 20 40 60 80 100 120 140 160 180
10 / 30
Outline
1 ttH Signal
2 Backgrounds for H → bb
3 Backgrounds for H →WW,ZZ, ττ →multi-leptons
4 Backgrounds for H → γγ
2→ 8 simulation of ttH(bb) signal and irred. backg. I
g
g
bt
W−u
dt
t
tb
W+
ℓ+
νℓ
Hb
b
g
g
bt
W−u
dt
t
W+
tb
W+
ℓ+
νℓ
b
b
g
g
bt
W−u
dt
t
tb
W+
ℓ+
νℓ
g
b
b
g
g
b
b
g
g
u
g
b
b
b
u
ud
W+
ℓ+
νℓ
O(α1S
)O
(α1S
)O
(α2S
)O
(α3S
)pp→ `ν + 2j + 4b at LO [Denner, Feger, Scharf, 1412.5290]
48 partonic channels and up to 78’000 diagrams/channel at LO
MEs of order α0S , α1
S , α2S , α3
S with interferences and non-resonant effects
11 / 30
2→ 8 simulation of ttH(bb) signal and irred. backg. II
g
g
bt
W−u
dt
t
tb
W+
ℓ+
νℓ
Hb
b
g
g
bt
W−u
dt
t
W+
tb
W+
ℓ+
νℓ
b
b
g
g
bt
W−u
dt
t
tb
W+
ℓ+
νℓ
g
b
b
g
g
b
b
g
g
u
g
b
b
b
u
ud
W+
ℓ+
νℓ
O(α1S
)O
(α1S
)O
(α2S
)O
(α3S
)O (αS) EW contributions to ttbb
ttH (70%), ttZ (∼ 10%), ttγ∗ (∼ 10%), and t-channel W -exchange (∼ 10%)
Interference between O(α2S
)QCD and O (αS) EW contributions to ttbb
−8% wrt ttbb QCD, mostly from t-channel W -exchange (tiny ttH interference)
Off-shell contributions with ≤ 1 top resonance (new qg and qq′ channels)
+12% wrt ttbb QCD (might be attributed to Wbb+jets?)
Non-standard background contributions of order 10–50% σttH
12 / 30
Irreducible ttbb QCD background at NLO
NLO ttbb [Bredenstein et al ’09/’10; Bevilacqua et al ’09];
ttbb dominates ttH(bb) systematics
NLO reduces uncertainty from 80% to 20–30%
NLO+PS ttbb 5F scheme (mb = 0) with Powhel [Garzelli et al ’13/’14]
ttbb NLO MEs cannot describe collinear g → bb splittings
⇒ inclusive tt+b-jets simulation requires parton shower in collinear bb region
⇒ NLO merging tt+ 0, 1, 2 jets (see later)
NLO+PS ttbb 4F scheme (mb > 0) with Sherpa+OpenLoops [Cascioli et al ’13]
ttbb NLO MEs cover full b-quark phase space
⇒ inclusive NLO accurate tt+b-jets simulation possible
13 / 30
S–MC@NLO ttbb 4F scheme [Cascioli et al ’13]
Good perturbative stability but unexpected MC@NLO enhancementttb ttbb ttbb (mbb > 100)
σLO[fb] 2644+71%−38%
+14%−11% 463.3+66%
−36%+15%−12% 123.4+63%
−35%+17%−13%
σNLO[fb] 3296+34%−25%
+5.6%−4.2% 560+29%
−24%+5.4%−4.8% 141.8+26%
−22%+6.5%−4.6%
σNLO/σLO 1.25 1.21 1.15
σMC@NLO[fb] 3313+32%−25%
+3.9%−2.9% 600+24%
−22%+2.0%−2.1% 181+20%
−20%+8.1%−6.0%
σMC@NLO/σNLO 1.01 1.07 1.28
Large enhancement (∼30%) in Higgs region from double g → bb splittings
matching, shower and 4F/5F systematicsremain to be understood!
14 / 30
tt+ jets background and merging at NLO
NLO tt+ 2 jets [Bevilacqua, Czakon, Papadopoulos, Worek ’10/’11]
reduces uncertainty from 80% to 15%
NLO merging (FxFx, MEPS@NLO, UNLOPS,. . . )
0-jet NLO+PS tt1-jet NLO+PS tt+ 1 j. . . . . .≥ n jets NLO+PS tt+ n j
NLO and log accuracy for 0, 1, . . . n jets
separated via kT-algo at merging scale Qcut
smooth PS–MEs transition ↔ MEs with PS-likescale and Sudakov FFs
NLO merging for tt+ 0, 1 jets
FxFx with Madgraph5/aMC@NLO[Frederix, Frixione ’12]
MEPS@NLO with Sherpa+GoSam[Hoche et al ’13]
15 / 30
MEPS@NLO for tt+ 0, 1, 2 jets (Sherpa+OpenLoops)[Hoche, Krauss, Maierhofer, S. P. , Schonherr, Siegert ’14] I
Sherpa+OpenLoops
pl-jet⊥ > 40 GeV
pl-jet⊥ > 60 GeV
×0.1
pl-jet⊥ > 80 GeV
×0.01
[email protected] × MEPS@LOS-MC@NLO10−5
10−4
10−3
10−2
10−1
1Inclusive light jet multiplicity
σ(≥
Nl-
jet)
[pb]
pl-jet⊥ > 40 GeV
1
1.5
2
Rat
ioto
ME
PS@
NL
O
pl-jet⊥ > 60 GeV
1
1.5
2
Rat
ioto
ME
PS@
NL
O
pl-jet⊥ > 80 GeV
0 1 2 3
1
1.5
2
Nl-jet
Rat
ioto
ME
PS@
NL
O
Consistency with LO merging and NLO+PS
decent (10–20%) mutual agreement
Reduction of µR, µF , µQ variationsNlight−jet ≥ 0 1 2
LO 48% 65% 80%NLO 17% 18% 19%
16 / 30
MEPS@NLO for tt+ 0, 1, 2 jets (Sherpa+OpenLoops)[Hoche, Krauss, Maierhofer, S. P. , Schonherr, Siegert ’14] II
Sherpa+OpenLoops
1st jet
2nd jet
3rd jet
[email protected] × MEPS@LOS-MC@NLO10−7
10−6
10−5
10−4
10−3
Light jet transverse momenta
dσ
/d
p T[p
b/G
eV]
1st jet0.5
1
1.5
Rat
ioto
ME
PS@
NL
O
2nd jet0.5
1
1.5
Rat
ioto
ME
PS@
NL
O
3rd jet
40 50 100 200 500
0.5
1
1.5
pT (light jet) [GeV]
Rat
ioto
ME
PS@
NL
O
Consistency with LO merging and NLO+PS
decent (10–20%) mutual agreement
Reduction of µR, µF , µQ variationsNlight−jet ≥ 0 1 2
LO 48% 65% 80%NLO 17% 18% 19%
Differential distributions
similarly mild scale dependence
small shape corrections
Next steps
improve CPU performance
consistent combination with ttbb
17 / 30
MEPS@NLO for tt+ 0, 1, 2 jets (Sherpa+OpenLoops)[Hoche, Krauss, Maierhofer, S. P. , Schonherr, Siegert ’14] III
Sherpa+OpenLoops
MEPS@NLOMEPS@NLO Qcut = 20 GeVMEPS@NLO Qcut = 30 GeVMEPS@NLO Qcut = 40 GeV1.65×MEPS@LOS-MC@NLO
1
10 1
10 2
10 31 → 2 k⊥ jet resolution
dσ
/d
d 12
[pb/
GeV
]
1
1.5
→ ⊥
Rat
ioto
Qcu
t=
30G
eV
0.8
1
1.2
→ ⊥
Rat
ioto
Qcu
t=
30G
eV
3 10 100 3000
0.2
0.4
0.6
0.8tt+ 0j excl.tt+ 1j excl.tt+ 2j incl.
→ ⊥
d12 [GeV]
Rat
ioto
Qcu
t=
30G
eV
Small merging scale choice
Qcut = 30 GeV such that exp. resolvedjets are described by MEs
Merging scale uncertainty
Qcut = 30± 10 GeV
⇒ � 10% dependence
does not spoil tt+ 0, 1, 2 jets NLO precision
18 / 30
Scale dependencies in NLO matching and merging
Scale variations
basis for definition of intrinsic theory precision of (N)NLO Monte Carlo
quantitative and realistic uncertainty estimate is a key advantage of (N)NLO MCwrt (tuned) LO MC
Four scale variations
standard prescription for renormalisation (µR) and factorisation (µF ) scales
no widely accepted prescription for resummation (µQ) and merging (Qcut) scales
19 / 30
Benefits of NLO matching and merging (top-pair pT in tt+ 0, 1 jets)LO
−→NLO
Sherpa+OpenLoops
SMCNLO R1F1Q1SMCNLO R0.5F1Q1SMCNLO R2F1Q1SMCNLO R1F0.5Q0.5SMCNLO R1F2Q2
10−2
10−1
1
pT of top-antitop pair
dσ/dpT[pb/GeV
]
0 50 100 150 200 250 300 350 400
0.6
0.8
1
1.2
1.4
pT [GeV]
MC/Data
Sherpa+OpenLoops
LOPS R1F1Q1LOPS R0.5F1Q1LOPS R2F1Q1LOPS R1F0.5Q0.5LOPS R1F2Q2
10−6
10−5
10−4
10−3
10−2
10−1
1
pT of top-antitop pair
dσ/dpT[pb/GeV
]
0 50 100 150 200 250 300 350 400
0.6
0.8
1
1.2
1.4
pT [GeV]
MC/Data
Sherpa+OpenLoops
MEPSNLO30 R1F1Q1MEPSNLO30 R0.5F1Q1MEPSNLO30 R2F1Q1MEPSNLO30 R1F0.5Q0.5MEPSNLO30 R1F2Q210−2
10−1
1
pT of top-antitop pair
dσ/dpT[pb/GeV
]
0 50 100 150 200 250 300 350 400
0.6
0.8
1
1.2
1.4
pT [GeV]
MC/Data
Sherpa+OpenLoops
MEPSLO30 R1F1Q1MEPSLO30 R0.5F1Q1MEPSLO30 R2F1Q1MEPSLO30 R1F0.5Q0.5MEPSLO30 R1F2Q2
10−2
10−1
1
pT of top-antitop pair
dσ/dpT[pb/GeV
]
0 50 100 150 200 250 300 350 400
0.6
0.8
1
1.2
1.4
pT [GeV]
MC/Data
µR = ξ mt
renorm. scale
µQ = µF = ξ mt
factorisation scale
resummation scale
NLO merging
10% accuracy overwhole spectrum!
matching −→ merging
20 / 30
Choice and variation of merging scale Qcut?!
Virtue of small Qcut
NLO accuracy over whole pT spectrum of experimentally resolved jets
Formal problem at (very) small Qcut [Hamilton, Nason, Oleari, Zanderighi ’12]
when Qcut approaches Sudakov peak, i.e. αS log2(Qcut/Qhard) ∼ 1
⇒ uncontrolled O(α1.5S
)subleading logs can spoil NLO precision
⇒ Optimal Qcut choice? Uncertainty?
Quantitative estimate (conservative)
consider jet-kT dist ⇒ max Qcut sensitivity
push Qcut down to Sudakov region
take local Qcut dependence as uncertainty estimate
stop before you exceed NLO scale dependence
Sherpa+OpenLoops
MEPSNLO30 R1F1Q1MEPSNLO5 R1F1Q1MEPSNLO10 R1F1Q1MEPSNLO20 R1F1Q1MEPSNLO40 R1F1Q1
10−4
10−3
10−2
10−1
1
10 1
10 2
log10(k⊥ jet resolution 0 → 1 [GeV])
dσ/dlog10(d
01/GeV
)[pb]
0.5 1 1.5 2 2.5 3 3.5
0.6
0.8
1
1.2
1.4
log10(d01/GeV)MC/Data
Example: MEPS@LO for tt+ 0, 1 jets
Qcut = 40 . . . 5 GeV ⇒ less than 10% uncertainty for Qcut ≥ 10 GeV
21 / 30
Outline
1 ttH Signal
2 Backgrounds for H → bb
3 Backgrounds for H →WW,ZZ, ττ →multi-leptons
4 Backgrounds for H → γγ
Most relevant Backgrounds to ttH(H → V V )
tt+W/Z/γ∗+extra jets
MC systematics dominated by ttW + 0, 1, 2 jets
same signature as ttH(H →WW → `νjj)!
tt+ Z/γ(→ `+`−) requires off-shell effects down to small m``
V V+jets
mainly WZ → 3` but also ZZ → 4`
requires inclusive sample with NLO precision up to 2 jets
HF jets crucial (genuine V V bb component and light-jet mistags)
22 / 30
NLO+PS tt+W±/Z with Powheg/Powhel [Garzelli et al ’12]
2
510
-22
510
-12
512
0.90.951.0
1.051.1
0.80.91.01.11.2
0 50 100 150 200 250 300 350 400 450 500
dσ
dp ⊥
,t[fb/G
eV]
√s = 7TeV
mt = 172.9GeVmZ = 91.1876GeVµ0 = mt +mZ/2CTEQ6.6M
(a) PowHel-NLOPowHel-LHEµ0/2 ≤ µ ≤ 2µ0
LHE/N
LO
K-fact
p⊥,t [GeV]
LHE/NLO
K-fact
Moderate NLO+PS corrections
K = 1.1–1.35 and mild uncertainties (10% scale, 8% PDFs)
top decays in NWA (uncorrelated)
ttV + 2 jets from PS
23 / 30
NLO+PS predictions for ttV and ttV V [Maltoni et al. ’15]
Thorough study with scale+PDF variations for all V = W,Z, γ
moderate K-factors and shape corrections at 13–14 TeV
small irreducible ttH(V V ) backgrounds: 2 fb ≤ σttV V ≤ 12 fb
[pb]
NLO
σ
110
1
10
210
H production at pp colliders at NLO in QCDtV, ttt
Ztt
±Wtt
γtt
Htt
, MSTW2008 NLO PDFs (68% cl)g
µ = r
µ = f
µcentral
MadGraph5_aMC@NLO
| | | | |
fa
cto
rsK 1
1.2
1.4
| | | | |
[TeV]s
facto
rsK
1
1.5
2
8 13|
14|
25|
33|
50|
100
[fb
]N
LO
σ
1
10
210
310
410 production at pp colliders at NLO in QCDtttVV, ttt
ZZtt
[4f]
W+
Wtt
γγtt
Z±Wtt
γZtt
γ±Wtt
, MSTW2008 NLO PDFs (68% cl)g
µ = r
µ = f
µcentral
tttt
MadGraph5_aMC@NLO
| | | | |
fa
cto
rsK 1
1.2
1.4
| | | | |
[TeV]s
facto
rsK
1
1.5
2
8 13|
14|
25|
33|
50|
100
24 / 30
Backgrounds to ttH →multileptons [Maltoni et al. ’15]
Simulation of 2-, 3- and 4-lepton CMS searches at 13 TeV
σNLO+PS [fb] 2 same-sign leptons 3 leptons 4 leptons
ttH(H → WW∗) 1.54+5.1%−9.0%
+2.2%−2.6%
1.47+5.2%−9.0%
+2.0%−2.4%
0.095+7.4%−9.7%
+2.0%−2.4%
ttH(H → ZZ∗) 0.0437+5.5%−9.2%
+2.3%−2.8%
0.119+6.3%−9.6%
+2.1%−2.5%
0.0170+5.0%−8.5%
+2.0%−2.4%
ttH(H → τ+τ−) 0.563+4.6%−8.8%
+2.2%−2.7%
0.669+6.0%−9.4%
+2.1%−2.6%
0.0494+7.1%−9.9%
+2.1%−2.5%
ttW± 5.77+15.1%−12.7%
+1.6%−1.2%
2.44+13.1%−11.6%
+1.7%−1.4%
-
ttZ/γ∗ (→ `+`−) 1.61+7.7%−10.5%
+2.0%−2.5%
2.70+9.0%−11.2%
+2.0%−2.5%
0.280+9.8%−11.0%
+1.9%−2.3%
ttW+W− 0.288+8.0%−11.1%
+2.3%−2.6%
0.201+7.4%−10.7%
+2.1%−2.3%
0.0116+6.9%−10.2%
+2.2%−2.3%
tttt 0.340+27.5%−25.8%
+5.5%−6.4%
0.211+27.4%−25.6%
+5.2%−6.1%
0.0110+27.0%−25.5%
+5.0%−5.9%
ttZZ 0.0096+3.5%−8.4%
+1.8%−1.8%
0.0050+3.7%−8.3%
+1.8%−1.7%
0.00025+7.2%−9.6%
+1.9%−1.8%
ttW±Z 0.0620+9.0%−10.2%
+2.2%−1.6%
0.0279+9.2%−10.3%
+2.3%−1.7%
0.00091+7.2%−9.2%
+2.4%−1.7%
B/S up to 350% for ttV and only 15–25% for ttWW + tttt
NLO precision (10–15%) seems sufficient apart from ttW (and ttW+jets!)
25 / 30
Outline
1 ttH Signal
2 Backgrounds for H → bb
3 Backgrounds for H →WW,ZZ, ττ →multi-leptons
4 Backgrounds for H → γγ
Backgrounds to ttH(γγ) and MC predictions
σttH(γγ) ∼ 1.5 fb contaminated by mild backgrounds
tt+ 1, 2γ, single-top+1, 2γ, jets+1, 2γ
H+jets with H → γγ and HF jets
NLO+PS for ttγ and ttγγ with Powhel [Kardos, Trocsanyi ’14]
mild NLO uncertainty (14%) and MC effects for inclusive observables
realistic IR-safe hard-photon isolation (without q → qγ fragmentation): looseFrixione isolation (generation cuts) + tighter cone isolation (analysis cuts)
2
5
10−3
2
5
10−2
dσ/dmγ1,γ
2[fb/G
eV]
PowHel LHEPowHel NLO
8 TeV, µ = mt
δ0 = 0.4
0.91.01.1
ratio
0 50 100 150 200 250 300 350 400 450 500
mγ1, γ2 [GeV]
5
10−3
2
5
10−2
2
dσ/dmγ1,γ
2[fb/G
eV]
SMC Rγ, q = 0.1SMC δ0 = 0.1
13 TeV, µ = H⊥/2
0.91.01.1
ratio
0 50 100 150 200 250 300 350 400 450 500
mγ1, γ2 [GeV]26 / 30
ttγγ background in ttH(γγ) searches [Maltoni et al. ’15]
MC@NLO simulations of ttH(γγ) and ttγγ see also [van Deurzen at al. ’15]
CMS-like cuts for ttH(γγ) search with one lepton
Higgs peak fully reconstructed within |mγγ −mH | < 2 GeV mass window
dσ
/dm
[p
b/b
in]
NLO+PS (µg), LHC13 ttγγ
ttH(γγ)
10−6
10−5
10−4
MadGraph5_aMC@NLO
ΚP
Stt H
LO unc. NLO unc.
0.5
1
1.5
ΚP
Stt γγ
m(γ1γ2) [GeV]
0.5
1
1.5
100 110 120 130 140 150 160 170 180
20 signal events/100 fb−1
S/Birred. ∼ 10
NLO+PS for S and Birred. well undercontrol: <∼ 10% uncertainty
⇒ MC better than side-band approach forbackground?
What about reducible backgrounds (fake photons, leptons,. . . )?
27 / 30
Subtleties in tt+multi-γ production
Radiative top decays (not included in recent NLO MC studies)
two photons can arise from tt+ 0, 1, 2 γ production plus 2, 1, 0 γs from top decays
photons with Mb`νγ < mt mainly from top-decay products (light leptons/quarks)
NLO study of tt+ 0, 1 γ with t→ b`+ν + 0, 1 γ [Melnikov, Schulze ’11]
0
0.2
0.4
0.6
0.8
1
0 20 40 60 80 100 120 140 160 180 200
dσX NLO
dpT(γ)
/dσNLO
dpT(γ)
pT(γ) [GeV]
γ in production
γ in decay
1 2 30
0.2
0.4
0.6
0.8
1
dσXNLO
dR
/dσNLO
dR
R(γ, bjet)
γ in production
γ in decay
photons from top decays dominate up to pT(γ) <∼ 60 GeV (smooth isolation, Rγi > 0.4)
ttH(γγ) searches require tt+X background with nγ = 1, 2
radiative top decays can be generated with QED parton shower
tt+ 0, 1, 2 γ MEs need to be merged (to avoid double counting)
28 / 30
Relevance of radiative top decays for ttH(γγ) searches
������ ���� n���� g
g
�� � ������ ���� n����g
g�� ������ ���� n���� g
g�� �Fig. 146: A subsample of the relative Feynman graphs illustrating the three types ofttγγ processes.
Table 27: Cross-sections at leading order (statistical errors in parentheses), number of events generated, numbers ofevents and
statistical weight/generated event for 30 and 100 fb−1 of integrated luminosity for the irreducible backgrounds considered.
Process σ× BR(1 W→ lν) Ngen N 30 fb−1 Wgt 30 fb−1 N 100 fb−1 Wgt 100 fb−1 Generator
ttγγ 1 1.6 fb (≤ 1/mil) 9296 48 .0052 160 .0172 AL,MGttγγ 2 6.1 fb (≤ 1%) 2310 183 .0792 610 .2641 ALttγγ 3 4.9 fb (≤ 1%) 914 147 .1608 490 .5361 ALbbγγ 318.1 fb 159829 9543 .0597 31810 0.1990 MG
Wγγ 4j 11.5 fb (1.2%) 4587 345 0.0752 1150 .2507 ALZγγ 29.0 fb 50005 870 0.0174 2900 0.0580 MGWγγ 23.6 fb 112000 708 0.0063 2360 0.0211 MG
identified due to incomplete detector coverage or other instrumental effects. This could arise if one ormore electrons or jets are misidentified as photons, or a jet as an electron or a muon. Therefore possiblebackground processes can be grouped into the following signature categories:llγ, llj, ljj, lγj, γγj, γjj,jjj, wherel is a lepton andj is a jet. Table 28 lists the reducible background processes to be considereedfor each category. It should be noted that several processescould contribute to more than one signaturecategory.
During the time horizon of the workshop, due to the implementation of the many new generatorprocesses, it has been possible to study only the irreducible backgrounds with acceptable statistics, soonly these will be presented in this report. Low-statisticstests on most of the processes in Table 28 havebeen performed, and as many of these processes as possible will be included with high statistics in adefinitive study now in progress with events fully simulatedand reconstructed in the CMS detector.
All generated signal and background events were fragmentedand hadronized with PYTHIA [27,284] version 6.227.
25.4 Description of preselections
No generator-level preselections were made on signal events. For the irreducible background events, thefollowing preselection was made:
• mγγ ≥80 GeV + where applicable:
• pTγ ≥ 20 GeV,|ηγ | ≤ 2.5 or pTγ ≥ 15 GeV,|ηγ | ≤ 2.7
• pTj,l,b ≥ 15 GeV,|ηj,l,b| ≤ 2.7,∆R(l,j or j,j or b,b orγ,j or γ, γ) ≥ 0.3
where pT refers to the transverse momentum of the particle,η its rapidity and ∆R =√(∆η2 + ∆φ2) whereφ is the azimuthal angle.
146
������ ���� n���� g
g
�� � ������ ���� n����g
g�� ������ ���� n���� g
g�� �Fig. 146: A subsample of the relative Feynman graphs illustrating the three types ofttγγ processes.
Table 27: Cross-sections at leading order (statistical errors in parentheses), number of events generated, numbers ofevents and
statistical weight/generated event for 30 and 100 fb−1 of integrated luminosity for the irreducible backgrounds considered.
Process σ× BR(1 W→ lν) Ngen N 30 fb−1 Wgt 30 fb−1 N 100 fb−1 Wgt 100 fb−1 Generator
ttγγ 1 1.6 fb (≤ 1/mil) 9296 48 .0052 160 .0172 AL,MGttγγ 2 6.1 fb (≤ 1%) 2310 183 .0792 610 .2641 ALttγγ 3 4.9 fb (≤ 1%) 914 147 .1608 490 .5361 ALbbγγ 318.1 fb 159829 9543 .0597 31810 0.1990 MG
Wγγ 4j 11.5 fb (1.2%) 4587 345 0.0752 1150 .2507 ALZγγ 29.0 fb 50005 870 0.0174 2900 0.0580 MGWγγ 23.6 fb 112000 708 0.0063 2360 0.0211 MG
identified due to incomplete detector coverage or other instrumental effects. This could arise if one ormore electrons or jets are misidentified as photons, or a jet as an electron or a muon. Therefore possiblebackground processes can be grouped into the following signature categories:llγ, llj, ljj, lγj, γγj, γjj,jjj, wherel is a lepton andj is a jet. Table 28 lists the reducible background processes to be considereedfor each category. It should be noted that several processescould contribute to more than one signaturecategory.
During the time horizon of the workshop, due to the implementation of the many new generatorprocesses, it has been possible to study only the irreducible backgrounds with acceptable statistics, soonly these will be presented in this report. Low-statisticstests on most of the processes in Table 28 havebeen performed, and as many of these processes as possible will be included with high statistics in adefinitive study now in progress with events fully simulatedand reconstructed in the CMS detector.
All generated signal and background events were fragmentedand hadronized with PYTHIA [27,284] version 6.227.
25.4 Description of preselections
No generator-level preselections were made on signal events. For the irreducible background events, thefollowing preselection was made:
• mγγ ≥80 GeV + where applicable:
• pTγ ≥ 20 GeV,|ηγ | ≤ 2.5 or pTγ ≥ 15 GeV,|ηγ | ≤ 2.7
• pTj,l,b ≥ 15 GeV,|ηj,l,b| ≤ 2.7,∆R(l,j or j,j or b,b orγ,j or γ, γ) ≥ 0.3
where pT refers to the transverse momentum of the particle,η its rapidity and ∆R =√(∆η2 + ∆φ2) whereφ is the azimuthal angle.
146
������ ���� n���� g
g
�� � ������ ���� n����g
g�� ������ ���� n���� g
g�� �Fig. 146: A subsample of the relative Feynman graphs illustrating the three types ofttγγ processes.
Table 27: Cross-sections at leading order (statistical errors in parentheses), number of events generated, numbers ofevents and
statistical weight/generated event for 30 and 100 fb−1 of integrated luminosity for the irreducible backgrounds considered.
Process σ× BR(1 W→ lν) Ngen N 30 fb−1 Wgt 30 fb−1 N 100 fb−1 Wgt 100 fb−1 Generator
ttγγ 1 1.6 fb (≤ 1/mil) 9296 48 .0052 160 .0172 AL,MGttγγ 2 6.1 fb (≤ 1%) 2310 183 .0792 610 .2641 ALttγγ 3 4.9 fb (≤ 1%) 914 147 .1608 490 .5361 ALbbγγ 318.1 fb 159829 9543 .0597 31810 0.1990 MG
Wγγ 4j 11.5 fb (1.2%) 4587 345 0.0752 1150 .2507 ALZγγ 29.0 fb 50005 870 0.0174 2900 0.0580 MGWγγ 23.6 fb 112000 708 0.0063 2360 0.0211 MG
identified due to incomplete detector coverage or other instrumental effects. This could arise if one ormore electrons or jets are misidentified as photons, or a jet as an electron or a muon. Therefore possiblebackground processes can be grouped into the following signature categories:llγ, llj, ljj, lγj, γγj, γjj,jjj, wherel is a lepton andj is a jet. Table 28 lists the reducible background processes to be considereedfor each category. It should be noted that several processescould contribute to more than one signaturecategory.
During the time horizon of the workshop, due to the implementation of the many new generatorprocesses, it has been possible to study only the irreducible backgrounds with acceptable statistics, soonly these will be presented in this report. Low-statisticstests on most of the processes in Table 28 havebeen performed, and as many of these processes as possible will be included with high statistics in adefinitive study now in progress with events fully simulatedand reconstructed in the CMS detector.
All generated signal and background events were fragmentedand hadronized with PYTHIA [27,284] version 6.227.
25.4 Description of preselections
No generator-level preselections were made on signal events. For the irreducible background events, thefollowing preselection was made:
• mγγ ≥80 GeV + where applicable:
• pTγ ≥ 20 GeV,|ηγ | ≤ 2.5 or pTγ ≥ 15 GeV,|ηγ | ≤ 2.7
• pTj,l,b ≥ 15 GeV,|ηj,l,b| ≤ 2.7,∆R(l,j or j,j or b,b orγ,j or γ, γ) ≥ 0.3
where pT refers to the transverse momentum of the particle,η its rapidity and ∆R =√(∆η2 + ∆φ2) whereφ is the azimuthal angle.
146
LO Monte Carlo study [LH2005, hep-ph/0604120]
inclusive σ(`νjjbbγγ) dominated by Type 2+3
# of events/100 fb−1 for ttH(γγ) selection with stringentγ-hardness+isolation cuts and |mγγ −mH | < 1.5 GeV
mH [GeV] 120 130 140
ttH(γγ) 13.5 13.1 9.5
ttγγ Type 1 0.38 0.48 0.53
ttγγ Type 2 0.5 0.3 0.5
ttγγ Type 3 0.5 <0.5 <0.5
⇒ radiative top decays still dominant
Implications on shapes and acceptance efficiencies deserve detailed NLO studies
29 / 30
Summary and Outlook
Theory precision for ttH and backgrounds
crucial for measurements of ttH rate and coupling
main priority: multi-particle backgrounds (tt+jets, tt+b-jets, ttV+jets,. . . )
Recent theory progress
powerful automated NLO MC tools
lot of results for ttH signal (soft resummation, NLO EW, off-shell effects)
all backgrounds available at NLO+PS level but not yet optimally understood
Sophisticated MC simulations and experimental analyses
solid understanding of matching/merging/shower uncertainties non-trivial
precision MC simulations in ATLAS/CMS require theory guidance
To follow/contribute to ongoing ttH HXSWG studies
⇒ https://twiki.cern.ch/twiki/bin/view/LHCPhysics/LHCHXSWGTTH
⇒ subscribe e-group mailing list lhc-higgs-xsbr-tth
30 / 30