Post on 27-Oct-2019
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
SHERPA status
Frank Krauss
Institute for Particle Physics PhenomenologyDurham University
MCnet meeting, CERN, April 2018
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
EW NLO corrections
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
NLO EW subtraction in SHERPA
adapt QCD subtraction (spl. fns. and colour-/spin-correlated MEs)
replacements: αs → α, CF → Q2f , CA → 0,
replacements: αs → α, TR → Nc,f Q2f , nf TR →
∑f Nc,f Q
2f ,
replacements:Tij ·Tk
T2ij→ QijQk
Q2ij
αFF
=α
FI=
αIF
=α
II=
1
αFF
=0.
001,
αFI
=α
IF=
αII
=1
αFI
=0.
001,
αFF
=α
IF=
αII
=1
αIF
=0.
001,
αFF
=α
FI=
αII
=1
αII
=0.
001,
αFF
=α
FI=
αIF
=1
αFF
=α
FI=
αIF
=α
II=
0.00
1
bc bc bc
bc
bc bc
pp → e+e− @√
s = 13 TeV, mee > 60 GeV, no γPDF
0.16
0.17
0.18
0.19
0.2
0.21
0.22
0.23
0.24
σIRS in dependence on α-parameter
σIR
S(α
)/σ
Bor
n[%
]
αFF
=α
FI=
αIF
=α
II=
1
αFF
=0.
001,
αFI
=α
IF=
αII
=1
αFI
=0.
001,
αFF
=α
IF=
αII
=1
αIF
=0.
001,
αFF
=α
FI=
αII
=1
αII
=0.
001,
αFF
=α
FI=
αIF
=1
αFF
=α
FI=
αIF
=α
II=
0.00
1
bcbc
bc bc bc
bc
pp → tt @√
s = 13 TeV, no γPDF
3.52
3.54
3.56
3.58
3.6
3.62
3.64
σIRS in dependence on α-parameter
σIR
S(α
)/σ
Bor
n[%
]
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
inclusion of electroweak corrections in simulation
incorporate approximate electroweak corrections in MEPS@NLO
1 using electroweak Sudakov factors
Bn(Φn) ≈ Bn(Φn) ∆EW(Φn)
2 using virtual corrections and approx. integrated real corrections
Bn(Φn) ≈ Bn(Φn) + Vn,EW(Φn) + In,EW(Φn) + Bn,mix(Φn)
real QED radiation can be recovered through standard tools(parton shower, YFS resummation)
simple stand-in for proper QCD⊕EW matching and merging→ validated at fixed order, found to be reliable,→ difference . 5% for observables not driven by real radiation
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
results: pp → `−ν + jets
Sher
pa+O
pen
Lo
ops
Qcut = 20 GeV
MEPS@LOMEPS@NLO QCDMEPS@NLO QCD+EWvirtMEPS@NLO QCD+EWvirt w.o. LO mix
100
10–3
10–6
10–9
pp → ℓ−ν + 0,1,2 j @ 13 TeV
dσ
/dp T
,V[p
b/G
eV]
50 100 200 500 1000 20000
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
pT,V [GeV]
dσ
/dσ
NL
OQ
CD
Sher
pa+O
pen
Lo
ops
Qcut = 20 GeV
MEPS@LOMEPS@NLO QCDMEPS@NLO QCD+EWvirtMEPS@NLO QCD+EWvirt w.o. LO mix
100
10–3
10–6
10–9
pp → ℓ−ν + 0,1,2 j @ 13 TeV
dσ
/dp T
,j 1[p
b/G
eV]
50 100 200 500 1000 20000
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
pT,j1 [GeV]
dσ
/dσ
NL
OQ
CD
⇒ particle level events including dominant EW corrections
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
results: pp → tt
Sher
pa+O
pen
Lo
ops
+0,1 (NLO QCD) + 2,3,4 (LO)+0,1 (NLO QCD+EWvirt) + 2,3,4 (LO)+0,1 (NLO QCD+EWvirt+LOsub) + 2,3,4 (LO)10−6
10−5
10−4
10−3
10−2
10−1
1
pp → tt + 0, 1(, 2, 3, 4) jets at 13 TeV
dσ
/dp T
,t[p
bG
eV−
1 ]
50 100 300 500 1000
0.7
0.8
0.9
1.0
1.1
1.2
pT,t [GeV]
1/M
EPS
@N
LO
QC
D
b
b
b
b
b
b
b
b
Sher
pa+O
pen
Lo
ops
µ = µCKKW
b ATLAS dataPRD 93 (2016) 3, 032009MEPS@NLO QCDMEPS@NLO QCD+EWvirt
10−2
10−1
1
10 1pp → tt (+ jet) at 8 TeV
dσ
/dp T
[pb
GeV
−1 ]
b b b b b b b b
300 400 500 600 700 800 900 1000 1100 1200
0.6
0.8
1
1.2
1.4
Particle top-jet candidate pT [GeV]
The
ory
/D
ata
⇒ particle level events including dominant EW corrections
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
loop-induced processes: gg → HH
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
comparison with literature: NLO+PS
10 20 30 40 50 60 70 80
0.2
0.3
0.4
0.5
dσ
/d
pHH⊥
[fb/
GeV
]
SHERPA, Dire ShowerSHERPA, CS Shower
Powheg-Box + PythiaMadGraph5 aMC@NLO + Pythia
0 200 400 600 800 1000
1.0
1.5
2.0
2.5Ratio to fixed-order
0 200 400 600 800 1000pHH⊥ [GeV]
1.0
1.5
2.0
2.5Ratio to fixed-order
largest differences in peak region dueto differences in algorithms beyondchoice of µPS: 15%
all NLO+PS results agree withinuncertainties in tail region
MADGRAPH uncertainties larger
POWHEG: flat excess in tail knownfeature of matching method(somewhat suppressed through“damping factor” hdamp = 250 GeV)
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
comparison with literature: analytic resummation
0.0
0.1
0.2
0.3
0.4
dσ
/d
pHH⊥
[fb/
GeV
]
p p → H H√
s = 14 TeV
Full SM
SHERPA+HHGRID+OPENLOOPS
NLO+NLL [Ferrera, Pires, 2017]
MC@NLO, Dire showerMC@NLO, CS shower
101 102
pHH⊥ [GeV]
0.7
0.9
1.1
1.3
Rat
ioto
NLO
+NLL
next-to-leading log (NLL)parametrically equivalent toparton shower
uncertainties on NLO+NLL:
3% near pHH⊥ ≈ 20 GeV
10% near pHH⊥ ≈ 100 GeV
good agreement withinuncertainties
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
example applications
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
single-top production: rates
MC@NLO
µ2 = t, s for t-/s-channel (by clustering to 2→ 2), tW : mtT
2
comparison to ATLAS/CMS results:
30 40 50 60
σtot [pb]
ATLAS
Nf = 5
}Sherpa
Nf = 4
ATLAS
Nf = 5
{Sherpa
Nf = 4Sherpa+
OpenL
oo
ps
CMS
CMS
4 6 8 10 12
σfid [pb]
t-ch
.(t
)t-
ch.
(t)
ATLAS
Nf = 5
}Sherpa
Nf = 4
ATLAS
Nf = 5
{Sherpa
Nf = 4
ppt-ch.→ tj / tj @ 8 TeV
µF,R PDF αS
0 10 20 30
σtot [pb]tW
-ch.
(t+t)
s-ch
.(t
+t)
ATLAS
Sherpa
ATLAS
Sherpa
Sherpa+
OpenL
oo
ps
pps-ch.→ tj / pp→ tW @ 8 TeV
CMS
CMS
µF,R PDF αS
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
single-top production: distributions
comparison to [ATLAS 1702.02859]
10−3
10−2
10−1
dσ/dpT
[pb
GeV−
1]
Sherpa+
OpenL
oo
ps
ppt-ch.→ tj @ 8 TeV
Nf = 4
Nf = 5
ATLAS
0 100 200 300
pT,t [GeV]
0.8
1.0
1.2
rati
oto
Nf5
1
2
3
4
dσ/dy
[pb
]
Sherpa+
OpenL
oo
ps
ppt-ch.→ tj @ 8 TeV
Nf = 4
Nf = 5
ATLAS
0 1 2 3 4
|yj1 |
0.8
1.0
1.2
rati
oto
Nf5
good agreement for distributions, both with NF4 & NF5
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
interference-induced Higgs peak shift
interference with background shifts Higgs diphoton peak[Martin 1208.1533]
can infer Higgs decay width from peak shift [Dixon Li 1305.3854]
with HL-LHC 3 ab−1 ΓH/ΓSMH < 15 at 95 % CL
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
on-the-fly reweighting
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
reweighting on-the-fly: NLOPS
110 variations in pp →W at MC@NLO
two hardest emissions included in reweighting
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
reweighting on-the-fly: Qcut (recent news)
variation of NTuple samples
10−4
10−3
10−2
10−1
100
101
102
dσ/dpT
[pb
/GeV
]
Sherpa, NLO QCDpp→ e+e−
@ 14 TeV
7-point scale variations
dedicated runs
Ntuple reweighting
101
102
103
` pT [GeV]
0.5
1.0
1.5
2.0
rati
o
Qcut variations in merging runs
1j 10GeV1j 20GeV1j 5GeVPS
/GeV
) [pb
] 0
→1(Q
10/d
log
σd
1−10
1
10
210
310
Rat
io
0.2−0.1−
00.10.2
+0 jet+1 jet+2 jet
Frac
tion
0.20.40.60.8
/GeV) 0→1
(Q10
log0.5 1 1.5 2 2.5
on-the-fly reweighting as the ubiquitous & default way to generatesystematic uncertainty bands
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
implementing DGLAP @ NLO
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
towards higher logarithmic accuracy
aim: reproduce DGLAP evolution at NLOinclude all NLO splitting kernels
expand splitting kernels as
P(z , κ2) = P(0)(z , κ2) +αS
2πP(1)(z , κ2)
three categories of terms in P(1):
cusp (universal soft-enhanced correction) (already included in original showers)
corrections to 1→ 2new flavour structures (e.g. q → q′), identifed as 1→ 3
new paradigm: two independent implementations
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
physical results: e−e+ → hadrons
b bb
b
b
b
b
bb
b b b b bb
bb
bb
bb
bb
bb
bb
b
b
b
b
b
Dir
ePS
b DataNLO1/4 t ≤ µ2
R ≤ 4 tLO1/4 t ≤ µ2
R ≤ 4 t10−1
1
10 1
10 2
Differential 2-jet rate with Durham algorithm (91.2 GeV)
dσ
/d
y 23
b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b
10−4 10−3 10−2 10−1
0.6
0.8
1
1.2
1.4
yDurham23
MC
/Dat
a
bb
b
b
b
b
bb
b b b b bb
bb
b
b
b
b
b
b
b
b
b
b
b
b
b
Dir
ePS
b DataNLO1/4 t ≤ µ2
R ≤ 4 tLO1/4 t ≤ µ2
R ≤ 4 t10−2
10−1
1
10 1
10 2
10 3Differential 3-jet rate with Durham algorithm (91.2 GeV)
dσ
/d
y 34
b b b b b b b b b b b b b b b b b b b b b b b b b b b b b
10−4 10−3 10−2 10−1
0.6
0.8
1
1.2
1.4
yDurham34
MC
/Dat
a
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
physical results: DY at LHC
b
b b
b
b
b
b
b
b
b
bb
bb b b b b b b b b b b b b
Dir
ePS
b DataNLO1/4 t ≤ µ2
R ≤ 4 tLO1/4 t ≤ µ2
R ≤ 4 t
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Z → ee ”dressed”, Inclusive
1σ
fidd
σfid
dp T
[GeV
−1 ]
b b b b b b b b b b b b b b b b b b b b b b b b b b
0 5 10 15 20 25 300.80.91.01.11.21.31.4
Z pT [GeV]
MC
/Dat
a
b
b b
b
b
b
b
b
b
b
bb
bb b b b b b b b b b b b b
Dir
ePS
b DataNLO1/4 t ≤ µ2
R ≤ 4 tLO1/4 t ≤ µ2
R ≤ 4 t
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.080.0 < |yZ| < 1.0, ”dressed”
1σ
fidd
σfid
dp T
[GeV
−1 ]
b b b b b b b b b b b b b b b b b b b b b b b b b b
0 5 10 15 20 25 300.80.91.01.11.21.31.4
Z pT [GeV]
MC
/Dat
a
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
parton shower accuracy
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
how to assess formal precision?
PS proven to be NLL accurate for simple observables, providedCatani,Marchesini,Webber, NPB349(1991)635
soft double-counting removed (↗ before) and2-loop cusp anomalous dimension included
not entirely clear what this means numerically, because
parton shower is momentum conserving, NLL is notparton shower is unitary, NLL approximations break this
differences can be quantified by
designing an MC that reproduces NLL exactlyremoving NLL approximations one-by-one
employ well-established NLL result as an example
observable: Thrust in e+e− →hadronsmethod: Caesar Banfi,Salam,Zanderighi, hep-ph/0407286
this discussion is technical, but it is needed to showthat equivalence at NLL does not mean identical numerics
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
differences between pure NLL and parton shower
Hoeche, Reichelt, Siegert, arXiv:1711.03497
isolated differences in terms of resolved/unresolved splittingprobability:
R’≶v (ξ) =α≶v,soft
s
(µ2≶
)π
∫ zmax≶v,soft
zmin dz CF
1−z −α≶v,coll
s
(µ2≶v
)π
∫ zmax≶v,coll
zmin dz CF1+z
2
NLL Parton Shower NLL Parton Shower
zmax>v ,soft 1− (ξ/Q2)
a+b2a zmax
>v ,coll 1 1− (ξ/Q2)a+b2a
µ2>v ,soft ξ(1− z)
2ba+b µ2
>v ,coll ξ ξ(1− z)2ba+b
α>v ,softs 2-loop CMW α>v ,coll
s 1-loop 2-loop CMW
zmax<v ,soft 1− v
1a 1− (ξ/Q2)
a+b2a zmax
<v ,coll 0 1− (ξ/Q2)a+b2a
µ2<v ,soft Q2v
2a+b (1− z)
2ba+b ξ(1− z)
2ba+b µ2
<v ,coll n.a. ξ(1− z)2ba+b
α<v ,softs 1-loop 2-loop CMW α<v ,coll
s n.a. 2-loop CMW
can cast pure NLL into PS language by using NLL expressions in PS
can study each effect in detail by reverting changes back to PS
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
Local momentum conservation and unitarity
single emission
Analytic NLL ǫ → 0Shower ǫ = 0.001z(1 − z) > k2
T/Q2
same plus µ2 = k2T
Shower ǫ = 0.001z(1 − z) > k2
T/Q2, η > 0same plus µ2 = k2
T
1
P(1
−T
<v)
-2 -1.5 -1 -0.5 0
0.850.9
0.951.0
1.051.1
log10(v)
Rat
io
NLL→PS in zmin/max
(4-momentum conservation)
NLL→PS in zcoll>v ,max
(phase-space sectorization)
NLL→PS in µ2>v ,coll
(conventional)
Analytic NLL ǫ → 0Shower ǫ = 0.001zmax<v,soft = 1 − (ξ/Q2)
a+b2a
Shower ǫ = 0.001µ2
<v,soft = k2T10−1
1
P(1
−T
<v)
-2 -1.5 -1 -0.5 0
0.60.70.80.91.0
log10(v)
Rat
io
NLL→PS in zsoft<v ,max
(from PS unitarity)
NLL→PS in µ2<v ,soft
(from PS unitarity)
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
Running coupling and global momentum conservation
Analytic NLL ǫ → 0Shower ǫ = 0.0012-loop (< v, soft)2-loop CMW (< v, soft)Shower ǫ = 0.0012-loop CMW
10−1
1
P(1
−T
<v)
-2 -1.5 -1 -0.5 0
0.60.70.80.91.0
log10(v)
Rat
io
NLL→PS in 2-loop CMW< v , soft(from PS unitarity)
NLL→PS in 2-loop CMWoverall(conventional)
Analytic NLL → 0unitary Shower ǫ = 0.001z(1 − z) > k2
T/Q2, η > 0v from 4-momentasoft kT & z definition
10−1
1
P(1
−T
<v)
-2 -1.5 -1 -0.5 0
0.60.70.80.91.0
log10(v)
Rat
io
NLL→PS in observable(use experimental definition)
NLL→PS in evolution variable
F. Krauss IPPP
SHERPA status
EW NLO corrections Loop-induced processes example applications reweighting NLO in DGLAP evolution Parton shower accuracy
Overall comparison NLL / PS / Dipole ShowerAnalytic
Resummation Shower
DGLAP Shower
CS Shower
-2
0
2
4
6
8
10
12
π αS〈n
em
〉
-10 -8 -6 -4 -2 00.9
0.95
1.0
1.05
lnk2t
Q2
Ratio
DGLAP kT orderedDGLAP ξ1−T orderedDipole kT orderedincl. gluon splittingDipole kT orderedNLL
10−1
1
P(1
−T
<v)
-2 -1.5 -1 -0.5 00.60.70.80.91.01.11.2
log10(v)
Rat
ioTuned comparison of differences between formally equivalentcalculations
Simplest process and simplest observable, but still large differences
Origin of differences traced to treatment of kinematics & unitarity
At NLL accuracy, none of the methods is formally superior→ Difference is a systematic uncertainty & needs to be kept in mind
F. Krauss IPPP
SHERPA status