HERPA - ippp.dur.ac.ukkrauss/Talks/MCnet18.pdf · MCnet meeting, CERN, April 2018 F. Krauss IPPP...

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

F. Krauss IPPP

SHERPA status


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


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


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


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


loop-induced processes: gg → HH

F. Krauss IPPP

SHERPA status


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


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


example applications

F. Krauss IPPP

SHERPA status


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


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


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


on-the-fly reweighting

F. Krauss IPPP

SHERPA status


reweighting on-the-fly: NLOPS

110 variations in pp →W at MC@NLO

two hardest emissions included in reweighting

F. Krauss IPPP

SHERPA status


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


implementing DGLAP @ NLO

F. Krauss IPPP

SHERPA status


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


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


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


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


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


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


parton shower accuracy

F. Krauss IPPP

SHERPA status


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


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


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


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


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

HERPA - ippp.dur.ac.ukkrauss/Talks/MCnet18.pdf · MCnet meeting, CERN, April 2018 F. Krauss IPPP...

Documents

Transcript of HERPA - ippp.dur.ac.ukkrauss/Talks/MCnet18.pdf · MCnet meeting, CERN, April 2018 F. Krauss IPPP...