Status of the Sherpa Event Generator · 2018. 11. 22. · Status of the Sherpa Event Generator Stefan H oche SLAC National Accelerator Laboratory Multi-Boson Interactions Workshop

Status of the Sherpa Event Generator

Stefan Hoche

SLAC National Accelerator Laboratory

Multi-Boson Interactions Workshop

Ann Arbor, 08/28/2018

The Sherpa Framework

[Gleisberg et al.] arXiv:0811.4622I Hard interactionLO, NLO QCD/EW 1, NNLO QCD 2

ME generators Amegic & Comix

I Radiative correctionsCatani-Seymour based PS,Dire, YFS QED resummation

I Multiple interactionsSjostrand-Zijl model

I HadronizationCluster hadronization model

I Hadron DecaysPhase space or EFTs,YFS QED corrections

Talk covers MBI related items and other news

1via interfaces to loop generators2pp→ Z/W±/h/W±W∓

1

Vector-boson scattering simulations with Sherpa

I VBF-like situations require judicious setting of color flow informationin interface between fixed-order calculation and parton shower

Correct Incorrect

I Current color selection in Sherpa based on hardcoded probabilitiesfor the most relevant processes, VBF topologies are not included

I Alternative, generic option in future version 3.0.0I Idenitify all possible color flows in core interaction

(after ME+PS clustering, e.g. pp→ e+e− in pp→ e+e−+jets)I Compute corresonding partial amplitudes [Gleisberg,SH] arXiv:0808.3674

I Select winner topology probabilistically

I Sherpa 3.0.0 also allows to specify different starting scalesfor parton-shower evolution of disconnected dipoles

2

Example: W+W+ production, e+µ+ channel

I Test new setup in simplest vector boson scattering scenario→ same-sign WW production (simplified to e+µ+ channel)

I Rivet analysis courtesy of Stefanie Todt (TU Dresden)I Two charged dressed leptons (same sign) pT > 27 GeV, |η| < 2.5I At least 2 jets with pT > 30 GeV, |η| < 4.5I Lepton isolation ∆Rll > 0.3, ∆Rlj > 0.3I Lepton invariant mass mll > 20 GeVI Missing transverse momentum pT,miss > 30 GeVI Tagjet (lead-pT ) invariant mass mjj > 500 GeVI Tagjet rapidity difference ∆yjj > 2

before cuts

3.0.0 PS2.2.2 PS

0 1 2 3 4 5

10−1

1

nj

dσ/

dnj

[fb]

after cuts

3.0.0 PS2.2.2 PS

0 1 2 3 4 5

10−2

10−1

nj

dσ/

dnj

[fb]

3

Example: W+W+ production, e+µ+ channel

3.0.0 PS3.0.0 MEPS 1j2.2.2 PS2.2.2 MEPS 1j

10−3

10−2dσ/

dy∗ j3

[fb]

-4 -3 -2 -1 0 1 2 3 40.5

1

1.5

2

2.5

y∗j3

Rat

ioto

v3PS

0 < |yj3| < 1

v3.0.0 PSv3.0.0 MEPS 1jv2.2.2 PSv2.2.2 MEPS 1j

10−5

10−4

10−3

dσ/

dpT

,j3[f

b/G

eV]

40 60 80 100 120 140 160 180 2000.60.8

11.21.41.61.8

2

pT,j3 [GeV]

Rat

ioto

v3PS

I Differential distributions confirm expectation:I Third jet produced more centrally and at higher rate in Sherpa 2.2.2I PS radiation pattern in Sherpa 2.2.0 corrected by ME+PS merging,

but breaking of PS unitarity in CKKW(L) decreases overall event rate

I Sherpa 3.0.0 predicts ∼20% larger cross section after cutsas a result of correct color flow and PS starting scales

4

Example: W+W+ production, e+µ+ channel

3.0.0 PS3.0.0 MEPS 1j2.2.2 PS2.2.2 MEPS 1j

10−4

dσ/

dmjj

[fb/

GeV

]

600 800 1000 1200 1400 1600 1800

0.50.60.70.80.91.01.11.2

mjj [GeV]

Rat

ioto

v3PS

3.0.0 PS3.0.0 MEPS 1j2.2.2 PS2.2.2 MEPS 1j

10−3

10−2

10−1

dσ/

dyjj

[fb]

2 3 4 5 6 7 8

0.50.60.70.80.91.01.11.2

yjj

Rat

ioto

v3PS

I Differential distributions confirm expectation:I Third jet produced more centrally and at higher rate in Sherpa 2.2.2I PS radiation pattern in Sherpa 2.2.0 corrected by ME+PS merging,

but breaking of PS unitarity in CKKW(L) decreases overall event rate

I Sherpa 3.0.0 predicts ∼20% larger cross section after cutsas a result of correct color flow and PS starting scales

4

Automated electroweak subtraction

[Schonherr] arXiv:1712.07975

I Extension of QCD subtraction formalism to QCD+EW caseI Automation of color / charge correlator insertionsI Automated handling of mixed QCD/EW Born processesI Introduction of photon splittings in dipole formalismI Introduction of subtraction terms for massive bosons

reco

ilsc

hem

e0

reco

ilsc

hem

e1

reco

ilsc

hem

e2

reco

ilsc

hem

e3

reco

ilsc

hem

e4

√s = 13 TeV

mee > 2 TeV, p⊥,e > 20 GeVγq/γq channel

bc bc bc bcbc

ld ld ld ld

ld

bc CT14qed (pγ0 = 0%)

ld CT14qed (pγ0 = 0.14%)

0.098

0.1

0.102

0.104

0.136

0.138

0.140

0.142

pp → e+e− @ O(α3)

σIR

D(r

ecoi

lsch

eme)

/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

√s = 13 TeV

ut ut utut

ut ut

bcbc bc

bcbc bc

ldld

ldld

ld ld

ut CT14 bc CT14qed (pγ0 = 0%)

ld CT14qed (pγ0 = 0.14%)

-0.312-0.31

-0.308

-0.246-0.244-0.242

-0.23-0.228-0.226

pp → tt @ O(α2s α)

I need a lot of white text here, more even, more, more

σIR

D({

αd

ip})

/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

√s = 13 TeV

ut subtraction as fermion

u subtraction as scalar

u u uu

u u

b b b b bb

l l l l ll

ut ut utut

ut ut

bc bc bc bc bcbc

ld ld ldld ld

ld

ut CT14bc CT14qed (pγ

0 = 0%)ld CT14qed (pγ

0 = 0.14%)0.140.13

0.920.93

1.211.221.23

pp → W+W− @ O(α3)

σIR

D({

αd

ip})

/σ

Bor

n[%

]

5

Electroweak corrections and multi-jet merging

I Incorporate approximate electroweak corrections in MEPS@NLOI Electroweak Sudakov factors [Thompson] PhD Thesis

Bn → Bn∆EW

I Virtual corrections and integrated NLO subtraction terms[Kallweit,Lindert,Maierhofer,Pozzorini,Schonherr] arXiv:1511.08692

Bn → Bn + Vn,EW + In,EW + Bn,mix

I Real-emission corrections can be recovered to a good accuracyby using standard tools (parton showers, YFS resummation)

I Validated in comparison to fixed-order calculationsDifference .5% in observables not dominated by real radiation

6

Electroweak corrections and multi-jet merging

[Kallweit,Lindert,Maierhofer,Pozzorini,Schonherr] arXiv:1511.08692

[Gutschow,Lindert,Schonherr] arXiv:1803.00950

MEPS@LO

MEPS@NLO QCD

MEPS@NLO QCD+EWvirt

MEPS@NLO QCD+EWvirt w.o. LO mix

100

10–3

10–6

10–9

pp → ℓ−ν + 0,1,2 j @ 13 TeV

dσ/dpT,V

[pb/GeV

]

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/d

σNLO

QCD

I W− production at 13 TeV LHC

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

I tt production at 8 TeV LHC

7

NLO+PS matching for loop-induced processes

[Jones,Kuttimalai] arXiv:1711.03319Loop-Induced Processes at NLO

General Treatment

Born and real corrections: Automated one-loop tools

IR-subtraction: standard techniques: Catani-Seymour, FKS, . . .

Parton shower matching: standard techniques: MC@NLO, Powheg, . . .

Availability of two-loop virtual amplitudes

gg → γγ [Bern et al.: hep-ph/0109078]

gg → VV → llll [Gehrmann et al.: 1503.04812, Manteuffel et al.: 1503.08835]

gg → HH [Borowka et al.: 1608.04798]

gg → Hj [Jones et al.: 1802.00349]2

I Automated part of simulationI Born and real corrections: Automated one-loop tools, e.g.

OpenLoops [Cascioli et al.] arXiv:1111.5206, MadLoop [Hirschi at al.] arXiv:1103.0621

I IR-subtraction: Catani-Seymour (new, dedicated implementation)I Parton shower matching: S-MC@NLO (new, dedicated implementation)

I Process-dependent part: Two-loop virtual amplitudes

Implemented and validated forI gg → HH [Borowka et al.] arXiv:1608.04798

Further interesting processesI gg → γγ [Bern et al.] hep-ph/0109078

I gg → V V → 4l [Gehrmann et al.] arXiv:1503.04812, [Manteuffel et al.] arXiv:1503.08835

I gg → Hj [Jones et al.] arXiv:1802.00349

8

NLO+PS matching for loop-induced processes

[Jones,Kuttimalai] arXiv:1711.03319

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

I Comparison to MC@NLO & POWHEG[Heinrich et al] arXiv:1703.09252

I Comparison to analytic resummation[Ferrera,Pires] arXiv:1609.01691

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

9

Towards an NLO parton shower

[Jadach,Skrzypek] hep-ph/0312355

I DGLAP equation for fragmentation functions

dxDa(x, t)

d ln t=∑b=q,g

∫ 1

0dτ

∫ 1

0dz

αs

2π

[zPab(z)

]+τDb(τ, t) δ(x− τz)

I Refine plus prescription[zPab(z)

]+

= limε→0

zPab(z, ε) & rewrite for finite ε

Pab(z, ε) =Pab(z) Θ(1− ε− z)− δab∑

c∈{q,g}

Θ(z − 1 + ε)

ε

∫ 1−ε

0dζ ζ Pac(ζ)

d lnDa(x, t)

d ln t=−

∑c=q,g

1−ε∫0

dζ ζαs

2πPac(ζ) +

∑b=q,g

1−ε∫x

dz

z

αs

2πPab(z)

Db(x/z, t)

Da(x, t)

I First term defined as derivative of Sudakov factor

∆a(t, Q2) = exp

{−∫ Q2

t

dt

t

∑c=q,g

∫ 1−ε

0dζ ζ

αs

2πPac(ζ)

}

I DGLAP evolution is recovered from parton shower as ε→ 0

10

Towards an NLO parton shower

[Curci,Furmanski,Petronzio] NPB175(1980)27, [Floratos,Kounnas,Lacaze] NPB192(1981)417

I Higher-order DGLAP evolution kernels obtained from factorization

D(0)ji (z, µ) = δijδ(1− z) ↔

jz

/i

1

D(1)ji (z, µ) =−

1

εP

(0)ji (z) ↔

izj

/i

1

D(2)ji (z, µ) =−

1

2εP

(1)ji (z) +

β0

4ε2P

(0)ji (z) +

1

2ε2

∫ 1

z

dx

xP

(0)jk (x)P

(0)ki (z/x)

↔(

izj

+i

zj

) /i

1

I P(n)ji not probabilities, but sum rules hold (↔ unitarity constraint)

In particular: Momentum sum rule identical between LO & NLO

I PS implements NLO DGLAP equation if Sudakov factor definedas on previous slide (used in [Krauss,Prestel,SH] arXiv:1705.00982)

I Combination with soft evolution problematic [Dasgupta et al] arXiv:1805.09327

11

Fully differential collinear evolution at NLO

[Prestel,SH] arXiv:1705.00742

I Simulation of exclusive states requires computing splitting functionson the fly using differential NLO calculation & collinear factorization

I Schematically very similar to Catani-Seymour (CS) dipole subtractionI Simplest example: Flavor-changing configuration q → q′

za

zizj

saisaij

q q′

Tree-level expression ↔ real-emission correction in CSSubtraction term (q→g)⊗(g→q’) ↔ differential subtraction term in CS

I Complete NLO result schematically given by

P(1)qq′ (z) =

[Cqq′ (z) + Iqq′ (z)

]+

∫dΦ+1

[Rqq′ (z,Φ+1)− Sqq′ (z,Φ+1)

]Both components finite in 4 dimensions → amenable to MC simulation

I Analytical computation of I not needed, as C + I finitegenerate as endpoint at sai = 0, starting from integrand at O(ε)

12

Leading color fully differential soft evolution at NLO

I Double-soft real-emission corrections given byı j

˜12i j

1 2

S(gg)ij (1, 2) = S(s.o.)ij (1, 2)CA

2

(1 +

si1sj1 + si2sj2

(si1 + si2)(sj1 + sj2)

)+

sij

(si1 + si2)(sj1 + sj2)

CA

s12

(− 2 + 4 (1− ε) z1z2 cos2 φ12,ij

)I Strongly ordered and spin correlation terms

S(s.o.)ij (1, 2) =sij

si1s12sj2+

sij

sj1s12si2−

s2ij

si1sj1si2sj2

cosφij12 = k⊥ jij,⊥(p1 + p2) where jµij,⊥(q) =(pjq) p

µi − (piq) p

µj√

2 (pipj)(piq)(pjq)

I Simplest way to implement → reweighting of LO parton showerI Account for finite recoil in emission probabilitiesI Account for first sub-leading color effectsI Account for spin correlations

I Subtracted real corrections remain non-zero in regions of phase spacenot covered by ordered emissions (according to parton-shower evolution)

13

Leading color fully differential soft evolution at NLO

[Dulat,Prestel,SH] arXiv:1805.03757

Dir

ePS

LO⊕NLO softκ/2 ≤ µ ≤ 2κ

LOκ/2 ≤ µ ≤ 2κ

LO no CMW5

10

15

20

25

dσ

/d

log 10

(y23

)[n

b]

-4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5

0.6

0.8

1

1.2

1.4

log10(y23)

Rat

io

Dir

ePS

LO⊕NLO softκ/2 ≤ µ ≤ 2κ

LOκ/2 ≤ µ ≤ 2κ

LO no CMW

5

10

15

20

25

30

35

dσ

/d

log 10

(y34

)[n

b]

-4 -3.5 -3 -2.5 -2 -1.5 -1

0.6

0.8

1

1.2

1.4

log10(y34)R

atio

I Impact on 2→ 3 and 3→ 4 Durham jet rate at LEP I

I Uncertainty bands no longer just estimatesbut perturbative QCD predictions for the first time

I Fair agreement with CMW scheme

14

Summary

Physics performance improvementsI New technique in Sherpa 3.0.0 for

I Selection of large-Nc color flowI Selection of parton-shower starting scales

Tested in vector boson scattering

Other new features

I Electroweak subtraction

I Approximate EW merging

I Loop-induced processes

I Partial NLO shower

15

Thank you for your attention