monte carlo event generators for the lhc - uni-mainz.de
TRANSCRIPT
Monte Carlo Event Generators for the LHCor How to relate Theory with Experiment?
Steffen Schumann
ITP, University of Heidelberg
EMG Annual Retreat 2010
Bingen am Rhein
27. - 29.09. 2010
Introduction & Monte Carlo Techniques
Hard Processes at (Next-to-)Leading Order
Parton Showers & Matching with Fixed Order
Multiple Interactions, Hadronization & Hadron Decays
Steffen Schumann Monte Carlo Event Generators for the LHC
Outline Lecture 2
Parton Showers & Matching with Fixed Order
The QCD parton-shower pictureMerging parton showers with multi-leg tree-level amplitudesMatching to One-Loop computations
Multiple Interactions, Hadronization & Hadron Decays
Remnant-Remnant interactionsThe cluster hadronization model
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Showers & Matching with Fixed Order
Steffen Schumann Monte Carlo Event Generators for the LHC
Recap: The Hard Process
The Hard Process
σpp→Xn =∑a,b
∫dxadxb fa(xa, µ
2F )fb(xb, µ
2F ) |Mab→Xn |2 dΦn
+ |Mab→Xn |2 fundamental physics, interferences, off-shell effects, full colour
+ accounts for high-pT , well separated partons
− few-parton final states only, poor for log-enhanced phase-space regions
'
&
$
%|Mqq→e+e−g|2
γ∗/Z0
0 25 50 75 100 125 150 175 200pT [GeV]
10-4
10-3
10-2
10-1
100
101
dσ/d
p T [p
b/G
eV]
CDF 2000e
+e
-g (scaled)
pT(e+e
-) @ Tevatron Run1
0 5 10 15 20pT [GeV]
5
10
15
20
25
30
dσ/d
p T [p
b/G
eV]
∣∣∣∣Mqq→e+e−g
∣∣∣∣2 ∼ ∣∣∣∣Mqq→e+e−∣∣∣∣2 αS (µ2
R )
p2T
σpp→e+e−g
∼ σpp→e+e−αS (µ2
R ) logpmaxT
pminT
|M|2 factorise in IR limes (universal) large logs need to be resummed to all orders
Steffen Schumann Monte Carlo Event Generators for the LHC
Recap: The Hard Process
The Hard Process
σpp→Xn =∑a,b
∫dxadxb fa(xa, µ
2F )fb(xb, µ
2F ) |Mab→Xn |2 dΦn
+ |Mab→Xn |2 fundamental physics, interferences, off-shell effects, full colour
+ accounts for high-pT , well separated partons
− few-parton final states only, poor for log-enhanced phase-space regions
'
&
$
%|Mqq→e+e−g|2
γ∗/Z0
0 25 50 75 100 125 150 175 200pT [GeV]
10-4
10-3
10-2
10-1
100
101
dσ/d
p T [p
b/G
eV]
CDF 2000e
+e
-g (scaled)
pT(e+e
-) @ Tevatron Run1
0 5 10 15 20pT [GeV]
5
10
15
20
25
30
dσ/d
p T [p
b/G
eV]
∣∣∣∣Mqq→e+e−g
∣∣∣∣2 ∼ ∣∣∣∣Mqq→e+e−∣∣∣∣2 αS (µ2
R )
p2T
σpp→e+e−g
∼ σpp→e+e−αS (µ2
R ) logpmaxT
pminT
|M|2 factorise in IR limes (universal) large logs need to be resummed to all orders
Steffen Schumann Monte Carlo Event Generators for the LHC
QCD Bremsstrahlung
accelerated charges radiate
QED: electrons (charged) emit photons
QCD: quarks (coloured) emit gluons
but, gluons coloured as well gluons emit gluons
QCD radiation enhanced in the infra-red: soft / collinear emissions
real-emission matrix elements factorize in collinear limit [universal]
|t| = |p2b|, z = Eb/Ea t = p2
a , z = Eb/Ea
dσn+1 = dσndt
tdzαS
2πPba(z)
iteration / Markov process
parton shower MC
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Shower: Toy Model
one particle species G only, starting scale t = tmax
GG
G
∝ PGG (z)dPG = PGG (z)
dt
tdz
given t→ I (t) =
∫ z+(t)
z−(t)
dzPGG (z)
probability of no-branching between tmax and t < tmax
PG ,no−branch(t, tmax) = exp
−tmax∫t
dt ′
t ′I (t ′)
→ ordering parameter t
probability for branching to occur at t < tmax
dPG ,branch
dt= −dPG ,no−branch
dt= I (t) exp
−tmax∫t
dt ′
t ′I (t ′)
lends itself to simulation, parton cascade
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Shower: Toy Model
a simple shower algorithm
determine scale of next emission by solving
# = exp
−tmax∫t
dt ′
t ′I (t ′)
for t
select energy fraction z according to PGG (z)
construct kinematics of emitted particle
reset tmax = t and start afresh
The full QCD picture
Pqq, Pgq, Pgq, Pgg , αS(z , t), choice of evolution variable [coherence effects]
shower has to stop at some infra-red cut-off to ∼ O(1GeV2)
below perturbative approach no-longer applies
invoke hadronization model
Steffen Schumann Monte Carlo Event Generators for the LHC
The QCD Parton Shower picture
Shower evolution as a probabilistic processconstruct explicitely the initial- & final-state partons history/fatesuccessive branching of incoming and outgoing legsencoded in QCD evolution of PDFs and Fragmentation Functions exclusive partonic final states with Ptot = Phard · PIS · PFS
What are Parton Showers good for?evolve parton ensemble from high- to low scale t0 ∼ O(1GeV2) link the hard process to universal hadronization models
in turn resummation of large kinematical logarithms [to (N)LL accuracy]
model intra-jet energy flows: jets become multi-parton objects
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Shower in real QCDQCD evolution of PDFs (Fragmentation Functions)
∂fa(z, t)
∂ log(t/t0)=
∫ ζmax
z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)− fa(z, t)
∫ ζmax
ζmin
dζ∑b=q,g
Kba(ζ, t)
Kba(ζ, t) - evolution kernels of the scheme IR factorization scheme e.g. αs/2π Pba(ζ)
Kba(ζ, t)IR→ 1
σ(n)a (Φn)
dσ(n+1)b (ζ, t; Φn)
d log(t/t0) dζ
ζ, t - splitting, evolution variable separate resolved from unresolved emissions
∆a(t0, t) = exp
−∫ t
t0
dt
t
∫ ζmax
ζmin
dζ∑b=q,g
Kba(ζ, t)
∂
∂ log(t/t0)
fa(z, t)
∆a(t0, t)=
1
∆a(t0, t)
∫ ζmax
z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)
fa(z, t) = ∆a(t0, t)fa(z, t0) +
t∫t0
dt
t
∆(t0, t)
∆(t0, t)
ζmax∫z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)
∆(t0, t) probability for evolving from t0 to t without branching [Sudakov Form Factor]
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Shower in real QCDQCD evolution of PDFs (Fragmentation Functions)
∂fa(z, t)
∂ log(t/t0)=
∫ ζmax
z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)− fa(z, t)
∫ ζmax
ζmin
dζ∑b=q,g
Kba(ζ, t)
Kba(ζ, t) - evolution kernels of the scheme IR factorization scheme e.g. αs/2π Pba(ζ)
Kba(ζ, t)IR→ 1
σ(n)a (Φn)
dσ(n+1)b (ζ, t; Φn)
d log(t/t0) dζ
ζ, t - splitting, evolution variable separate resolved from unresolved emissions
∆a(t0, t) = exp
−∫ t
t0
dt
t
∫ ζmax
ζmin
dζ∑b=q,g
Kba(ζ, t)
∂
∂ log(t/t0)
fa(z, t)
∆a(t0, t)=
1
∆a(t0, t)
∫ ζmax
z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)
fa(z, t) = ∆a(t0, t)fa(z, t0) +
t∫t0
dt
t
∆(t0, t)
∆(t0, t)
ζmax∫z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)
∆(t0, t) probability for evolving from t0 to t without branching [Sudakov Form Factor]
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Shower in real QCDQCD evolution of PDFs (Fragmentation Functions)
∂fa(z, t)
∂ log(t/t0)=
∫ ζmax
z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)− fa(z, t)
∫ ζmax
ζmin
dζ∑b=q,g
Kba(ζ, t)
Kba(ζ, t) - evolution kernels of the scheme IR factorization scheme e.g. αs/2π Pba(ζ)
Kba(ζ, t)IR→ 1
σ(n)a (Φn)
dσ(n+1)b (ζ, t; Φn)
d log(t/t0) dζ
ζ, t - splitting, evolution variable separate resolved from unresolved emissions
∆a(t0, t) = exp
−∫ t
t0
dt
t
∫ ζmax
ζmin
dζ∑b=q,g
Kba(ζ, t)
∂
∂ log(t/t0)
fa(z, t)
∆a(t0, t)=
1
∆a(t0, t)
∫ ζmax
z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)
fa(z, t) = ∆a(t0, t)fa(z, t0) +
t∫t0
dt
t
∆(t0, t)
∆(t0, t)
ζmax∫z
dζ
ζ
∑b=q,g
Kba(ζ, t) fb(z
ζ, t)
∆(t0, t) probability for evolving from t0 to t without branching [Sudakov Form Factor]
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Shower in real QCD
Initial- and Final-State Parton Showers
Starting from dσ, generate more radiation
P(IS)no, a(z , t, t′) =
∆a(t0, t′) fa(z , t)
∆a(t0, t) fa(z , t′)& P(FS)
no, a(t, t′) =∆a(t0, t
′)
∆a(t0, t)
+ accounts for multiple soft/collinear emissions resummation of large logs
+ exclusive radiation pattern/ hard cross section preserved
− lacks hard/large angle emissions
− leading-colour contributions only
0 25 50 75 100 125 150 175 200pT [GeV]
10-4
10-3
10-2
10-1
100
101
dσ/d
p T [p
b/G
eV]
CDF 2000e
+e
-g (scaled)
pT(e+e
-) @ Tevatron Run1
0 5 10 15 20pT [GeV]
5
10
15
20
25
30
dσ/d
p T [p
b/G
eV]
0 25 50 75 100 125 150 175 200p
T [GeV]
10-4
10-3
10-2
10-1
100
101
dσ/d
p T [
pb/G
eV]
CDF 2000CS show. + Py 6.2 had.CS show. + Py 6.2 had. (enhanced start scale)
0 5 10 15 20p
T [GeV]
5
10
15
20
25
30
dσ/d
p T [
pb/G
eV]
Steffen Schumann Monte Carlo Event Generators for the LHC
Subtraction formalism based ShowersCatani–Seymour local subtraction term∫
m+1
dσA =∑
dipoles
∫m
dσB ⊗∫
1
dVdipole
→ universal dipole term←spin- & color correlation
Ansatz: Complete factorisation through
projection onto leading term in 1/Nc
spin averaged dipole terms Vdipole → 〈Vdipole〉
Shower Algorithm
color connected emitter–spectator ’dipoles’
subsequent branchings of type II, IF, FI, FF
exact momentum mappings invertable
emissions ordered in k2⊥
i j k
m-parton LO-ME Splitting operator
4 2e
m-parton state splitting operator
pi pk
(pi pj )(pk pj )=
pi pk
(pi pj )(pi + pk )pj
+pi pk
(pk pj )(pi + pk )pj
Dipole subtraction based: Dinsdale et al. Phys. Rev. D 76 (2007) 094003, Krauss, S. JHEP 0803 (2008) 038
Antenna subtraction based: Krauss, Winter JHEP 0807 (2008) 040, Giele et al. Phys. Rev. D 78 (2008) 014026
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Showers at work
Dijet azimuthal decorrelation [Krauss, S. JHEP 0803 (2008) 038, DØ data Phys. Rev. Lett. 94 (2005) 221801]
∆φ dijet (rad)
1/σ
dije
t dσ
dije
t / d∆φ
dije
t
pT max > 180 GeV (×8000)
130 < pT max < 180 GeV (×400)
100 < pT max < 130 GeV (×20)
75 < pT max < 100 GeV
LO
NLO
NLOJET++ (CTEQ6.1M)
µr = µf = 0.5 pT max
DØ
10-3
10-2
10-1
1
10
102
103
104
105
π/2 3π/4 π
only the two leading jets need to be reconstructed strong test of the initial- and final-state radiation pattern
Steffen Schumann Monte Carlo Event Generators for the LHC
Parton Showers at work
Dijet azimuthal decorrelation [Krauss, S. JHEP 0803 (2008) 038, DØ data Phys. Rev. Lett. 94 (2005) 221801]
π/2 3π/4 π
∆φdijet
(rad)
10-3
10-2
10-1
100
101
102
103
104
105
1/σ
dij
et dσ
dij
et/d∆φ
dij
et
75 < pTmax
< 100 GeV
100 < pTmax
< 130 GeV (x20)
130 < pTmax
< 180 GeV (x400)
pTmax
> 180 GeV (x8000)
∆φdijet
distribution @ Tevatron Run II
points: D0 data 2005
histo: CS show. + Py 6.2 had.
only the two leading jets need to be reconstructed strong test of the initial- and final-state radiation pattern
Steffen Schumann Monte Carlo Event Generators for the LHC
Combining ME & PS
Problem: QCD MEs and PS deal with the same physics!double counting of phase-space configurations
unpopulated phase-space regions
Aim: Consistent description of real QCD emissions!proper description of soft/collinear and hard emissions
combine QCD matrix elements of different parton multiplicity with showers
[CKKW: Catani et al. ’01, MLM: Mangano et al. ’01, CKKW-L: Lonnblad ’01]
Construction criteria:describe few hardest emissions through full matrix elements
Kba(z, t) →1
dσ(n)a (Φn)
dσ(n+1)b
(z, t; Φn)
d log(t/t0) dz
preserve shower-evolution equation i.e. logarithmic accuracy
avoid double counting or empty phase-space regions
slice emission phase space by parton-separation criterion Qba(z, t)
Steffen Schumann Monte Carlo Event Generators for the LHC
Combining ME & PS: sneak preview
Drell-Yan pT distribution [Krauss,Schalicke, S., Soff Phys. Rev. D 70 (2004) 114009]
/ GeV Z
P0 20 40 60 80 100 120 140 160 180 200
10-3
10-2
10-1
1
10pt Z
Z + 0 jet
Z + 1 jet
Z + 2 jet
Z + 3 jet
CDF
Ge
Vp
b
/
dPσ
d
/ GeV Z
P0 5 10 15 20 25 30 35 40 45 50
G
eV
pb
/
d
Pσd
1
10
pt Z
Z + 0 jet
Z + 1 jet
Z + 2 jet
Z + 3 jet
CDF
Steffen Schumann Monte Carlo Event Generators for the LHC
Solution part 1: Slicing the phase space
Phase-space separation
KPSba (z , t) = Kba(z , t) Θ
[Qcut − Qba(z , t)
]← shower regime
KMEba (z , t) = Kba(z , t) Θ
[Qba(z , t)− Qcut
]← matrix-element regime
⇒ Qba(z , t) has to identify logarithmically enhanced phase-space regions
Consequences
Sudakov form factor and shower no-branch probabilities factorize
∆a(t0, t) = ∆PSa (t0, t) ∆ME
a (t0, t)
P(IS)no, a(z, t, t′) = P(IS) PS
no, a (z, t, t′) P(IS) MEno, a (t, t′) =
∆PSa (t0, t
′) fa(z, t)
∆PSa (t0, t) fa(z, t′)
∆MEa (t0, t
′)
∆MEa (t0, t)
need to constrain shower emissions to Q < Qcut
matrix elements need to be reweighted [made exclusive quantities]
→ think of ME’s as predetermined shower emissions, truncated shower
Steffen Schumann Monte Carlo Event Generators for the LHC
Solution part 2: Defining PS histories
Interpret ME as if produced by PS
Identify most likely splittingacc. to PS branching probability
Combine partons into mother partonacc. to inverse PS kinematics
Continue until 2→ 2 core process
shower specific cluster algorithm
predetermined shower emissions
PS starts at core processcan radiate “between” ME emissions
ME branchings must be respectedevolution-, splitting- & angular variable preserved
truncated shower
Example branching history
NNLO
cluster once
find some t
NLO
t
cluster twice
find some t′
LOt′
t
Steffen Schumann Monte Carlo Event Generators for the LHC
Solution part 3: Truncated shower
Assume ME splittings at t and t ′ > t
Shower emission below Qcut
t′Q > Qcut
tQ > Qcut
tQ < Qcut
> >
emission accepted
large-angle soft emissions
soft color coherence
approx. in CKKW only
Shower emission above Qcut
↔ N3LOLO
t′
t entire event is rejected
Sudakov suppression PMEno, a(t, t ′)
to be described by ME instead
σtot preserved at LO
Steffen Schumann Monte Carlo Event Generators for the LHC
Matrix Elements and Parton Showers: ME⊕PS
How to attach shower to an N-parton ME?
The emerging algorithm [Hoche, Krauss, S., Siegert JHEP 0905 (2009) 053]
ME legs pre-determined shower emissionsdetermined by clustering inverse to the shower
→ pseudo shower history for MEs
PS starts off a reconstructed 2→ 2 corecan radiate gluons off “intermediate” lines→ Truncated Shower
ME branchings must be respectedevolution-, splitting- & angular variables k2
⊥, z, φ preserved
veto event if shower emission above Qcut
preserves the log-accuracy of the shower
implementations
Sherpa-1.2 [Hoche, Krauss, S., Siegert JHEP 0905 (2009) 053]
Herwig for e+e− [Hamilton, Richardson, Tully JHEP 0911 (2009) 038]
pseudo shower history
NNLO
cluster once
find k2⊥;z;φ
NLO
k2⊥
cluster twice
find k′2⊥ ;z′;φ′
LOk′2⊥
k2⊥
Truncated ShowerQ < Qcut Q > Qcut
N3LOLO
k′2⊥
k2⊥
Steffen Schumann Monte Carlo Event Generators for the LHC
ME⊕PS: facts & figures
Qcut and/or Nmax variation should affect σtot only beyond (N)LL
Example: DY-pair production σtot @ Tevatron
Nmax0 1 2 3 4 5 6
Qcut
20 GeV192.6(1)
191.0(3) 190.5(4) 189.0(5) 189.4(7) 188.2(8) 189.9(10)30 GeV 192.3(2) 192.7(2) 192.6(3) 192.9(3) 192.7(3) 193.2(3)45 GeV 193.6(1) 194.4(1) 194.3(1) 194.4(1) 194.6(2) 194.4(1)
LO
Nmax = 6
Nmax = 5
Nmax = 4
Nmax = 3
Nmax = 2
Nmax = 1
20 30 45
0.96
0.98
1.0
1.02
1.04
Qcut/GeV
σ/
σ(LO)
“merging systematics” of σtot < ±3%
Steffen Schumann Monte Carlo Event Generators for the LHC
ME⊕PS: facts & figures
Jet rates and -spectra improved compared to pure PS simulationdue to exact real emission ME’s
Example: DY-pair production σe+e−+NjetCDF Data: PRL 100 (2008) 102001
b
b
b
Nmax = 0
Nmax = 1
Nmax = 2
Nmax = 3
datab
10 1
10 2
10 3
10 4
σ(N
jet)
(scaledto
firstbin)
1 2 3
0.6
0.8
1
1.2
1.4
Njet
MC/data
Note: σtot preserved but big effectson rare events!
e+ e−e+
e− e+ e−
e+ e−e+
e− e+ e−
e+ e−e+
e− e+ e− e+ e−
Qcut = 30 GeV
Steffen Schumann Monte Carlo Event Generators for the LHC
ME⊕PS: facts & figures
Variation of Qcut should affect distributions only beyond (N)LLbut Qcut must be in range where PS approximation is valid
Example: All-jets pT ’s in DY-pair production CDF Data: PRL 100 (2008) 102001
b
bbb
bb
b
b
b
b
b
Qcut = 20GeV
Qcut = 30GeV
Qcut = 45GeV
datab
10−1
1
10 1
10 2
dσ/d
p⊥(jet)forN
jet≥
1
50 100 150 200 250 300 350 400
0.6
0.8
1
1.2
1.4
p⊥(jet) [GeV]
MC/data
b
b
b
b
b
b
b
Qcut = 20GeV
Qcut = 30GeV
Qcut = 45GeV
datab
10−1
1
10 1
10 2
dσ/d
p⊥(jet)forN
jet≥
2
50 100 150 200 250 300
0.6
0.8
1
1.2
1.4
p⊥(jet) [GeV]
MC/data
Nmax = 5 Nmax = 5
Steffen Schumann Monte Carlo Event Generators for the LHC
Matching Parton Showers with NLO: NLO⊕PS
Objectives:
Born observable (e.g. total rate) accurate to NLO
first/hardest emission according to real emission
further collinear/soft emissions according through shower
→ resolve double counting for the first/hardest emission
→ preserve log-resummation of the shower scheme
The MC@NLO approach
[Frixione, Webber JHEP 0206 (2002) 029]
modified subtraction in σNLO
O(αS) shower exp. removed
shower specific
first emission gets corrected
implemented for Herwig shower
many processes [also BSM]
The POWHEG approach
[Nason JHEP 0411 (2004) 040]
dσB → dσB to get NLO rate
hardest shower emission from σR
vetoed & truncated shower
used with Pythia/Herwig
Powheg-Box
recently automated in Sherpa
Steffen Schumann Monte Carlo Event Generators for the LHC
Matching Parton Showers with NLO calculations
Example: top-pairs @ LHC
[Frixione et al. JHEP 0308 (2003) 007]
Example: Z − pT @ Tevatron
[Hoche et al. arXiv:1008.5399]
[DØ data arXiv:1006.0618]
b bbbbb
b
b
b
b
b
b
b
DØ datab
POWHEG
ME+PS (1-jet) × 1.2LO+PS × 1.2
10−8
10−7
10−6
10−5
10−4
10−3
10−2
10−1Z boson pT
1/
σd
σ/dp⊥(Z
)
0 50 100 150 200 250 300
0.6
0.8
1
1.2
1.4
p⊥(Z) [GeV]
MC/data
Steffen Schumann Monte Carlo Event Generators for the LHC
Pull the rabbit: MEnloPS
Objective: combine virtues of ME⊕PS & NLO⊕PS MEnloPS
inclusive rate accurate to NLOhigher emissions from tree-level matrix elementsall supplemented/combined with parton showers
Hamilton, Nason JHEP 1006 (2010) 039
jet multiplicities in pp → W− + X @ LHC
Hoche et al. arXiv:1009.1127
2nd jet pT in gg → h → ττ + X @ LHC
MENLOPS (3-jet)ME+PS (3-jet) × 2.1POWHEG
10−10
10−9
10−8
10−7
10−6
10−5
10−4
10−3
10−2
Transverse momentum of second jet
dσ/dp⊥(jet
2)[pb/GeV
]
10 2 10 3
0.6
0.8
1
1.2
1.4
p⊥(jet 2) [GeV]
Ratio
Steffen Schumann Monte Carlo Event Generators for the LHC
Pull the rabbit: MEnloPS
Objective: combine virtues of ME⊕PS & NLO⊕PS MEnloPS
inclusive rate accurate to NLOhigher emissions from tree-level matrix elementsall supplemented/combined with parton showers
Hamilton, Nason JHEP 1006 (2010) 039
jet multiplicities in pp → W− + X @ LHC
Hoche et al. arXiv:1009.1127
2nd jet pT in gg → h → ττ + X @ LHC
MENLOPS (3-jet)ME+PS (3-jet) × 2.1POWHEG
10−10
10−9
10−8
10−7
10−6
10−5
10−4
10−3
10−2
Transverse momentum of second jet
dσ/dp⊥(jet
2)[pb/GeV
]
10 2 10 3
0.6
0.8
1
1.2
1.4
p⊥(jet 2) [GeV]
Ratio
Steffen Schumann Monte Carlo Event Generators for the LHC
Summary: Parton Showers & Matching with Fixed Order
exclusive generation of multiple emissions
guided by QCD evolution equations
Multijet ME-PS merging sustainable approach to describe multijet events
hard emissions through exact tree-level matrix elements(intra) jet evolution through truncated QCD parton showers
complementary ansatz: showers matched to NLO matrix elements
latest: combination of both approaches MEnloPS
ultimate goal: combine NLO for different multiplicities plus showers
Steffen Schumann Monte Carlo Event Generators for the LHC
Leaving the perturbative ground:
The Underlying Event & Hadronization
Steffen Schumann Monte Carlo Event Generators for the LHC
The Underlying Event: remnant-remnant interactions
Definition: An attempteverything but the hard interaction including showers & hadronization→ soft & hard remnant-remnant interactions
Beyond factorization: Multiple-Parton Interactions
σ2→2QCD(p2
T ,min) =
s/4∫p2T,min
dp2T
dσ2→2QCD (p2
T )
dp2T
=
∫ ∫ ∫ s/4
p2T,min
dxadxbdp2T fa(xa, p
2T )fb(xb, p
2T )
dσ2→2QCD
dp2T
∼ α2S(p2
T )
p4T
for low pT ,min: 〈σ2→2QCD(p2
T ,min)/σNDpp 〉 = 〈n〉 > 1
there might be many interactions per event Pn =〈n〉nn!
e−〈n〉
strong dependence on cut-off pT ,min energy dependent!
Steffen Schumann Monte Carlo Event Generators for the LHC
The Underlying Event: remnant-remnant interactions
Definition: An attempteverything but the hard interaction including showers & hadronization→ soft & hard remnant-remnant interactions
Beyond factorization: Multiple-Parton Interactions
σ2→2QCD(p2
T ,min) =
s/4∫p2T,min
dp2T
dσ2→2QCD (p2
T )
dp2T
=
∫ ∫ ∫ s/4
p2T,min
dxadxbdp2T fa(xa, p
2T )fb(xb, p
2T )
dσ2→2QCD
dp2T
∼ α2S(p2
T )
p4T
for low pT ,min: 〈σ2→2QCD(p2
T ,min)/σNDpp 〉 = 〈n〉 > 1
there might be many interactions per event Pn =〈n〉nn!
e−〈n〉
strong dependence on cut-off pT ,min energy dependent!
Steffen Schumann Monte Carlo Event Generators for the LHC
Experimental Evidence
direct: DPS in γ + 3jets
CDF Phys. Rev. D56 (1997) 3811
indirect: jet shapes
r/R 0 0.2 0.4 0.6 0.8 1
(r/R
)
Ψ
0
0.2
0.4
0.6
0.8
1
CDF Run II Preliminary
-1 Data L = 1.7 fb
Pythia Tune A
Pythia w/o UE
Pythia Tune DW
ee + jets→Z2
< 116 GeV/cee66 < M
| < 1e1
η > 25 GeV, |eT
E
| < 2.8e2
η| < 1 || 1.2 < |e2
η|
| < 2.1jet
> 30 GeV/c, |yjetT
p
R(e,jet) > 0.7∆
Statistical uncertainties only
Steffen Schumann Monte Carlo Event Generators for the LHC
Multiple Interactions: A simple modelSjostrand, Zijl Phys. Rev. D 36 (1987) 2019
hard process defines scale pT ,hard
generate sequence of additional 2→ 2 QCD scatterings ordered in pT
P(pT ) =1
σND
dσ2→2QCD
dp2T
exp
−p2T,hard∫p2T
1
σND
dσ2→2QCD
dp2′T
dp2′T
with σ2→2
QCD regulated according to
dσ2→2QCD
dp2⊥→
dσ2→2QCD
dp2⊥× p4
⊥(p2⊥ + p2
⊥0)2
α2S (p2⊥ + p2
⊥0)
α2S (p2⊥)
[parameter pT,0 ≈ 2 GeV]
further featuresimpact parameter dependence [typically double Gaussian]
central collisions more active, Pn broader than Poissonian
use rescaled PDFs taking into account used up momentum Pn narrower than Poissonian
attach parton showers/hadronization
Steffen Schumann Monte Carlo Event Generators for the LHC
The Underlying Event: comparison to Tevatron data
Ncharged vs. p⊥,jet1 in different ∆φ regions w.r.t the leading jet
SHERPASHERPASHERPASHERPASHERPA
Min Bias Run IJet20 Run I
Sherpa w/o MIPYTHIA w/ MI
Sherpa w/ MI
in
1 G
eV
bin
Ch
arg
ed
N
1
2
3
4
5
6
7
8
9
10
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
-0.2
-0.10
0.1
0.2
GeV T, jet1
P
0 5 10 15 20 25 30 35 40 45 50
SHERPASHERPASHERPASHERPASHERPA
Min Bias Run IJet20 Run I
Sherpa w/o MIPYTHIA w/ MI
Sherpa w/ MI
in
1 G
eV
bin
Ch
arg
ed
N
0.5
1
1.5
2
2.5
3
3.5
4
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
Track Finding Efficiency: 0.92
>0.5T
p |<1.0η|Theory / Data - 1
-0.2
-0.10
0.1
0.2
GeV T, jet1
P
0 5 10 15 20 25 30 35 40 45 50
Steffen Schumann Monte Carlo Event Generators for the LHC
From partons to hadrons: Hadronization Models
Objectives: dynamical hadronization of multi-parton systemscapture main non-perturbative aspects of QCDuniversality→ robust extrapolation to new machines, higher energies
→ should not depend on specifics of the hard process
model (un)known decays of (un)known hadrons→ hadron multiplicities, meson/baryon ratios→ decay branching fractions
→ hadron-momentum distibutions
Lund string fragmentation
implemented in Pythia
Cluster-hadronization model
implemented in Herwig & Sherpa
Steffen Schumann Monte Carlo Event Generators for the LHC
From partons to hadrons: Cluster-Hadronization Model
Cluster-formation model
Cluster-decay model
features
preconfinement [colour neighboring partons after shower close in phase space]
parametrization of primary-hadron generation
locality and universality
Steffen Schumann Monte Carlo Event Generators for the LHC
From partons to hadrons: Cluster-Formation Model
Steffen Schumann Monte Carlo Event Generators for the LHC
From partons to hadrons: Cluster-Decay Model
Steffen Schumann Monte Carlo Event Generators for the LHC
Point of reference: LEP @√s = 91.2 GeV
particle multiplicities: Herwig++[Gieseke et al. JHEP 0402 (2004) 005]
Particle Measured LEP Herwig++
All Charged 20.924 ± 0.117 20.814
γ 21.27 ± 0.6 22.67
π0 9.59 ± 0.33 10.08
ρ(770)0 1.295 ± 0.125 1.316
π± 17.04 ± 0.25 16.95
ρ(770)± 2.4 ± 0.43 2.14η 0.956 ± 0.049 0.893ω(782) 1.083 ± 0.088 0.916
η′(958) 0.152 ± 0.03 0.136
K0 2.027 ± 0.025 2.062
K∗(892)0 0.761 ± 0.032 0.681
K∗(1430)0 0.106 ± 0.06 0.079
K± 2.319 ± 0.079 2.286
K∗(892)± 0.731 ± 0.058 0.657φ(1020) 0.097 ± 0.007 0.114
p 0.991 ± 0.054 0.947
∆++ 0.088 ± 0.034 0.092
Σ− 0.083 ± 0.011 0.071Λ 0.373 ± 0.008 0.384
Σ0 0.074 ± 0.009 0.091
Σ+ 0.099 ± 0.015 0.077
Σ(1385)± 0.0471 ± 0.0046 0.0312∗Ξ− 0.0262 ± 0.001 0.0286
Ξ(1530)0 0.0058 ± 0.001 0.0288∗Ω− 0.00125 ± 0.00024 0.00144
... ... ...
event shapes: Sherpa[Sherpa unpublished]
T = max|n|=1
∑i n · pi∑i |pi |
b
b bbbbbbbb
bb
bb
b
b
b
b
b
b
DELPHI datab
Njet = 2
Njet = 3
Njet = 4
Njet = 5
Sherpa1.2.110−3
10−2
10−1
1
10 1
1− Thrust
Nd
σ/d(1
−T)
0 0.1 0.2 0.3 0.4 0.5
0.6
0.8
1
1.2
1.4
1− T
MC/data
Steffen Schumann Monte Carlo Event Generators for the LHC
Summary: Monte Carlo event generators for the LHC
Stochastic simulation of exclusive events
Hard Processes at Fixed Order
Initial- & Final-State Parton Showers
Underlying Event/Multiple Interactions
Hadronization
...
produce exclusive events at the rate produced in nature
Ptot = PHard · PShower · PUE · PHadronization · PDecays · PQED
Steffen Schumann Monte Carlo Event Generators for the LHC
First LHC tests passed ...
Dijet Mass (GeV)200 400 600 800
|<1.
3)η
|<0.
7)/N
(0.7
<|η
N(|
0.5
1
1.5 CMS Preliminary = 7 TeVs
(M<838 GeV)-1Data 120 nbNLONLO+Non-Pert. CorrectionNLO UncertaintyPYTHIA6PYTHIA6 x NLO/LO
dijetϕ∆
dijet
ϕ∆
dd
N
N1
310
210
110
1
10
210
)3
200 GeV (x10≥ T
max p
)2 200 GeV (x10≤ T
max p≤ 120
120 GeV (x10)≤ T
max p≤ 90
90 GeV≤ T
max p≤ 70
)GENSMR Herwig++ (
)GENSMR Pythia 6 (
)GENSMR MadGraph (
= 7 TeVs@ pp
1L = 72 nb
1.1≤|y|
CMS Preliminary
/2π /3π2 /6π5 π
η-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
η/d
chdN
0
1
2
3
4
5
6
7
8Data 7 TeVPYTHIA-6 D6TPYTHIA-6 DWPYTHIA-6 P0PYTHIA-6 CWPYTHIA-8
CMS preliminary
> 3 GeV/cT
leading track-jet p
> 0.5 GeV/c)T
charged particles (p
Steffen Schumann Monte Carlo Event Generators for the LHC
LHC data is coming!
Frigate Monte Carlo
Steffen Schumann Monte Carlo Event Generators for the LHC
Aside: Logarithmic accuracy Catani–Seymour shower
Catani–Seymour Shower Sudakov form factor
∆CSq (µ2,Q2) = exp
−∫ Q2
µ2
dk2T
k2T
αS(λRk2T )
π
[CF log
(Q2
k2T
)− 3
2CF + . . .
]
∆CSg (µ2,Q2) = exp
−∫ Q2
µ2
dk2T
k2T
αS(λRk2T )
π
[CA log
(Q2
k2T
)− 1
6(11CA − 2nf ) + . . .
]Analytic QCD Sudakov form factor
∆QCDa (µ2,Q2) = exp
−∫ Q2
µ2
dk2T
k2T
[Aa(αS(k2
T )) log
(Q2
k2T
)+ Ba
(αS(k2
T ))]
Aa(αS ) =∞∑n=1
(αS
π
)nA
(n)a and Ba(αS ) =
∞∑n=1
(αS
π
)nB
(n)a
a=q: A(1)q = CF , B
(1)q = − 3
2CF , A
(2)q = 1
2CFK , with K = CA
(67
18− π2
6
)− 5nf
9
q=g: A(1)g = CA, B
(1)g = − 1
6(11CA − 2nf ), A
(2)g = 1
2CAK
set λR = exp(−K/2β0) = exp(− 67−3π2−10nf /3
33−2nf
)[Catani et al. Nucl. Phys. B 349 (1991) 635]
Steffen Schumann Monte Carlo Event Generators for the LHC
Aside: Phase-space slicing
The proposed measure [example final-state splitting]
Q2ij = 2 pipj min
k 6=i,j
2
C ki,j + C k
j,i
; C ki,j =
pipk
(pi + pk )pj− m2
i
2 pipjif j = g
1 else
→ minimize over color partners k
IR limitae
soft limit: pj = λq, λ → 0
1
Q2ij
→ 1
2λ2
1
2 pi qmaxk 6=i,j
[pipk
(pi + pk ) q− m2
i
2 piq
]quasi-collinear limit: k⊥ → λk⊥, m → λm
1
Q2ij
→ 1
2λ2
1
p 2ij −m2
i −m2j
(Ci,j + Cj,i
); Ci,j =
z
1− z− m2
i
2 pipjif j = g
1 else
measure correctly identifies enhanced phase-space regions
Steffen Schumann Monte Carlo Event Generators for the LHC