MC Tuning from TeV to LHC based on Dijet Azimuthal...
Transcript of MC Tuning from TeV to LHC based on Dijet Azimuthal...
MC Tuning from TeV to LHCbased on Dijet Azimuthal Decorelations
Markus Wobisch, Fermilab
TeV4LHC Workshop – September 16, 2004
• Motivation / Observable
• Experimental Results
• Fixed Order pQCD Description
• MC Tuning to TeV Data
• Extrapolation to LHC Energies?
in collaboration with: A. Kupco, M. Begel, C. Royon, M. Zielinski
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 1
Monte Carlos – what can be tuned?
event topology: fundamental signature + broad features + fine details
Monte Carlo Event Generators:
• LO Matrix Elements for fundamental process (e.g. 2 → 2) → normalization uncertainty!
• perturbative parton cascade models (based on soft and collinear approximations)
• phenomenological models for non-perturbative phase (hadronization, underlying event)
Tuning MCs?
• LO MEs are exact (use most recent αs and MRST/CTEQ pdfs)
• everything else can be tuned!! – plenty of parameters in PYTHIA / less in HERWIG
How to tune MCs? (don’t use “hard” parameters to fix “soft” physics)
• first: find k-factor to fix normalization problem – or use normalized differential distrib.
• second: use “broad” event features to tune “hard” physics ⇐ scope of this talk
• third: tune soft physics to describe finer details
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 2
Dijet Azimuthal Decorrelations
∆φdijet
Dijet Production:
• limit: exactly two jets, no further radiation
azimuthal opening anglebetween both leading pT jets:
⇒ ∆φ dijet = π
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 3
Dijet Azimuthal Decorrelations
∆φdijet
Dijet Production:
• limit: exactly two jets, no further radiation
• additional soft radiation outside the jets
azimuthal opening anglebetween both leading pT jets:
⇒ ∆φ dijet = π
⇒ ∆φ dijet small deviations from π
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 4
Dijet Azimuthal Decorrelations
∆φdijet
Dijet Production:
• limit: exactly two jets, no further radiation
• additional soft radiation outside the jets
• one additional high pT jet
azimuthal opening anglebetween both leading pT jets:
⇒ ∆φ dijet = π
⇒ ∆φ dijet small deviations from π
⇒ ∆φ dijet as small as 2π/3
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 5
Dijet Azimuthal Decorrelations
∆φdijet
Dijet Production:
• limit: exactly two jets, no further radiation
• additional soft radiation outside the jets
• one additional high pT jet
• multiple additional hard jets in the event
azimuthal opening anglebetween both leading pT jets:
⇒ ∆φ dijet = π
⇒ ∆φ dijet small deviations from π
⇒ ∆φ dijet as small as 2π/3
⇒ small ∆φ dijet – no limit
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 6
Dijet Azimuthal Decorrelations
∆φdijet
Dijet Production:
• limit: exactly two jets, no further radiation
• additional soft radiation outside the jets
• one additional high pT jet
• multiple additional hard jets in the event
azimuthal opening anglebetween both leading pT jets:
⇒ ∆φ dijet = π
⇒ ∆φ dijet small deviations from π
⇒ ∆φ dijet as small as 2π/3
⇒ small ∆φ dijet – no limit
⇒ ∆φ dijet distribution is sensitive to higher order pQCD effects without requiring thereconstruction of additional jets (Yes: this is an experimental advantage!!)
⇒ ∆φ dijet: examine transition between soft and hard physics, based on single observable
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 7
Defining the Observable
• define the jets using iterative, seed-based midpoint cone algorithmwith Rcone = 0.7 in rapidity and azimuthal angle(the “Run II cone algorithm” ⇐ Run II Workshop)
• Define the observable to be the normalized differential ∆φ dijet distribution:
1
σdijet
·dσdijet
d∆φ dijet
• measure the observable as a function of a hard scale:⇒ in four different regions of the leading jet pT – starting at pmax
T > 75 GeV
requiring the second leading pT jet to have pT > 40 GeV
• require that both leading pT jets have central rapidity |yjet| < 0.5
the ∆φ dijet distribution is a three-jet observable! the ratio is ∝ α3s/α2
s
recently available: NLO pQCD predictions for 3-jet observables: pQCD at O(α4s)
(NLOJET++, Z. Nagy)
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 8
Non-Perturbative Effects
Hadronization Corrections: Obs.hadronObs.parton
Underlying Event: Obs.with UEVTObs.w/o UEVT
0.951
1.05
1.5707 1.9634 2.3561 2.7489 3.1416
hadr
oniz
atio
n co
rrec
tion
c =
σha
dron
/σpa
rton
pT max > 180 GeV
0.951
1.05
1.5707 1.9634 2.3561 2.7489 3.1416
130 < pT max < 180 GeV
0.951
1.05
1.5707 1.9634 2.3561 2.7489 3.1416
100 < pT max < 130 GeV
0.951
1.05
1.5707 1.9634 2.3561 2.7489 3.1416
∆φ dijet (rad)
75 < pT max < 100 GeV
π/2 3π/4 π
PYTHIAHERWIG
0.95
1
1.05
1.5707 1.9634 2.3561 2.7489 3.1416
unde
rlyin
g ev
ent c
orre
ctio
n
pT max > 180 GeV
0.95
1
1.05
1.5707 1.9634 2.3561 2.7489 3.1416
130 < pT max < 180 GeV
0.95
1
1.05
1.5707 1.9634 2.3561 2.7489 3.1416
100 < pT max < 130 GeV
0.95
1
1.05
1.5707 1.9634 2.3561 2.7489 3.1416
∆φ dijet (rad)
75 < pT max < 100 GeV
PYTHIA 6.225
π/2 3π/4 π
Non-perturbative effects are below 5% =⇒ only sensitive to perturbative effects
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 9
Experimental Results
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
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
DØ
10-3
10-2
10-1
1
10
10 2
10 3
10 4
10 5
0.5 0.625 0.75 0.875 1
π/2 3π/4 π
First Tevatron Run II QCD Jet Publicationhep-ex/0409040 submitted to PRL today!
• data in four pmaxT regions:
⇒ more strongly peaked at high pmaxT
⇒ decreasing by more than 4 orders ofmagnitude from ∆φ dijet = π to π/2
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 10
Experimental Results
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
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
LONLO
NLOJET++ (CTEQ6.1M)µr = µf = 0.5 pT
max
DØ
10-3
10-2
10-1
1
10
10 2
10 3
10 4
10 5
0.5 0.625 0.75 0.875 1
π/2 3π/4 π
First Tevatron Run II QCD Jet Publicationhep-ex/0409040 submitted to PRL today!
• data in four pmaxT regions:
⇒ more strongly peaked at high pmaxT
⇒ decreasing by more than 4 orders ofmagnitude from ∆φ dijet = π to π/2
• LO pQCD prediction is poor
⇒ reasonable only in limited ∆φ dijet range
⇒ ∆φ dijet < 2π/3 no phase space
⇒ ∆φ dijet → π divergence
• NLO pQCD prediction is very good
⇒ (see ratios for details)
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 11
Quantitative Comparison: Data and NLO
1
2
1.5707 1.9634 2.3561 2.7489 3.1416
pT max > 180 GeV
Dat
a / N
LO T
heor
y
DØ
1
2
1.5707 1.9634 2.3561 2.7489 3.1416
130 < pT max < 180 GeV
µr, µf dependencePDF uncertainty
1
2
1.5707 1.9634 2.3561 2.7489 3.1416
100 < pT max < 130 GeV
1
2
1.5707 1.9634 2.3561 2.7489 3.1416
75 < pT max < 100 GeV
∆φ dijet (rad)
NLOJET++ (CTEQ6.1M)
π/2 3π/4 π
• NLO pQCD:
⇒ good description of the dataon average 5–10% below data
⇒ except at ∆φ dijet close to π
(soft processes – needs resummation)
• renormalization and factorizationscale dependence:0.25pmax
T < µr,f < pmaxT
⇒ small at intermediate ∆φ dijet
⇒ large at ∆φ dijet → π (soft region)
⇒ large at ∆φ dijet < 2π/3
(only tree-level four parton final states)
• PDF uncertainty using CTEQ6.1M pdfs
⇒ dominant at intermediate ∆φ dijet
larger in high pmaxT region
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 12
Comparison: Data and MCs
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
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
HERWIG 6.505PYTHIA 6.225
DØ
(CTEQ6L)
10-3
10-2
10-1
1
10
10 2
10 3
10 4
10 5
0.5 0.625 0.75 0.875 1
π/2 3π/4 π
• HERWIG v6.505 (default)
⇒ good description of the dataover whole ∆φ dijet range
• PYTHIA v6.225 (default)
⇒ significantly too low at small ∆φ dijet
⇒ too narrowly peaked at π
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 13
Comparison: Data and MCs
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
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
HERWIG 6.505PYTHIA 6.225PYTHIAincreased ISR
(CTEQ6L)
DØ
10-3
10-2
10-1
1
10
10 2
10 3
10 4
10 5
0.5 0.625 0.75 0.875 1
π/2 3π/4 π
• HERWIG v6.505 (default)
⇒ good description of the dataover whole ∆φ dijet range
• PYTHIA v6.225 (default)
⇒ significantly too low at small ∆φ dijet
⇒ too narrowly peaked at π
• changing maximum pT in ISR shower(remember: Rick Field’s PYTHIA “tune A”)
⇒ change: PARP(67)=1.0 → 4.0
PARP(67) × hard scale (' pT ) definesthe maximum virtuality in ISR shower
⇒ directly related to max. pT in ISR shower
⇒ huge effect for ∆φ dijet distribution
⇒ best value somewhere betweenPARP(67)=1.0 and =4.0
⇒ hard processes can be adjusted!
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 14
Data and MCs — looking at ∆φ dijet ≈ π
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
d∆φ
dije
t
pT max > 180 GeV (×30)130 < pT
max < 180 GeV (×10)
100 < pT max < 130 GeV (×3)
75 < pT max < 100 GeV
HERWIG 6.505PYTHIA 6.225 default increased pT max ISR
DØ
(CTEQ6L)
1
10
10 2
0.8125 0.875 0.9375 1
15π/167π/813π/16 π
– zoom into the peak– this is where NLO fails (soft processes!)– where parton shower should work
use same MCs as before:
• HERWIG (default)
⇒ slightly to narrow — but reasonable
• PYTHIA (default)
⇒ much too narrowly peaked at π
too low everywhere else
• PYTHIA with PARP(67)=4.0
⇒ too narrow in peak
⇒ too low at ∆φ dijet ≈ 15π/16
(low ∆φ dijet tail slightly high)
⇒ more tuning needed for PYTHIA to describe peak region (soft processes)
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 15
Tuning PYTHIA – soft processes
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
d∆φ
dije
t
pT max > 180 GeV (×8)130 < pT
max < 180 GeV (×4)100 < pT
max < 130 GeV (×2)
75 < pT max < 100 GeV
PYTHIA 6.225 default increased pT max ISR
decreased xµ ISR
increased prim. kT
DØ
(CTEQ6L)
1
10
10 2
0.875 0.9062 0.9375 0.9688 1
15π/167π/8 π
vary PYTHIA parameters related to ISR
• pT max ISR PARP(67)=4.0 (D=1.0)
⇒ small effect at high ∆φ dijet for low pmaxT
• xµ ISR PARP(64)=0.5 (D=1.0)
⇒ effect is negligible
• primordial kT PARP(91)=4.0 (D=1.0)
upper cut-off PARP(93)=8.0 (D=5.0)
⇒ very small effect at high ∆φ dijet for low pmaxT
⇒ nothing helps!
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 16
Tuning PYTHIA – soft processes
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
d∆φ
dije
t
pT max > 180 GeV (×150)130 < pT
max < 180 GeV (×25)100 < pT
max < 130 GeV (×5) 75 < pT
max < 100 GeV
HERWIG 6.505PYTHIA 6.225 default increased pT max FSR
DØ
(CTEQ6L)10
-1
1
10
10 2
10 3
0.75 0.8125 0.875 0.9375 1
3π/4 7π/8 π
vary PYTHIA parameters related to FSR:
• pT max ISR ↔ PARP(67) was so successful
⇒ try the same thing for FSR
• pT max FSR ↔ PARP(71)
⇒ increase: PARP(71)=8.0 (D=4.0)
⇒ zero effect!
⇒ Here we ran out of ideas...
More suggestions for PYTHIAparameters variations are welcome!!
HERWIG: we tried PTRMS=1.5 GeV (D=0)
⇒ no effect
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 17
From Tevatron to the LHC
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
d∆φ
dije
t
75 < pT max < 100 GeV
100 < pT max < 130 GeV
130 < pT max < 180 GeV
pT max > 180 GeV
180 < pT max < 500 GeV
500 < pT max < 1200 GeV
pT max > 1200 GeV
LHC
Tevatron
NLO pQCD
(all curves: from top to bottom)
NLOJET++ (CTEQ6.1M)µr = µf = 0.5 pT
max10-4
10-3
10-2
10-1
1
0.5 0.625 0.75 0.875 1
π/2 3π/4 π
NLO gives a very good description⇒ use NLO as a reference
compare NLO predictions for ∆φ dijet
at Tevatron and LHC
for both: Run II cone algorithm, |yjet| < 0.5
• Tevatron Run II (as: hep-ex/0409040)
⇒ pT 2 > 40 GeV
⇒ four pmaxT regions
• LHC
⇒ pT 2 > 80 GeV
⇒ three pmaxT regions
⇒ The chosen pmaxT ranges for the LHC results cover the spread of the Tevatron results
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 18
A last look at the Tevatron ...
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
d∆φ
dije
t
pT max > 180 GeV (×103)130 < pT max < 180 GeV (×102)100 < pT max < 130 GeV (×10) 75 < pT max < 100 GeV
from top to bottom:
75 < pT max < 100 GeV (×8)
100 < pT max < 130 GeV (×4)
130 < pT max < 180 GeV (×2)
pT max > 180 GeV
NLOHERWIGPYTHIA (TeV-tuned)
Tevatron Run II
10-4
10-3
10-2
10-1
1
10
0.5 0.625 0.75 0.875 1
π/2 3π/4 π
best description by PYTHIAfor PARP(67) between D=1.0 and 4.0
• tune PARP(67) to NLO
⇒ result: PARP(67)=2.5 (D=1.0)
⇒ this setting is now referred to as “TeV-tuned”
• (ignore the peak region...)
⇒ good agreement:HERWIG ≈ PYTHIA ≈ NLO
Question:Can this good agreement (and the tune)be transferred to the LHC?
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 19
... and a first look at the LHC
∆φ dijet (rad)
1/σ di
jet
dσdi
jet /
d∆φ
dije
t
from top to bottom:
180 < pT max < 500 GeV (×4)
500 < pT max < 1200 GeV (×2)
pT max > 1200 GeV
NLOHERWIGPYTHIA (TeV-tuned)
LHC
10-4
10-3
10-2
10-1
1
10
0.5 0.625 0.75 0.875 1
π/2 3π/4 π
... a huge success!!! – ... expected??
• PYTHIA (TeV-tuned)
⇒ the good agreement with NLOat Tevatron Run II energiesis reproduced at LHC energies!!
• HERWIG (default)
⇒ small differences:broader at low pmax
T
narrower at large pmaxT
⇒ Both Monte Carlos are in good agreement with NLO predictions
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 20
Summary and Conclusions
using Dijet Azimuthal Decorrelations to test & tune Monte Carlo event generators:
• normalized distribution⇒ not affected by poor absolute normalization of LO Matrix Elements
• not sensitive to non-perturbative effects (hadronization, underlying event)⇒ allows to tune perturbative parameters in MCs w/o interference of “soft parameters”
• strategy must be:⇒ tune “hard parameters” first — then the “soft parameters”⇒ e.g. order: Dijet Azimuthal Decorrelations → Jet Shapes → Underlying Event
• HERWIG: not much to tune (no parameters / but also: not necessary)• PYTHIA: only sensitivity: pT max ISR – result: PARP(67)=2.5 (D=1.0)
⇒ this should be the basis for a new Tevatron tune (“tune A-prime”?)
surprise: PYTHIA tuning can be transferred to LHC energies=⇒ very promising for tuning MCs for LHC!
Markus Wobisch, Fermilab MC Tuning from TeV to LHC TeV4LHC Workshop, 09/15/04 21