AcerMC and ISR/FSR systematics at ATLAS
Download
Report
Transcript AcerMC and ISR/FSR systematics at ATLAS
AcerMC and ISR/FSR systematics at ATLAS
Top Physics Workshop (Grenoble 2007)
Liza Mijovic, Borut Kersevan
Jozef Stefan Inst.
Univ. of Ljubljana
ATLAS approach:
• Generator level studies
• Parameters treated
• Interesting examples
Impact of different models
Recently a study of top mass reconstruction using tt~ was done using:
MC@NLO (Herwig+Jimmy)
AcerMC (Pythia – new showering and UE model)
Full detector simulation
The observed discrepancy caused quite a few raised eyebrows..
We cannot know
offhand which answer
is correct!
2
Impact of different models cont’d
The first thought was that NLO corrections impact the event
shapes more than anyone suspected...
...But the difference turns out to be purely parton-shower
related!
We just plugged the AcerMC events into Herwig and..
We cannot know
offhand which answer
- Herwig/Pythia showers and UE is correct! We need the
data!
The difference becomes really small..
3
QCD-activity related systematics
From the above example one can see that the predictions are by
no means unique; using the standard division one needs to have a
look at:
Initial state radiation
Final state radiation
Underlying event modeling
PDFs, etc...
I think I don’t need to stress that the precision in top
measurements at the LHC will be systematics-dominated..
At ATLAS the UE is handled by tuning the available models to
the Tevatron data so it was excluded from these studies.
4
Strategy
First thing to estimate is the prediction range the models on the
marked provide – and the experimentalists can tune on the data.
In this respect HERWIG is rather unflexible: It has excellent
theoretical basics but very few parameters in the shower activity
that are allowed to have at least some uncertainty..
Pythia is in this respect much more flexible so at least as a start a
detailed study was made on how (and how much) we can ‘push around’
the Pythia showering activity.
AcerMC + Pythia with varying parameters was thus used to check on the
prediction uncertainties w.r.t. the QCD
(parton-)showering
activity.selection criteria:
Semileptonic decay
Studies
done< 2.5
on generator/truth
level
•pT > 40first
GeV, |eta|
for 2 b-jets and at least
twoonly.
light jets.
> 20variation
GeV for thetypical/simple
lepton from W, isolation
requirement
anywere
jet)>0.4
In•pT
each
analysis
cuts and~cone(lepton,
procedures
•No cuts on missing ET .
used
for top mass reconstruction and the distributions were
•No jet energy rescaling.
compared.
•W reconstruction: truth / 2 light jets with MJJ closest to MW and MJJ < 120 GeV.
•The JJB1,2 combination with highest pT chosen as the top candidate.
5
Pythia ISR parameters
ATLAS uses the Pythia new pT-ordered showering and UE model!
A lot of switches:
MSTP(62): level of coherence
MSTP(70): regularization scheme, pT ->0
PARP(62), PARP(81), PARP(82)
MSTP(72):max. pT for FSR of ISR partons
PARP(61) : Λ(QCD)
PARP(64) : evolution scale factor
ISR master switch, ME corrections ...
6
Rregularization scheme: PARP(81)
Default : PARP(81)=1.9 GeV=D
PARP(81)=D/2
PARP(81)=2*D
t-tbar rel. pT
# of truth-jets
7
Rregularization scheme: PARP(81)
Default : PARP(81)=1.9 GeV=D
PARP(81)=D/2
PARP(81)=2*D
b jet / b quark E ratio
light jet / light q. E ratio
8
Rregularization scheme: PARP(81)
Default : PARP(81)=1.9 GeV=D
PARP(81)=D/2
PARP(81)=2*D
top mass: truth W + bjet
top mass: W(jets)+bjet
9
PARP(61): Λ(QCD)
Default : PARP(61)=0.192 GeV
PARP(61)=0.1 GeV
PARP(61)=0.4 GeV
t-tbar rel. pt
truth-jet number
10
PARP(61): Λ(QCD)
Default : PARP(61)=0.192 GeV
PARP(61)=0.1 GeV
PARP(61)=0.4 GeV
high-pt truth jet n.
top mass: W(jets)+bjet
11
PARP(64): evolution scale factor
Default : PARP(64)=1=D
PARP(64)=D/2
PARP(64)=2*D
t-tbar rel. pt
top mass: W(jets)+bjet
12
Pythia FSR parameters
Tunable & relevant parameters for the new showering:
PARJ(81): Λ(QCD) (for external processes)
(D=0.25 GeV)
PARP(71): scale of the hard scattering (D=4)
PARJ(82): mass cut-off below which partons don’t radiate (D=1 GeV)
13
PARJ(81): Λ(QCD)
Top mass: final W + b jet
AcerMC + Pythia:
Parj(81) = 0.25 GeV
Parj(81) = 0.14 GeV
MCatNLO+Herwig
Reconstruction:
True W from + truth b-jet.
MT(AcerMC) <
MT(AcerMC-FSR) < MT(MC@NLO)
The FSR change in Pythia goes in the right
direction!
14
PARP(71) (scale of the hard scattering)
b quark pt, number of high-pt jets, pt of the jets, pt of b-jets
15
PARP(71)
top mass (final W + bjet)
PARP(71) has no/little effect
16
Pythia ISR and FSR parameters
PARAMETER
ISR
MSTP(70): reg. scheme
0 : PARP(62)
1 : PARP(81)
2 : PARP(82)
PARP(61) : Λ(QCD)
PARP(64) : evol. Factor
FSR
PARJ(81) : Λ(QCD)
TOP MASS
low
high
f
f
f
g
f
f
These parameters were then combined into two samples which
lower/increase the reconstructed mass to obtain endpoints..
17
Systematics sample proposal
AcerMC + Pythia, new showering
Minimum top mass:
2*PARJ(81)
(Λ(QCD) , FSR)
0.5* PARP(61) (Λ(QCD) , ISR)
2*PARP(62)
(pt -> 0, kt cut-off, ISR)
Maximum top mass:
0.5*PARJ(81)
2* PARP(61)
0.5*PARP(62)
(Λ(QCD) , FSR)
(Λ(QCD) , ISR)
(pt -> 0, kt cut-off, ISR)
18
B-fragmentation studies
19
B-fragmentation studies II
20
B-fragmentation studies III
21
B-fragmentation studies IV
The overall effect of
varying b-fragm. Parameters
gives ~stable results. However:
Need to compare with other
models like Herwig or EvtGen
22
Further studies/plans
I don’t have time to go through all of the things we had done but
to summarize other results:
The impact of QCD uncertainties on the ttbar cross-section
measurements (efficiency) seems to be small (percent order)
so not crucial at the initial measurement stages. Further
investigation ongoing...
Studies on the impact of PDF uncertainties is planned.
Impact of other models like EvtGen on b-tagging and top
reconstruction needs to be studied.
We need to develop methods on extracting the QCD model
parameters from the data and/or validate different
showering models.
A lot of work...
23