MadGraph5_aMC@NLO

PYTHIA8 shower settings

Asked by jannisgebbo

Dear MadGraph team,

I am generating the NLO process
p p > j j [QCD]
and shower with PYTHIA8 through the MadGraph interface.

Because I discovered poor agreement with experimental data, I checked the shower settings used internally, given in MCatNLO/Scripts. I was a bit surprised to see some non-default PYTHIA settings, especially:

TimeShower:QEDshowerByQ = off ! Prevent quarks from radiating photons
                                                                          ! ** USE on FOR REALISTIC SHOWER SIMULATION ** !!!
TimeShower:QEDshowerByL = off ! Prevent leptons from radiating photons
                                                                         ! ** USE on FOR REALISTIC SHOWER SIMULATION ** !!!
TimeShower:QEDshowerByGamma = off ! Prevent photons from branching
                                                                                 ! ** USE on FOR REALISTIC SHOWER SIMULATION ** !!!

Why are these chosen to be off by default? It even says that for realistic shower simulation, they should be on. This information is quite hidden in a place that standard users do not alter. So I am asking myself why is that and whether it is safe to turn these on?

Thanks and best,
Jannis

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
davide.pagani.85 Edit question
Solved by:
jannisgebbo
Solved:
Last query:
Last reply:
Revision history for this message
marco zaro (marco-zaro) said : 		#1

Hi,
I do not think these settings make any difference for the process at hand.
Long ago, i played with dijet @NLO+PS and I recall very serious issues due to the low efficiency and very steeply-falling spectra, plus the fact that one must use generation cuts which are looser than those in the analysis.
Playing with the bias function may help, hoping it does not introduce too large fluctuations.
Best,

Marco

Revision history for this message
jannisgebbo (jannisgebbo) said : 		#2

Thanks for the reply!

Indeed, I checked and the differences are marginal. Plus, you are right with the low efficiency. However, flattening the steeply falling spectra is necessary to get high enough statistics at high p_T.

Best,
Jannis