MadGraph5_aMC@NLO

Crash in Mint_MC with added cut

Asked by Josh Kunkle

Hello,

I am attempting to simulate W+2 photons at NLO using a cut that I added into NLO/SubProcesses/cuts.f. The cut is on the sum of the mass of the lepton+neutrino + the sum of the lepton+neutrino+lead photon [M(l nu) + M(l nu gamma)]. We require the cut to be greater than 165 to select ISR-like events and less than 165 to select FSR-like events. The point of adding the cut is that the inclusive cross section is dominated by FSR events, so this cut allows us to better sample ISR events. When running with the ISR case I see a crash during the integration. In the FSR case the generation finishes successfully. Presumably this is due to a poor sampling in some phase space, but I'm not sure how to debug further in this case. Any help is appreciated. The procedure to replicate the crash (on lxplus) is below. You can see the code added by looking at the patch file. You can find the cut value in the run_card. It is called mlnpmlnacut. Set it to 0. to disable the cut, set it to -165. to run the FSR like generation (which should run to completion). Let me know if I can provide additional information.

Thank you
Josh

# To reproduce the crash from a clean madgraph installation.....

#get Madgraph 2.2.2 (you can get the tar from my public, its not modified at all)

cd /tmp # or some temporary directory of your choice

cp /afs/cern.ch/user/j/jkunkle/public/MG5_aMC_v2.2.2.tar.gz .
gunzip MG5_aMC_v2.2.2.tar.gz

tar -xvf MG5_aMC_v2.2.2.tar

cd MG5_aMC_v2_2_2/

# get the patch file

cp /afs/cern.ch/user/j/jkunkle/public/mg5_amc_v2_2_2_wggcut.patch .

# apply patch

patch -l -p0 -i mg5_amc_v2_2_2_wggcut.patch

# open madgraph (It may ask to update versions, say no)

./bin/mg5_aMC

> define l- = e- mu- ta- ; define l+ = e+ mu+ ta+ ; generate p p > l+ vl a a [QCD] ; add process p p > l- vl~ a a [QCD] ; output LNuAA_ISR_0j_PtG10

> exit

# copy the cards

cp /afs/cern.ch/user/j/jkunkle/public/proc_card_mg5.dat LNuAA_ISR_0j_PtG10/Cards/

cp /afs/cern.ch/user/j/jkunkle/public/run_card.dat LNuAA_ISR_0j_PtG10/Cards/

cp /afs/cern.ch/user/j/jkunkle/public/madspin_card.dat LNuAA_ISR_0j_PtG10/Cards/

# enter the directory and generate events

cd LNuAA_ISR_0j_PtG10

./bin/generate_events

> 3 (turn shower off because we don't run the shower to make gridpack)

It should run for approximately 30 minutes before crashing

Question information

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

Hi Josh,
Indeed, it is strange... Iadd Rikkert to the thread.
Rik, can you please have a look and tell us what you think?

Many thanks,

Marco

Revision history for this message
Rikkert Frederix (frederix) said : 		#2

Hello Josh,

I think the problem might be that the two photons can become very collinear in your case. Even though I don't think there is a singularity related to this, it makes the computation of the loops unstable leading to a crash.

You therefore have to change the cuts and do not allow for these very collinear photons. Either by setting " .true. = isoEM" in the run_card.dat or by implementing your own cut in the <Your_Processes>/SubProcesses/cuts.f

Best regards,
Rikkert

Revision history for this message
Josh Kunkle (jmkunkle) said : 		#3

Hi Rikkert,

Thanks for your response. I forgot to mention this in my post, but indeed if I set .true. = isoEM I do not see the crash with the cut at 165. But, if I keep isoEM to true and increase the cut (I tried 300) I again see the crash. So I suppose that with high enough cuts I push it into a bad region again.

I originally set isoEM to false because I assumed that the Frixione isolation is implemented to handle photon fragmentation properly and would have little to do with EM objects. But it seems in this case it may be a good idea to keep it. So I think I'll just do that.

best
Josh