MadGraph5_aMC@NLO

Too few events for ee~ > tt~ at NLO and pythia showering

Asked by RomainM

Hello,

I am trying to generate e e~ > t t~ events and decay/shower them with Pyhtia 8. While everything works well at LO, I obtain something strange at NLO: only few events are passing the pythia step. I am unsure whether it's a Pythia-related issue or not, and I apologize if I am not asking at the right place.

The problem is that I generate 10k events and only ~250 are out from pythia (while I get the 10k for the LO generation). Looking at logfile, I beleive that this is due to this error (from pythia logfile):
 > Error in BeamRemnants::setKinematics: kinematics construction failed

I am not sure which logfile/info would be useful for you to help me, but I am happy to provide any of them. The commands I am using to generate the events are:
 - LO: 'generate e+ e- > t t~ '
 - NLO: 'generate e+ e- > t t~ [QCD]'
Then I go in the corresponding directories and I execute manually the command
 > ./bin/generate_events
after which I enable the showering with Pythia8. I tried to setttings for the PDF used in the showering: pdfcode=0 and pdfcode=1 but both fail at NLO.

Finally, it might be worth noting that I am using a slightly older version of MG (v2.4.31) to be able to use an ISR plugin (https://github.com/qliphy/MGISR, which I wasn't able to make it work wit the last version).

Thanks a lot for you help,
Cheers,
Romain

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
Xiaoran Zhao Edit question
Last query:
Last reply:
Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#1

Hi,

I guess that the issue is that some change need to be done at the shower interface to Pythia8, but I would not be able to tell which one needs to be done.

I have assigned this to our expert in this area but this is the first time that I assign him a question so let's hope that he will be notified/... so do not hesitate to ping me back if you do not receive an answer from him in a couple of days and I will ping him via other method.

However this is likely something that we have to change internally to our code and that will require an update.
Now this is not an issue for you since NLO computation are not compatible with the plugin anyway so while for LO computation you might be stuck to outdated version this should not be the case for NLO compuations

Cheers,

Olivier

Revision history for this message
RomainM (romain-madar) said : 		#2

Hello,

Thank a lot for your feedback - at least I didn't do anything wrong.

Also, I wasn't unsure about the validity of the ISRs together with the NLO computation, and this is now clear. For now, I'll probably only use the cross-section for NLO computation w/o ISR (since this should be correct even before parton showering), and stick to the LO prediction of events generation.

Thanks again,
Romain.

Revision history for this message
Xiaoran Zhao (zxrlha) said : 		#3

Hello,

I cannot find the version 2.4.31 you mentioned. So I just tried the version 2.9.2, and I successfully reproduced the bug you mentioned(of course, without that plugin).

After looking into the code, I realized that it is a bug related to numerical inaccuracy: in NLO mode of MG5_aMC, only 8 digits of four momenta are output into LHE file. In Pythia8 there is a check which requires 10^-9 precision and thus it failed.
To fix this, just modifing the file "handling_lhe_events.f" under directory "SubProcesses", in line 809, changing the output format for LHE files from:
 504 format(1x,i8,1x,i2,4(1x,i4),5(1x,e14.8),2(1x,e10.4))
to:
 504 format(1x,i8,1x,i2,4(1x,i4),5(1x,e24.17),2(1x,e10.4))

i.e. output 17 digits instead of 8 digits.
Then Pythia8 works fine.
Note that the LO mode of MG5_aMC output 11 digits instead, so pythia8 works fine.

I haven't test the plugin you are using since I'm facing some problems with it.

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#4

Thanks Xiaoran,

I have pushed your fix in the development version of the code:

https://bazaar.launchpad.net/~maddevelopers/mg5amcnlo/2.9.3/revision/311

Cheers,

Olivier

Revision history for this message
RomainM (romain-madar) said : 		#5

Many thanks Xiaoran and Olivier for the fix!

Can you help with this problem?

Provide an answer of your own, or ask RomainM for more information if necessary.

To post a message you must log in.