MadGraph5_aMC@NLO

calculation fails for certain cuts

Asked by Roberto Franceschini

Hi
  I am computing p p > e+ ve mu- mu+ [QCD]
and I have a cut on the pT of the sum of mu+ and mu- four-vectors.
For low values of this cut, e.g. pT>200 GeV, I get results.
However if I set this cut to above 400 GeV there are errors as in the following:

INFO: Idle: 16, Running: 16, Completed: 16 [ 3m 0s ]
WARNING: program /afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_uxd_emvexmupmum/ajob1 6 F 0 0 launch ends with non zero status: 1. Stop all computation
INFO: Idle: 16, Running: 15, Completed: 17 [ 3m 2s ]
INFO: Idle: 16, Running: 14, Completed: 18 [ 3m 2s ]
INFO: Idle: 16, Running: 13, Completed: 19 [ 3m 2s ]
INFO: Idle: 16, Running: 12, Completed: 20 [ 3m 2s ]
/afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_dux_emvexmupmum/ajob1: line 34: 869798 Terminated ../madevent_mintMC > log.txt < input_app.txt 2>&1
date: write error: Broken pipe
INFO: Idle: 16, Running: 11, Completed: 21 [ 3m 2s ]
/afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_dux_emvexmupmum/ajob1: line 34: 869811 Terminated ../madevent_mintMC > log.txt < input_app.txt 2>&1
date: write error: Broken pipe
INFO: Idle: 16, Running: 10, Completed: 22 [ 3m 2s ]
INFO: Idle: 16, Running: 9, Completed: 23 [ 3m 2s ]
INFO: Idle: 16, Running: 8, Completed: 24 [ 3m 3s ]
/afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_uxd_emvexmupmum/ajob1: line 34: 870287 Terminated ../madevent_mintMC > log.txt < input_app.txt 2>&1
date: write error: Broken pipe
INFO: Idle: 16, Running: 7, Completed: 25 [ 3m 3s ]
/afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_uxd_emvexmupmum/ajob1: line 34: 870328 Terminated ../madevent_mintMC > log.txt < input_app.txt 2>&1
date: write error: Broken pipe
INFO: Idle: 16, Running: 6, Completed: 26 [ 3m 3s ]
/afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_uxd_emvexmupmum/ajob1: line 34: 870425 Terminated ../madevent_mintMC > log.txt < input_app.txt 2>&1
date: write error: Broken pipe
INFO: Idle: 16, Running: 5, Completed: 27 [ 3m 3s ]
/afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_uxd_emvexmupmum/ajob1: line 34: 870679 Terminated ../madevent_mintMC > log.txt < input_app.txt 2>&1
date: write error: Broken pipe
INFO: Idle: 16, Running: 4, Completed: 28 [ 3m 3s ]
/afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_uxd_emvexmupmum/ajob1: line 34: 870761 Terminated ../madevent_mintMC > log.txt < input_app.txt 2>&1
date: write error: Broken pipe
INFO: Idle: 16, Running: 3, Completed: 29 [ 3m 3s ]
/afs/cern.ch/work/r/rfrances/www/MG5_aMC_v2_5_5/DibosonPrecision/NLO/pp2elmumu_EWdim6NLO_NLO_duplicate_B/SubProcesses/P0_uxd_emvexmupmum/ajob1: line 34: 870774 Terminated ../madevent_mintMC > log.txt < input_app.txt 2>&1
date: write error: Broken pipe
INFO: Idle: 16, Running: 2, Completed: 30 [ 3m 3s ]
INFO: Idle: 16, Running: 0, Completed: 32 [ 3m 3s ]

I tried to make 100K events instead of 10K hoping that the phase-space of the intial grids were sampled more extensively in the tails, but it did not help.
Any idea what could go wrong here?

Thanks for your help.
Best,
Roberto

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
Rikkert Frederix Edit question
Solved by:
Rikkert Frederix
Solved:
Last query:
Last reply:
Revision history for this message
Best Rikkert Frederix (frederix) said : 		#1

Dear Roberto,

The problem is that in order to have an efficient phase-space generation, it is assumed that the phase-space points have just enough energy to be produced: that where the most of the events will be in the end. Unfortunately, the code does know about user-defined cuts in cuts.f. Hence, the code throws many phase-space points in which it thinks most of the events will be in the end, but none of these points pass your user-defined cut. (It tries several 100000 before calling it quits and saying it's too difficult). There are two ways out.

1. Simply lower your cuts and generate more events so that after the more stringent analysis cuts (i.e., the one with 400 GeV) you still have a sizeable event sample.

2. Hack the code, in order to force it to have more events at higher energies. In the <YourProc>/SubProcesses/setcuts.f file, around line 493 you find something like:
            taumin_s(iFKS,ichan)=
     & max(taumin_j(iFKS,ichan),taumin_s(iFKS,ichan))

Before this statement set
       taumin_s(iFKS,ichan)=max(taumin_s(iFKS,ichan),400d0)

where, in this case, I assume that 400 GeV is (roughly) the minimum energy that's needed for an event to pass the cuts. (Better set this value on the low side, since putting it harder than strictly needed might bias the final results).

Best,
Rikkert

Revision history for this message
Roberto Franceschini (franceschini-roberto) said : 		#2

Hi,
  in the end I went for solution 1, pushed a bit with the statistics, dealt with bigger output but works.
Thanks for the suggestions and for confirming this is somehow expected to happen.
Best,
Roberto

Revision history for this message
Roberto Franceschini (franceschini-roberto) said : 		#3

Thanks Rikkert Frederix, that solved my question.