MadGraph5_aMC@NLO

FxFx diphoton production in MG2.9 with m_yy cuts

Asked by Ana Rosario Cueto Gomez

Dear experts,

we have been trying for some time now to have a diphoton production FxFx, while we understand the new features of this process in MG3, we have tried to set it up using MG2.9.5.

The process works without problems when we do an inclusive generation. However, when we make m_yy slices, the low myy ones work ok (sometimes need a change of seed to avoid errors in the integration), but the high myy ones (myy>175 GeV) they continuously fail. The crash is always the same, while setting up the grids it throws something like:

program /userdata3/acueto/MadGraphAthena/Diphoton_2j_FxFx_masswindow/Otherslices/rungrid_175_2000/PROCNLO_loop_sm-no_b_mass_0/SubProcesses/P2_gdx_aadxg/ajob1 2 F 0 0 launch ends with non zero status: 1. Stop all computation

the process in which it fails can vary depending on the seed. And when I look into the SubProcesses/P2_gdx_aadxg folder. In this case, the folder GF2 log.txt file contains the following error:

 ERROR in genps_fks.f: cannot set s-channel without BW -2 NaN 1936.0000000000000
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_DIVIDE_BY_ZERO
STOP 1

We have tried to impose dphi_yy cuts or restrict the QED order, but none of them works. We saw also this ticket:
https://bugs.launchpad.net/mg5amcnlo/+bug/1918031
but we did not see any possible solution.

Do you have any ideas on how to solve this issue? or at least have a clue about why this happens? Also, let me know which are the relevant pieces of code that would be interesting for you to look at (I did not want to fill the report with many irrelevant information).

Thanks in advance,
Ana

Question information

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

Dear Ana,

Clearly, there is somewhere a numerical inaccuracy in the code that results in a not-a-number ('NaN'). Assuming that you have proper IR-safe cuts (i.e., with proper photon isolation criteria and the like), it should be okay to simply skip the point and let the code continue.

Around line 1930 of SubProcesses/genps_fks.f changing the if statement from

         if (s(i).lt.smin .or. s(i).gt.smax) then

to

         if (s(i).lt.smin .or. s(i).gt.smax .or. s(i).ne.s(i)) then

should prevent the code from stopping.

Alternatively, slicing the phase-space is not really the recommended way of increasing the number of events in tails of distributions. What typically works better is to bias the event generation by unweighting the events against a non-constant function: for more details, look towards the bottom of the SubProcesses/cuts.f file, where you can define this 'bias_weight_function'.

Best,
Rikkert

Can you help with this problem?

Provide an answer of your own, or ask Ana Rosario Cueto Gomez for more information if necessary.

To post a message you must log in.