subprocesses with invalid results

[Solano Felicio]

I am trying to run the parton-level process p p > t t~ h h on the Minimal Composite Higgs Model (MCHM5), with an UFO I generated with FeynRules. Everything runs smoothly (I can generate the diagrams, output, and launch) up until the refinement step, where MG5 crashes with the error in the attached log file. I looked into the results.dat for some subprocesses and they contain the message "end code not correct 1" instead of the results.

[Olivier Mattelaer (olivier-mattelaer)]

The usefull error message here is
 Problem in the multi-channeling. All amp2 are zero but not the total matrix-element

The amp2 are the definition of the multi-channel weight, their sum should be normalised to one, which is not possible if some phase-space point have all the amp2 set to zero.
This indicates that the phase-space integration is badly setup.
Given how early the crash occur, this is likely not a local issue in the phase-space but a global issue that the phase-space integration is wrongly setup.

This is possible for your process since I do observe coupling set to zero:
        GC_48 = 0.00000E+00 0.00000E+00
        GC_51 = 0.00000E+00 0.00000E+00
         GC_9 = 0.00000E+00 0.00000E+00
        GC_10 = 0.00000E+00 0.00000E+00
        GC_46 = 0.00000E+00 0.00000E+00
        GC_50 = 0.00000E+00 0.00000E+00
You should (normally) have seen a warning associated to the presence of zero coupling.
One risk of having such zero coupling is that the phase-space parametrization are based on such coupling and therefore all amp2 are proportional to such coupling and lead to the above crash.

Couple of possible solution:
1) use a restricted model to remove, from the start, all diagrams which correspond to a zero coupling
2) change the phase-space strategy to "sde_strategy=2", While that strategy might not be optimal for this process, it will certainly avoid the above issue.

Cheers,

Olivier

[Solano Felicio (sfelicio)]

Thank you, both solutions worked. First one is better as the phase-space strategy sde_strategy=2 results in really slow computations.

Cheers,
Solano