Optimizer detects zero coupling

Asked by Amin Aboubrahim on 2021-04-12

Hi Olivier,

I am running this process in MG5 v3.1.0:

import model MSSM_SLHA2-no_b_mass -name

define eweakino = n2 x1+ x1-

generate p p > eweakino eweakino @0
add process p p > eweakino eweakino j @1
add process p p > eweakino eweakino j j @2

The event generation is slow and at some point the following message is shown:

Working on SubProcesses
INFO: Compiling for process 1/21.
INFO: P2_qq_n2n2qq
WARNING: The optimizer detects that you have coupling evaluated to zero:
GC_21 GC_57 GC_73 GC_102 GC_56 GC_72 GC_675 GC_690
This will slow down the computation. Please consider using restricted model:
https://answers.launchpad.net/mg5amcnlo/+faq/2312
INFO: Compiling for process 2/21.
INFO: P2_qq_x1px1mqq

A restriction card is already in use so I don't understand what this warning is saying. I also used the default param_card just in case the one I'm using is erroneous but still the same problem.

Thank you,
Amin

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
Last query:
Last reply:
Amin Aboubrahim (amin83) said : #1

As a follow-up:

I let it run overnight and it did eventually finish. So I am trying now with gridpacks to see if I can generate events faster. It is running now but with this warning:

Working on SubProcesses
INFO: P2_qq_n2n2qq
INFO: P2_gg_n2n2qq
Note: The following floating-point exceptions are signalling: IEEE_DENORMAL
Note: The following floating-point exceptions are signalling: IEEE_DENORMAL
Note: The following floating-point exceptions are signalling: IEEE_DENORMAL
INFO: P2_gq_n2n2gq
INFO: P2_qq_n2n2gg

Is that something to worry about?

Best,
Amin

Interesting,

If you look at the value of such couplings at the python level:
(display couplings)

        GC_21 = mdl_complexi*mdl_G__exp__2*mdl_I12x36 + mdl_complexi*mdl_G__exp__2*mdl_I13x36 = 8.80959233340306e-09j
        GC_57 = -(mdl_complexi*G*mdl_I51x36) - mdl_complexi*G*mdl_I52x36 = -3.3987396230727995e-09j
        GC_73 = -(mdl_complexi*G*mdl_I5x36) - mdl_complexi*G*mdl_I6x36 = -7.234522592103332e-09j
        GC_102 = mdl_complexi*mdl_G__exp__2*mdl_I74x63 + mdl_complexi*mdl_G__exp__2*mdl_I75x63 = 4.13869893911567e-09j
        GC_72 = (mdl_ee*mdl_complexi*mdl_I5x36)/3. + (mdl_ee*mdl_complexi*mdl_I6x36)/3. = 6.206612446213455e-10j
        GC_675 = mdl_ee*mdl_complexi*mdl_VV2x1*mdl_conjg__VV1x1 + mdl_ee*mdl_complexi*mdl_VV2x2*mdl_conjg__VV1x2 = -2.3189050818439583e-09j
        GC_690 = mdl_ee*mdl_complexi*mdl_VV1x1*mdl_conjg__VV2x1 + mdl_ee*mdl_complexi*mdl_VV1x2*mdl_conjg__VV2x2 = -2.318905109599534e-09j

So they are indeed not removed by the restriction file.

However at fortran level they are evaluated at zero :
        GC_57 = 0.0000000E+00 0.0000000E+00
        GC_73 = 0.0000000E+00 0.0000000E+00
And therefore the optimiser complains about it.
The optimiser will remove (after the warning) all the evaluation of amplitude associated to those coupling in warning.
However the phase-space integration strategy will not be modified accordingly (which means that you will likely gain speed if you can remove such couplings at the model level via modifying the model (at the Feynrules level).

Cheers,

Olivier

> On 13 Apr 2021, at 00:20, Amin Aboubrahim <email address hidden> wrote:
>
> New question #696538 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/696538
>
> Hi Olivier,
>
> I am running this process in MG5 v3.1.0:
>
> import model MSSM_SLHA2-no_b_mass -name
>
> define eweakino = n2 x1+ x1-
>
> generate p p > eweakino eweakino @0
> add process p p > eweakino eweakino j @1
> add process p p > eweakino eweakino j j @2
>
> The event generation is slow and at some point the following message is shown:
>
> Working on SubProcesses
> INFO: Compiling for process 1/21.
> INFO: P2_qq_n2n2qq
> WARNING: The optimizer detects that you have coupling evaluated to zero:
> GC_21 GC_57 GC_73 GC_102 GC_56 GC_72 GC_675 GC_690
> This will slow down the computation. Please consider using restricted model:
> https://answers.launchpad.net/mg5amcnlo/+faq/2312
> INFO: Compiling for process 2/21.
> INFO: P2_qq_x1px1mqq
>
> A restriction card is already in use so I don't understand what this warning is saying. I also used the default param_card just in case the one I'm using is erroneous but still the same problem.
>
> Thank you,
> Amin
>
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Amin Aboubrahim (amin83) said : #3

Hi Olivier,

Thanks for the answer.
The model I'm using is the built-in MSSM that comes with Madgraph. I don't think I have seen this behavior in previous versions.
Is it possible to fix the available UFO files rather than going back to FeynRules?

Best,
Amin

> I don't think I have seen this behavior in previous versions.

The check is indeed quite new (since 2.9.0) before it was sub-optimal (even worse than now actually) but this was not notified to the user.

> Is it possible to fix the available UFO files rather than going back to FeynRules?

If you change the analytical formula for the above coupling to '0.0'
then the restriction will remove those coupling and all the associated interactions.

Cheers,

Olivier

Amin Aboubrahim (amin83) said : #5

Thanks Olivier Mattelaer, that solved my question.