MadGraph v2.7.0 fails to reach target number of events

Asked by Mohamed Aly on 2020-03-09

I am attempting to calculate the interference between diagrams within a Manifest Left-Right symmetric model (MLRSM): https://drive.google.com/folderview?id=0BxMAGX_Tlpi9X0RUZW9tS2RaQ0E&usp=sharing

I am interested in the interference contributions to the process:
p p > mu+ mu+ j j

The final state in the model can be achieved via Majorana Neutrino mediated diagrams or via BSM-Higgs mediated diagrams. To generate the interefrence between the two terms, I added two new coupling orders to the model:

BSMHIGGS = CouplingOrder(name='BSMHIGGS',
 expansion_order = 99,
 hierarchy = 2 )

N = CouplingOrder(name='N',
 expansion_order = 99,
 hierarchy = 2 )

and used the following command in MG5:

generate p p > mu+ mu+ j j BSMHIGGS^2>1 N^2>1 @1
add process p p > mu+ mu+ j j BSMHIGGS^2>1 N^2>1 @2

My problem comes in when, for some combinations of theory model parameters, MadGraph fails to reach the required number of events (10000). The info message appears as shown below:

INFO: Combining runs
INFO: finish refine
INFO: Combining Events
INFO: fail to reach target 10000
  === Results Summary for run: run_01 tag: tag_1 ===

     Cross-section : 9.033e-10 +- 2.98e-12 pb
     Nb of events : 186

store_events
INFO: Storing parton level results
INFO: End Parton

I have not applied any cuts at all. The only changes I made to the param_card.dat were changing masses in the model:

set MH01 = 5000
set MH02 = 5000
set MH03 = 5000
set MHP1 = 5000
set MHP2 = 5000
set mhppr = 5000
set mhppl = 5000
set A02 = 5000
set MN5 = 7000
set MW2 = 4000

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
2020-03-09
Last reply:
2020-03-09

Hi,

>generate p p > mu+ mu+ j j BSMHIGGS^2>1 N^2>1 @1
>add process p p > mu+ mu+ j j BSMHIGGS^2>1 N^2>1 @2

I guess you made a typo above since those are twice the same line (or I miss the difference at least)

>I have not applied any cuts at all.

Are you sure that you do not have singularity (even integrable singularity)?

>My problem comes in when, for some combinations of theory model parameters, MadGraph fails to reach the required number of events (10000). The info message appears as shown below:

This is actually not surprising... The phase-space integration for interference is something that we do not know how to do in an efficient way. We are using the standard madevent technique to do such phase-space integration/event generation and as expected the machine learning layer (VEGAS) can compensate for some cases but not always (as usual machine learning is great to refine but if you are too bad to start with...)

In 2.7.1 [going to be release this week], we have a new (hidden) option of the run_card
hard_process if you set it on True, it will perform the computation with more event per iteration. This might allow VEGAS to better compensate in those cases.

Cheers,

Olivier

Mohamed Aly (moaly) said : #3

Hi Olivier,

Thank you for your quick reply.

I am not aware of any singularities, but is there a way I can identify singularities from the MadGraph output?

Hopefully with the new realese the hard_process feature helps. Until then, am I right to assume the cross-section produced is not trustworthy?

Cheers,
Mohamed

> , am I right to assume the cross-section produced is not
> trustworthy?

It is not necessarily wrong. but indeed not having the correct number of events is a sign of difficulty of the phase-space integrator and this means it is more likely than usual to face bias for the integrator.
Running multiple run with various seed can be a test if this is the result is stable or not.

Cheers,

Olivier

> On 9 Mar 2020, at 21:37, Mohamed Aly <email address hidden> wrote:
>
> Question #689230 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/689230
>
> Mohamed Aly posted a new comment:
> Hi Olivier,
>
> Thank you for your quick reply.
>
> I am not aware of any singularities, but is there a way I can identify
> singularities from the MadGraph output?
>
> Hopefully with the new realese the hard_process feature helps. Until
> then, am I right to assume the cross-section produced is not
> trustworthy?
>
> Cheers,
> Mohamed
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Can you help with this problem?

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

To post a message you must log in.