Different cross section and event generation with different sde_strategy?

Asked by Ruoquan Wang on 2021-05-13

I meet some problem in event generation and calculating cross section.

My process is
mu- e- > mu- e- vec
via QED process, where the vec is a massive vector gauge boson that couples with muon and electron.

The diagram is purely QED with a t-channel photon.

This diagram should have no singularity. However, when I try to generate 10,000 events and calculating cross section with sde_strategy=1, I found that if the mass of the vector gauge boson is small, the event generation is slow or event number cannot reach the target. hard_survey cannot save the situation. (large vector boson mass is totally fine)

If I set sde_strategy=2, then it quickly generates 10,000 events and gives me a cross section. However, this cross section is different from the previous one. The previous one is around 3 times larger than this one. I do not believe this is within the uncertainty. The difference of the cross section even exists for the large boson mass case.

Which result should I believe? If I should believe the cross section with the old integral strategy, how can I generate enough number of events? I already tried the parameter hard_survey, or iseed, to the old strategy, but no one works.

Any idea about this?

Question information

English Edit question
MadGraph5_aMC@NLO Edit question
No assignee Edit question
Solved by:
Olivier Mattelaer
Last query:
Last reply:
Revision history for this message
Best Olivier Mattelaer (olivier-mattelaer) said :

Well your process seems exactly why I have spent 10 years of my time to devellop sde_strategy=2.

In those 10 years I have tried a lot of strategy (including hard_survey, t channel ordering, dumping t-channel, ...) and after 10 years I finally find one satisfactory one: sde_strategy 2.

So the feeling here is that
1) you can not fix the situation for sde_strategy =1 (at least I would not be on that before 2030)
2) I would trust the cross-section from sde_strategy=2

I would suggest to play with the t channel ordering option introduced in the same paper as sde_strategy, this should help you to build confidence on the fact that the integration is stable and correct.



Revision history for this message
Ruoquan Wang (isaac95) said :

Thanks a lot for your answer! This solved my problem.

According to what I tried, the parameter tmin_for_channel seems to give a similar result. But I found no references and discussions about this parameter. Could you give some hint about this parameter?

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :

This parameter changes the multi-channel weight (in a similar as sde_strategy).
By adding an cut veto on T-channel variable to forbid the associated channel to cover the non-singular region.

the idea is that if t is the invariant mass square of the t-channel
if t/stot < 2* tmin_for_channel then the channel weight is set to zero
if 2* tmin_for_channel < t/stot < tmin_for_channel then a linear suppression is used
if tmin_for_channel < t/stot , the multi-channnel weight is the one defined by the sde_strategy.



Revision history for this message
Ruoquan Wang (isaac95) said :

Great! Thanks a lot.

Revision history for this message
Ruoquan Wang (isaac95) said :

Thanks Olivier Mattelaer, that solved my question.