Recommendation for generating QCD Multijet events (many final-state legs)

Asked by Anthony Badea on 2021-05-11

Hello,

I'm interested in generating events with 4, 5, or 6 partons in the final state at matrix element level (only madgraph, no pythia).

Using MadGraph 2.7.3 I tried generating 10k events of the following processes on a 28 core machine.

generate p p > j j j j
add process p p > j j j j j
add process p p > j j j j j j

I let it run for >1 week and it did not finish, so I tried only gluon events using both Madgraph 2.7.3 and 3.1.0.

Generating with MadGraph 2.7.3
Process, NEvents, Run time
gg-> 4g , 10k , 1 hour
gg->5g , 1 , 2.5 hours
gg->6g , 1 , 3 hours (didn’t finish)

Generating with MadGraph 3.1.0.
Process, NEvents, Run time
gg-> 4g , 100 , 40 seconds -> scale by 100 roughly 1 hour
gg->5g , 1 , ~2.5 hours
gg->6g , 1 , >3 hours (didn’t finish)

What is the recommendation for generating such events with many final state legs, specifically p p > 4,5, or 6 j?

Thank you very much,
Anthony

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Anthony Badea
Solved:
Last query:
Last reply:
Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#1

Hi,

6 jet is currently out of reach.
For the four jet, I can do them on a 8 core laptop (10K event overnight) so that one is not a problem and for 5 jet my laptop is likely not enough, so I would advise to use a cluster for that.

Cheers,

Olivier

Revision history for this message
Anthony Badea (badeaa3) said :
#2

Hi Oliver,

Many thanks for the reply!

Gotcha so I'll go forward with generating 5 jet events on a cluster.

Is there a resource to read to understand the limitations of madgraph (like 6 jet being out of reach) and why?

Best,
Anthony

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#3

MG5aMC computes the full color-matrix and this computations scales
like ((Njet+1)!)**2 (check any QCD paper on that topic) in top of the number of processes that grows like 6**(Njet)
so going from 5 jet to 6 jet makes the computation
7*7*6 ~300 times more complicated than the 5 jet one.
(and this does not even take into account the additional helicity (times 2) and the growing number of Feynman diagram (which is not that relevant).

Cheers,

Olivier

> On 16 May 2021, at 16:05, Anthony Badea <email address hidden> wrote:
>
> Question #697012 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/697012
>
> Anthony Badea posted a new comment:
> Hi Oliver,
>
> Many thanks for the reply!
>
> Gotcha so I'll go forward with generating 5 jet events on a cluster.
>
> Is there a resource to read to understand the limitations of madgraph
> (like 6 jet being out of reach) and why?
>
> Best,
> Anthony
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Anthony Badea (badeaa3) said :
#4

Hi Oliver,

Thanks a lot for the information. This solves my question.

Best,
Anthony