Low efficiency for e+ e- > multijets

Hi,

Looks like you have pushed MG5_aMC to its limit.

> Why does MG5 terminate before reaching the requested number of events?

They are an hardcoded limit on the number of PS point that is used in order to reach the target.
The main motivation of such limit is to be able to stop the code in a decent amount of time even if the user ask us to integrate a divergency.
Such limitation can be change but I do not think that this is going to really help you.

> And is there a way to improve the efficiency without distorting the ycut distribution?

As usual, you should in this case change the phase-space parametrisation such that the integral is more flat in those variable.
But as you know this is certainly not something simple to do.

Sorry,

Olivier

On 02 Mar 2015, at 18:11, Bryan Webber <email address hidden> wrote:

> New question #263088 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/263088
>
> I am generating e+ e- > d d~ g g g to compute the (LO) 5 (kt-)jet rate at large log(1/ycut)~18, for comparison with various parton shower approximations. On the run card I set beam energies 500 GeV, min ej=min mjj = 0.1 GeV and request 10^4 events. But after 20 hours on my 2-core Mac I get only 66 unweighted events (1404 weighted). Why does MG5 terminate before reaching the requested number of events? And is there a way to improve the efficiency without distorting the ycut distribution?
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

OK thanks. I will just do several runs with different random number
seeds and combine the events. Bryan

On 03/03/2015 01:26, Olivier Mattelaer wrote:
> Your question #263088 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/263088
>
> Status: Open => Answered
>
> Olivier Mattelaer proposed the following answer:
> Hi,
>
> Looks like you have pushed MG5_aMC to its limit.
>
>> Why does MG5 terminate before reaching the requested number of events?
> They are an hardcoded limit on the number of PS point that is used in order to reach the target.
> The main motivation of such limit is to be able to stop the code in a decent amount of time even if the user ask us to integrate a divergency.
> Such limitation can be change but I do not think that this is going to really help you.
>
>> And is there a way to improve the efficiency without distorting the
> ycut distribution?
>
>
> As usual, you should in this case change the phase-space parametrisation such that the integral is more flat in those variable.
> But as you know this is certainly not something simple to do.
>
> Sorry,
>
> Olivier
>
>
> On 02 Mar 2015, at 18:11, Bryan Webber
> <email address hidden> wrote:
>
>> New question #263088 on MadGraph5_aMC@NLO:
>> https://answers.launchpad.net/mg5amcnlo/+question/263088
>>
>> I am generating e+ e- > d d~ g g g to compute the (LO) 5 (kt-)jet rate at large log(1/ycut)~18, for comparison with various parton shower approximations. On the run card I set beam energies 500 GeV, min ej=min mjj = 0.1 GeV and request 10^4 events. But after 20 hours on my 2-core Mac I get only 66 unweighted events (1404 weighted). Why does MG5 terminate before reaching the requested number of events? And is there a way to improve the efficiency without distorting the ycut distribution?
>>
>> --
>> You received this question notification because you are an answer
>> contact for MadGraph5_aMC@NLO.

On approach which might be more efficient, is to use the event generated by the parton-shower and reweight them by the LO matrix-element.

Cheers,

Olivier