precision for p p > b b~ at very high invariant mass

Hi,

Your cut are probably too strong for our phase-space integrator to converge.

Cheers,

Olivier

On 11 Jun 2015, at 22:07, Ezequiel Alvarez <email address hidden> wrote:

> New question #268048 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/268048
>
> Dear All,
>
> I am having a problem of precision and/or accuracy which seems that could be summarized in the following example.
>
> I generate p p > b b~ (LHC@13TeV) requiring very large invariant mass for the b-pair in the run_card.dat:
>
> 2500 = mmbb ! min invariant mass of a b pair
> 2600 = mmbbmax ! max invariant mass of a b pair
>
> After I generate 10^5 events I count in the unweighted_events.lhe file how many of the events were produced by initial states gg and how many by qqbar. What happens is that for some runs I get something reasonable as gg=85% and qqbar=15%, and for many other runs I get gg=0% and qqbar=100%. I stress that I'm generating 10^5 events. In fact, I have verified that the computed cross-section by Madgraph for each production channel has a enormous fluctuation in each run.
>
> I have read some previous posts in which there are variables as "req_acc_FO" for the accuracy, but I don't see this variable in my run_card.dat. (I guess that this variable is for NLO?)
>
> I guess that my problem may have to do with a systematic uncertainty in the PDF. In any case I'd like to have a way to get a number (and eventually an uncertainty) for this kinf of processes.
>
> All the best, Ezequiel.
>
>
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

HI Olivier,

is there any way to make it converge? Do I have to go to the hard code?

Thank you! Cheers, Ezequiel.

Hi Ezequiel,

I think that the easiest is that you implement your own Phase-space integrator.
and only use MG for the matrix-element (or you can use analytical formula)

Cheers,

Olivier

On 12 Jun 2015, at 02:51, Ezequiel Alvarez <email address hidden> wrote:

> Question #268048 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/268048
>
> Status: Answered => Open
>
> Ezequiel Alvarez is still having a problem:
> HI Olivier,
>
> is there any way to make it converge? Do I have to go to the hard code?
>
> Thank you! Cheers, Ezequiel.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.