Z+bb~ generation at NLO and 5 flavor

Hi Alexandre,
this depends on the observable you want to look at.
If, as you said, you are interested in final states with 2 b quarks, i think you can forget about the process with 0 extra jets.
however, if you want to create a FxFx (or UNLOPS) merged sample (therefore including different final state multiplicities), then i think you may need to include also the 0 jets sample, as currentrly the merging works only if you have no jets at the process with the lowest multiplicity.

Can you tell us exactly what you want to generate?

Thanks,

Marco

Hello,

Thanks for your answer!

We already produced a NLO 4-flavors Z+bb sample and we want to compare now with a NLO 5-flavors.
To produce this 5 flavors sample, we wanted to use amc@NLO+Pythia8 with FxFx merging indeed.
We will then look at variables such that: invariant mass of the two b's, and angular distributions.

What do you think?

Cheers,
Alexandre

Hi Alexandre,
i have also subscribed Rikkert to this discussion.
I think it may be sensible to go for a 5 flavour merged sample, yet it has to be quite a large sample, as typically the fraction of events with b quarks is small.
I remember having produced unmerged samples for ZJ and ZJJ for previous analyses, i can try to find them if you want (again, those only contain a given multiplicity).

Let me know,

Marco

Hi Marco,

Actually, we are still working at 7 TeV in the center of mass.
At which energy did you produced them?

Alexandre

i think 7 gev too

On 02 Dec 2014, at 14:41, Alexandre Mertens <email address hidden> wrote:

> Question #258463 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/258463
>
> Alexandre Mertens gave more information on the question:
> Hi Marco,
>
> Actually, we are still working at 7 TeV in the center of mass.
> At which energy did you produced them?
>
> Alexandre
>
> --
> You received this question notification because you are a direct
> subscriber of the question.

Ah,

I just discussed with Ludivine Ceard, and she told me that the 7 TeV sample you produced are not anymore compatible with LHE interface in cmssw. Do you know why?

No i don't, the only thing i could suspect was the format of the scale/pdf
uncertainties but it looks like it is the same as present files...
cheers,

Marco

Marco Zaro

On Tue, Dec 2, 2014 at 2:56 PM, Alexandre Mertens <
<email address hidden>> wrote:

> Question #258463 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/258463
>
> Status: Answered => Open
>
> Alexandre Mertens is still having a problem:
> Ah,
>
> I just discussed with Ludivine Ceard, and she told me that the 7 TeV
> sample you produced are not anymore compatible with LHE interface in
> cmssw. Do you know why?
>
> --
> You received this question notification because you are a direct
> subscriber of the question.
>

Hey Marco,

Sorry to bother again, we are using Rivet to compare different generators. So it is important to have "the best" possible generation procedure with amc@NLO and Pythia8. That's why we thought of a NLO merged sample with FxFx merging.

So can you confirm that the merging works only if you have no jets at the process with the lowest multiplicity?
And if yes, how would you generate this sample, taken into account that our deadline are within two weeks ?

Cheers,

Alexandre

Ciao Alexandre,

yes, i confirm that in order to have the merging working, you must not have
jets in the lowest multiplicity sample.
i think two weeks is still a reasonable amount of time, you can generate
10M events in a merged sample and see if the statistics is enough.

However, note that if you are interested in observables which have 2 b's or
b jets, i am not sure that a merged sample will be better than just the Zjj
sample alone...

Rikkert, can you comment on this?

Thanks,

Marco

Marco Zaro

On Tue, Dec 2, 2014 at 4:01 PM, Alexandre Mertens <
<email address hidden>> wrote:

> Question #258463 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/258463
>
> Status: Answered => Open
>
> Alexandre Mertens is still having a problem:
> Hey Marco,
>
> Sorry to bother again, we are using Rivet to compare different
> generators. So it is important to have "the best" possible generation
> procedure with amc@NLO and Pythia8. That's why we thought of a NLO
> merged sample with FxFx merging.
>
> So can you confirm that the merging works only if you have no jets at the
> process with the lowest multiplicity?
> And if yes, how would you generate this sample, taken into account that
> our deadline are within two weeks ?
>
>
> Cheers,
>
> Alexandre
>
> --
> You received this question notification because you are a direct
> subscriber of the question.
>

Hello,

The best predictions in the 5FS are indeed obtained using the FxFx merging. Unfortunately, this means that you have to generate many events to have enough events with 2 b-jets. You can generate separate samples for Z+0j, Z+1j and Z+2j (including FxFx merging) and shower them separately. This way, you can generate more events for Z+2j than for the other multiplicities (which will then have smaller weights) so that the statistics for the multi-jet multiplicity is enhanced.

Best,
Rikkert