gridpack generation

Dear Andrew,

That parameter is completely ignored when creating the gridpack. The code basically creates the gridpack to allow for an 'infinite' number of events to be generated with it.

Best regards,
Rikkert

Then I am confused. Olivier wrote in

https://answers.launchpad.net/mg5amcnlo/+question/240906

post #15 that the cross section depends on the number of events that you generate.

One of the results of the gridpack generation is a cross section. Is this cross section meaningful, since 0 events are produced during gridpack generation.

One of the results of the gridpack generation is a cross section. Is this cross section meaningful, since 0 events are produced during gridpack generation.

Dear Andrew,

As I wrote above, when generating a gridpack, you are "only" interested in creating a code that allows for event generation by skipping the time-consuming initialization of the integration grids and all that. Remember that you are still doing a calculation that's only Leading Order accurate (which is at best only an order of magnitude estimate of the real cross section), and therefore the question on precision of the cross section doesn't really play a role when doing mass production for a gridpack.

Best regards,
Rikkert

Hi,

Let me try to rephrase my question.

When you do a gridpack generation, there is a gridpack produced and there is also a cross section produced (e.g. in the crossx.html file).

Is this cross section the correct LO cross section or is it meaningless?

Thanks.

Andrew

That is the correct LO cross section

Best regards,
Rikkert

Hi,

I would like to clarify my doubt.

From the discussion above I got that we can trust the cross-section given by the log file of the Gridpack.

What I understand from the discussion is that for the Gridpack generation it is not generating actual number of events. May be small number say 100 or 500. So, If the cross-section given in the Gridpack is got from the less statistics then I think we should not trust this value. Right?

The statistical error on that number should always be much smaller than that systematical error (scale uncertainty) for any QCD production. So I would not see the problem.

For the non QCD processes, the statistical error associated to that value is always associated so you can have an idea of the precision on that value but this should be quite good.

Cheers,

Olivier

Dear Olivier,

Thanks for your information.

You are mentioning about QCD process. What about the QED processes? As I am looking into the VBS QED process.

with regards,
Ram

Well you still have PDF and factorization scale for that process.

But independently the (stristical) error associate to the cross-section is reported such that you know the accuracy.

Cheers,

Olivier

> On 15 Nov 2020, at 03:05, Ramkrishna <email address hidden> wrote:
>
> Question #244207 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/244207
>
> Ramkrishna requested more information:
> Dear Olivier,
>
> Thanks for your information.
>
> You are mentioning about QCD process. What about the QED processes? As I
> am looking into the VBS QED process.
>
> with regards,
> Ram
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Let me add something here that might clarify a few things.

The precision in the cross section computed in the creation of the gridpack is independent from the number of events requested at this stage. This precision is a fixed value in the code (I can't remember the exact numerical value on the top of my head), but it's chosen such that it will be of subdominant importance in comparison to any other uncertainties in the calculation. Where the other uncertainties are typically theoretical systematic uncertainties, such as renormalisation and factorisation scale dependence, or statistical uncertainties due to generating only a few events with the gridpack, resulting in large fluctuation e.g. in tails of distributions.

Best,
Rikkert