zero cross section at high invariant mass cut

Hi,

They are no approximation, the problem is that your cut as so tight that the code fails to find any point passing the cuts.
After trying for 500k phase-space points, we decide to stop and returns zero.
This variable (max_events) is define in Source/dsample.f if you want to increase the associate value.

Having loose cut allows to find points and therefore allow the code to improve the various grid of integration
allowing to generate events in part of the phase-space which were not possible with the original grid.
This is a standard feature of Monte-Carlo method applying strong cut requires dedicated method of integration.

Except the simple change of max_try that I suggest above, all other method will be very difficult to perform and will require that you go deep into the code and understand it, in order to make some change. You will basically be on your own for that since I’m not expert on that part of the code and will require a week for me to master it enough to be able to help you.
But in short, you need to identify how the integration is done. In this case the integration is done like a T-channel but If I’m correct one of the variable of integration is anyway M_LL so it should be possible to implement a priori your cut to have a good efficiency.
This should be done in a way or another in
Source/dsample.f : main controller of the integration and handling of the grid
SubProcesses/genps.f: generation of the momenta following the random number provide by sample.f
Source/myamp.f : Not expert at all on this file, but this try to guess the dependence in the cut on the variable of integration.

Another method is to add a S-channel resonance where you want to integrate.
and then edit the code such that the integration is only using that channel of integration but do not include the contribution of the S-channel in the final result.
That method in top of editing the model only requires to change
matrix.f and on that one, I can really help.
The first thing to understand here is the notion of multi-channel which is describe here:
hep-ph/0208156
and also in this lecture:
https://cp3.irmp.ucl.ac.be/projects/madgraph/attachment/wiki/Lund2014/Lund_3_MCGeneration.pdf

Cheers,

Olivier

> On Oct 26, 2015, at 00:33, Mamutjan <email address hidden> wrote:
>
> New question #273051 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/273051
>
> Hi,
>
> I was trying to generate events for aa>e+e- at 13TeV with cut on invariant mass 2.8<M(e+e-)<3Tev and I got zero cross section. After I had weakened the cut to 1TeV<M<3TeV, I got non-zero result and was able to analyze events for the region I wanted.
>
> I noticed that MadGraph would estimate the cross section before it actually calculates the cross section for the phase space with cuts and matrix elements and in my case mentioned above the estimated cross section is zero. But when I analyzed the events that MG had generated with weaker cuts I see events in the region 2.8<M<3TeV.
>
> Why does MG give zero cross section for the region it would give non-zero result after weakening the cuts? Could you tell me where should I look for the codes that calculate the estimated and final cross sections?
> Thank you.
>
> Regards,
> Mamutjan.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.