Cross sections of e+ e- > z h > mu+ mu- b b~

Which PDF are you using? (Both for LO an NLO?)
How did you compute LO, it is within the NLO framework? or within a dedicated LO framework? in the second case many choice (scale/cut/...) can be different.

Cheers,

Olivier

Hi Olivier,

Since this is a e+e- process, I do not consider the PDF at present. The lpp1 and lpp2 are zero in both cases. The pdlabel values are their default value (nn23lo1 for LO and nn23nlo for NLO).

I run the LO within a dedicated framework.

In both LO and NLO cases, I remove the default cuts for jets and leptons, and set a fixed scale at Z mass.

Cheers,
Zhijie

Hi Zhijie,
I am not able to reproduce your issue, at least not with with 3.5.2 (can you please check with that branch?).
I did two fixed order runs with different accuracies, and two event-generation runs at NLO, and all look consistent to me:
FO:
4.193e-03 +- 3.0e-05 pb
4.241e-03 +- 1.1e-05 pb

Events:
10k:
      Total cross section: 4.265e-03 +- 4.4e-05 pb
   --------------------------------------------------------------
      Scale variation (computed from LHE events):
          Dynamical_scale_choice -2 (envelope of 9 values):
              4.177e-03 pb +6.1% -7.6%
100k:
      Number of events generated: 100000
      Total cross section: 4.229e-03 +- 2.9e-05 pb
   --------------------------------------------------------------
      Scale variation (computed from LHE events):
          Dynamical_scale_choice -2 (envelope of 9 values):
              4.216e-03 pb +5.7% -7.1%

Also, with this setup, we get the LO cross section of about 6.78e-3pb

The only thing I could see in the run_card is an invariant mass of the lepton pair of at least 30 GeV, which should be irrelevant since there is only the z boson...

The only thing peculiar about this process is that you have a lot of negative weights, being the abs(xsect) > 2 xsec...

Let me know

Thanks,

Marco

Hi Marco,

Thanks a lot. I can get a fixed order cross section around 4.2e-03 pb with the v3.5.2.

But the cross section was changed a lot when I switched on the parton shower with Pythia8.

Moreover, the cross section is negative when I generate 10K events with parton shower.

Cheers,
Zhijie

Hi,
indeed, with pythia8, I get after the first step:
      Intermediate results:
      Random seed: 40
      Total cross section: 1.381e-03 +- 1.1e-03 pb
      Total abs(cross section): 3.884e-02 +- 1.1e-03 pb

Which means that the fraction of negative weights is just huge.
When you unweight events, and if you have not many of them, indeed the total cross section may be negative (although it should be compatible with the true xsection if you estimate the statistical error).
I do not know what to suggest here, I include Rikkert to the thread so that he may be able to comment more about that.

Best wishes,

Marco

Hi again,
the run has just finished.
I get this
      Intermediate results:
      Random seed: 40
      Total cross section: 3.457e-03 +- 4.0e-04 pb
      Total abs(cross section): 4.077e-02 +- 3.9e-04 pb
quite far from the cross section we had before (still within 2 sigma), and plagued by a lot of neg. weights.
Marco

Hi Marco,

I fixed the random seed to 40 as you, and I can get a cross section 0.0004319 ± 0.0013 pb. It seems good.

But if I choose another random seed (44 in this case), I get a cross section 0.003125 ± 0.00039 pb.

Is there any method to solve the negative weights problem?

Cheers,
Zhijie

Dear Zhijie,
there are two ways to reduce the negative weights:
the first is to reduce the shower scale, by e.g. setting shower_scale_factor =0.5 (instead of the default of 1)
The second one is to use the new method, so-called MC@NLO-Delta, there should be a flag in the run_card
https://inspirehep.net/literature/1783017
I am not sure this has ever been tested for lepton collisions, but I don’t see any problem specific to these…

About your results, they are not totally inconsistent (given how much one can trust the numerical-integration error…)
Let me know if either of the two methods above works.

Best,

Marco

> On 25 Sep 2023, at 18:00, Zhijie Zhao <email address hidden> wrote:
>
> Question #707920 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/707920
>
> Status: Answered => Open
>
> Zhijie Zhao is still having a problem:
> Hi Marco,
>
> I fixed the random seed to 40 as you, and I can get a cross section
> 0.0004319 ± 0.0013 pb. It seems good.
>
> But if I choose another random seed (44 in this case), I get a cross
> section 0.003125 ± 0.00039 pb.
>
> Is there any method to solve the negative weights problem?
>
> --
> You received this question notification because you are assigned to this
> question.

Dear Marco,

It seems better now. I think it is sufficient to solve my problem.

Thank you very much.

Cheers,
Zhijie