Difference for photon-induced dilepton production b/w MG5 2.6/2.7.X vs 2.9.3

Yes we have found the same issue at the begin of this month.
This should be fixed in the next Long Term Stable release (2.9.5) which should be release soon (when I'm back from holliday)
you can already test that version via the command
bzr branch lp:~maddevelopers/mg5amcnlo/LTS_dev

This version will also allow more control on the choice of the scale with the possibility to go back to dynamical scale

Cheers,

Olivier

> On 19 Aug 2021, at 21:05, Kristin Lohwasser <email address hidden> wrote:
>
> New question #698432 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/698432
>
> Dear all,
>
> we have noticed an about 10% change of cross section in a a > l+ l- going from MG5 2.7.X to 2.9.3 and a much more steeply falling off pT(ll) spectrum and would like to understand, where this is coming from.
>
> We set fixed factorization scale (as needed for the newer MG5 version):
> 'dsqrt_q2fact1' : 2.,
> 'dsqrt_q2fact2' : 2.,
> 'fixed_fac_scale' : 'T',
>
> varying the scale from 2. -> 20. or even 200. does not change the results, so we are wondering, if you have any idea of what else has changed that could be causing this.
>
> Thanks & Best
> Kristin
>
>
>
>
>
>
> Some T-channel width have been set to zero [new since 2.8.0]
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Actually, I realize that you do not specify for which type of collider you are doing the stuff.

So maybe we were not talking about the same issue.

If I run the following script:

generate a a > l+ l-
output
launch
set lpp1 3
set lpp2 3
set fixed_fac_scale T
set dsqrt_q2fact1 2
set dsqrt_q2fact2 2
launch
set dsqrt_q2fact1 200
set dsqrt_q2fact2 200

I do have very different cross-section.
Cross-section : 24.4 +- 0.07128 pb
Cross-section : 43.87 +- 0.1175 pb

I have actually printed the cutoff value for both the ee collision case and the pp case and the scale is correctly set to the valued picked by the script above.

So I guess that I would need that you provide the list of command that you pass to MG5 to understand what your issue is.

Cheers,

Olivier

Dear Olivier,

oh sorry about that. This is in general valid for a a > l+ l- with LPP1/2 = 2.

The commands are as follows below. (I have extracted the relevant ones, as this is within the ATLAS framework. I should note, that I did not set cut_decays T/F consistently, but double checked, that this does not make a difference (at least in MG5 v2.9.5).

Let me know if you have further questions.

Best
Kristin

-------- MG5 2.7.X :
generate a a > l+ l-

set lpp1 2
set lpp2 2

-------- MG5 2.9.3:
generate a a > l+ l-

set lpp1 2
set lpp2 2
set dsqrt_q2fact1 2.
set dsqrt_q2fact2 2.
set fixed_fac_scale T

-------- MG5 2.9.5:
generate a a > l+ l-

set lpp1 2
set lpp2 2
set fixed_fac_scale1 F,
set fixed_fac_scale2 F
set ptl 3.5

But which scale do you want? a fix scale computation or a dynamical one?
If it is a dynamical one, which dynamical_scale_choice did you choose?

Cheers,

Olivier

Dear Olivier,

we use the default "dynamical_scale_choice" in those cases, where the scale is dunamical, i.e.:

-1 = dynamical_scale_choice ! Choose one of the preselected dynamical choices.

We do not have an a priori preference (and would be open to suggestions).

So far, we have compared the dilepton-sample to H7 and LPAIR samples, that both have the pT spectrum of the MG 2.7.3 version.

Best
Kristin

Hi Kristin,

I have run some comparison at the cross-section level between 2.9.7 and 2.7.3 for
generate a a > l+ l-
output
launch
set lpp1 2
set lpp2 2
set dsqrt_q2fact1 2.
set dsqrt_q2fact2 2.
set fixed_fac_scale T
launch
set dsqrt_q2fact1 200.
set dsqrt_q2fact2 200.
launch
set dsqrt_q2fact1 2000.
set dsqrt_q2fact2 2000.

and got consistent (order 1%) cross-section for the three cases between the two versions (and a very small dependency on that scale).

I will also run with dynamical scale, but when doing that, I wanted to look at the distribution that you mentioned and I realize that such distribution is exactly 0 at mg5amc level. Since your computation is at tree-level, the two lepton are exactly back to back and the associated pt of the pair is exactly 0. (and this is fully independent of the version of MG5aMC).

So I'm clearly confused about something. Either you are not looking at that process or the plot that you do are done after parton-shower (and is fully dominated by the parton-shower). In that case you should check the change in setup between the parton-shower program to explain that difference in shape.

Cheers,

Olivier

Dear Olivier,

I am SO SORRY and would like to sincerely apologize!

Of course, for the elastic-elastic samples, there pt-ll distribution should be entirely determined by Pythia8 (and I should have thought of this!!!).

I got confused, because there were other effects that clearly stemmed from MG5 (admittedly for single-dissociative samples, ie. lpp1 1 and lpp2 2) and we had not changed the Pythia *settings* (though of course might have changed the Pythia version, but this is a bit more subtle for us).

Thanks a lot for your help and again, I am sorry, and hope that you did not spend too much time on this!

Best wishes and relaxing holidays!

No problem,

Sorry that it took me that long to spot that.

Just for completeness, I do reproduce the dynamical scale number of 2.7.0 with 2.9.7 (only check the cross-section)

Cheers,

Oliver

> On 13 Dec 2021, at 16:30, Kristin Lohwasser <email address hidden> wrote:
>
> Question #698432 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/698432
>
> Status: Answered => Solved
>
> Kristin Lohwasser confirmed that the question is solved:
> Dear Olivier,
>
> I am SO SORRY and would like to sincerely apologize!
>
> Of course, for the elastic-elastic samples, there pt-ll distribution
> should be entirely determined by Pythia8 (and I should have thought of
> this!!!).
>
> I got confused, because there were other effects that clearly stemmed
> from MG5 (admittedly for single-dissociative samples, ie. lpp1 1 and
> lpp2 2) and we had not changed the Pythia *settings* (though of course
> might have changed the Pythia version, but this is a bit more subtle for
> us).
>
> Thanks a lot for your help and again, I am sorry, and hope that you did
> not spend too much time on this!
>
> Best wishes and relaxing holidays!
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.