3photons_0j@NLO

Hi,

1) g g > a a a is actually zero due the Furry theorem (all loop with odd number of photon are vanishing)
We use this theorem at the time of diagram generation to avoid to have to compute those vanishing loop.
(So indeed g g > a and g g > 5a are also zero)

2) g g > a a a (if is was non zero) would have been a contribution to the NNLO and not a contribution to the NLO.
The loop contribution at NLO are the interference term between the loop and the born matrix-element.
Since in the case of g g > a a a, you do not have a born, you do not have any interference and this does not contribute at NLO.

The Loop-square is a NNLO contribution. However in case of loop-induced processes they can be much more important than typical NNLO contribution.

3) For your problem concerning the comparaison of LO and NLO, you should be careful that if you do
Generate p p > a a
And
Generate p p > a a [QCD]
The two computation will not have the same dynamical scale/ the same cuts/the same PDF/... So you need to be careful when comparing those numbers.
This is not what we did in 1405.0301v2
We ONLY use the "Generate p p > a a [QCD]" syntax, and at "launch" level, you have a switch to run at LO accuracy rather than at NLO accuracy.
This way you are sure that you use the same cut/scale/...

Cheers,

Olivier

So when you do
generate g g > a a a [QCD]
You are actually computing a contribution which is NNLO compare to Q \bar q > a a a
That contribution is NOT include in the NLO computation reported in arXiv:1405.0301v2 [hep-ph])
Loop-induced contribution can only be computed since arXiv:1507.00020<http://arXiv.org/abs/arXiv:1507.00020>

On 22 Sep 2017, at 16:13, Adam Jinaru <<email address hidden><mailto:<email address hidden>>> wrote:

New question #658531 on MadGraph5_aMC@NLO:
https://answers.launchpad.net/mg5amcnlo/+question/658531

Dear Madgraph users and experts,

I tried to generate p p> 3 photons@NLO with Madgraph. With my phase space cuts I get a cross-section for 3a_0j@NLO which is smaller than the 3a_0j@LO case - which is in disagreement with the previous results obtained in other publications, such as the 2014 Madgraph paper (arXiv:1405.0301v2 [hep-ph]). The reason is well known: the process at NLO gets important contributions from loops.
Then the process g g > a a a [QCD] must contribute a lot to the 3photons production (g q @NLO only with one extra jet).

So when I generate g g > a a a [QCD], I get this message:
Command "generate g g > a a a [QCD]" interrupted with error:
NoDiagramException : No amplitudes generated from process Process: g g > a a a WEIGHTED=12 [ noborn = QCD ]. Please enter a valid process

What should I do to get non-zero amplitudes (some MHV amplitudes surely must exist...)?
For the processes g g > a [QCD] and g g > a a a a a [QCD] I also get the same messages.
Is this a Madgraph problem, something that must be carefully considered and implemented?

Thanks a lot,
Adam

--
You received this question notification because you are an answer
contact for MadGraph5_aMC@NLO.

Hi Olivier,

ok, thanks a lot, thing are nice and clear now!
I didn't know about the Furry theorem... I suspected it was something related to mhv amplitudes.

Many thanks!
Adam

Hi,

sorry.. one more question: how does one do this straight from the command line? I think we'll need to proceed like this.
Thanks,
Adam

Hi Adam,

I'm not sure about what your question is.

If you want to compute the K-factor between LO and NLO cross-section, you can use the following script
generate p p > a a a [QCD]
output MYRUN
launch MYRUN
order=LO
fixed_order=ON
#here you can set all the parameter that you want like
# set ptgmin 40
launch MYRUN
order=NLO

You will get the first cross-section at LO, and the second at NLO, the ratio of the two will give you the K-factor

Cheers,

Olivier

Hi Olivier,

Let's say I want just the first part of that, from the command line, but I need to update the run card to adjust the random number seed, or number of events, or whatever. Then I should presumably run mg5_aMC on a text file with:

generate p p > a a a [QCD]
output MYRUN

in a first pass, then edit the run card, then in a second pass run mg5_aMC on a text file with just:

launch MYRUN
order=LO
fixed_order=ON

Is that right?

Thanks,
Zach