Zero pT for W boson production process

Asked by Maksym Ovchynnikov on 2018-09-19

I simulate a process p p > w+, w+ > e+ ve at sqrt[s] = 13 TeV. The output gives the cross-section 9.5 nb which is close to total cross-section 11.8 pb obtained in CMS paper https://arxiv.org/abs/1603.09222 by extrapolating the experimental data. However, mean pT for W bosons is zero, while ATLAS paper https://arxiv.org/pdf/1108.6308.pdf shows non-trivial pT distribution of W, with most of W having large pT.

Could you please tell me what is a reason of such discrepancy?

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
2018-09-26
Last query:
2018-09-26
Last reply:
2018-09-26

Since you ask a 2>1 process and that the computation is done with PDF without any transverse component
You will have all W generated without any PT by construction.

Now your events are INCLUSIVE over all radiation. If you use a parton-shower (like you should to compare to data), it will make specify what are those radiation and this will make the PT of the W explicit.
Now the precision of such PT distribution is quite low with such method.

If you are interested in such distribution, you should use a match/merged method at LO or even better at NLO accuracy.

Cheers,

Olivier

> On 19 Sep 2018, at 22:43, Maksym Ovchynnikov <email address hidden> wrote:
>
> Question #674155 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/674155
>
> Description changed to:
> I simulate a process p p > w+, w+ > e+ ve at sqrt[s] = 13 TeV. The
> output gives the cross-section 9.5 nb which is close to total cross-
> section 11.8 pb obtained in CMS paper https://arxiv.org/abs/1603.09222
> by extrapolating the experimental data. However, mean pT for W bosons is
> zero, while ATLAS paper https://arxiv.org/pdf/1108.6308.pdf shows non-
> trivial pT distribution of W, with most of W having large pT.
>
> Could you please tell me what is a reason of such discrepancy?
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Maksym Ovchynnikov (name-xxx) said : #2

Dear Olivier, you say that although total cross-section is comparable to data, the spectrum is completely different since the actual spectrum is evaluated for 2 -> n process due to parton showers production, while MadGraph doesn't take into account this?

Maksym Ovchynnikov (name-xxx) said : #3

Also, should I believe the FONLL predictions for cross-sections of the process p p > w+, w+ > ve e+ for specific cuts on e+ or ve pseudo-rapidity?

Hi,

The parton-shower is a unitary process. It does not change the cross-section. But indeed it change the shape.
The change in the shape are in general quite mild but for observables where the shape is "trivial" like the pt of the W in your case.

Now for such process, you will typically use a NNLO computation to get the total cross-section correct
since at lower order the cross-section has quite large theoretical uncertainty (for your process, at LO, this error is around 12% from scale and 3% from PDF as reported by MG5aMC)

Cheers,

Olivier

Maksym Ovchynnikov (name-xxx) said : #5

Dear Olivirer, thank you. But I still have doubts: can I use MadGraph for obtaining the cross-section p p > w+, w+ > ve e+ for specific cuts on e+ or ve pseudo-rapidity? Naively I think that MadGraph will give completely incorrect result since it predicts wrong pT and probably eta distribution of W bosons, and therefore wrong eta distribution of neutrinos and positrons. However, I'm not quite sure in this.

An if MadGraph can't be used, does there exist an estimation of the error of its prediction in comparison to realistic one?

Sorry if you have already answered this question.

Hi,

MG5aMC can perform many type of computation with many different accuracy.
If you want to stick with LO accuracy,
then it might be better to generate your W boson with one (hard) jet
generate p p > w+ j , w+ > ve e+

I would advise to use some type of matching/merging (like MLM for example)
via
generate p p > w+ , w+ > ve e+
add process p p > w+ j , w+ > ve e+
add process p p > w+ j j , w+ > ve e+

With the second syntax, the MG5aMC cross-section will be wrong due to duble counting
which will be removed by the parton-shower by vetoing some events. (please read about matching/merging before doing this since the removal of the double counting is not always working nicely (it depends of the matching scale that you choose).

So we are certainly able to be used for such prediction (also at NLO accuracy actually) but in all cases if would advise to go trough a PS after MG5aMC in order to have correctly the soft radiation (this is required for matching/merging and for NLO run anyway)

Cheers,

Olivier

> On 20 Sep 2018, at 08:32, Maksym Ovchynnikov <email address hidden> wrote:
>
> Question #674155 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/674155
>
> Status: Answered => Open
>
> Maksym Ovchynnikov is still having a problem:
> Dear Olivirer, thank you. But I still have doubts: can I use MadGraph
> for obtaining the cross-section p p > w+, w+ > ve e+ for specific cuts
> on e+ or ve pseudo-rapidity? Naively I think that MadGraph will give
> completely incorrect result since it predicts wrong pT and probably eta
> distribution of W bosons, and therefore wrong eta distribution of
> neutrinos and positrons. However, I'm not quite sure in this.
>
> An if MadGraph can't be used, does there exist an estimation of the
> error of its prediction in comparison to realistic one?
>
> Sorry if you have already answered this question.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Maksym Ovchynnikov (name-xxx) said : #7

Dear Olivier,

thank you for this detailed answer! I have only one question regarding the topic: should I with more or less error normalize the total cross-section for particular pseudo-rapidity cuts obtained in MadGraph (as you wrote, incorrect due to total counting) on the experimental total cross-section, or this is clearly incorrect?

Hi,

Well this is a experimental driven approach which has it's own merit (and drawback).
This highly depends of what you want to do/search/...
When doing that you certainly hide any type of new physics that scale the total cross-section for example.

Cheers,

Olivier

Maksym Ovchynnikov (name-xxx) said : #9

Thanks Olivier Mattelaer, that solved my question.