MadGraph5_aMC@NLO

Matching and ME cut on leading pt jet

Asked by Marc Thomas

Is it possible to have a large cut on the leading jet pt at the ME level, and still have matching in a consistent way (with a meaningful cross-section)?

I'm performing post-Delphes cuts of pt_j1 > 550 GeV, and as expected, only a very small number of generated
events survive cuts. Therefore, I'm trying to generate events with a cut already at the ME level (e.g. xptj > 100 GeV).

My problem is that I'm not sure if the post-matching cross section is accurate when the cut has been applied?
(which I plan to use in CX_after_post_Delphes_cuts = CX_post_matching * (Number_pass_post_Delphes_cuts / Number_generated_post_matching)

If I generate,
import model mssm-full
define sq = ul ul~
generate p p > sq sq QED=0 QCD=99 @0
add process p p > sq sq j QED=0 QCD=99 @1
add process p p > sq sq j j QED=0 QCD=99 @2
with ickkw = 1, xptj > 100, xqcut = 30
then I miss the area of phase space with 30 < xqcut < 100, but xqcut < 30 is still filled in by Pythia.

If I use xptj > 100, xqcut > 100
then it's as if there is no cut, as Pythia fills in the phase space with xqcut < 100.

However, if I instead miss out the zero jet diagram, and just do
generate p p > sq sq j QED=0 QCD=99 @1
add process p p > sq sq j j QED=0 QCD=99 @2
then the cross-section is highly dependent on xqcut.

Any advice would be appreciated. Thanks.

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
Last query:
Last reply:
Revision history for this message
Best Olivier Mattelaer (olivier-mattelaer) said : 		#1

I would personally generate such sample in the following way

1) do not define any cut, and check which value of the xqcut and qcut are valid for the matching
(with 0j-2j sample)

If as I expect qcut is lower than 100.
then I generate only the 1 jet and 2 jet sample with the same value for xqcut and qcut, but with
pth1min set on 100 (at generation time)

This should be the perfect way to generate such sample.
In both generation, I would do the typical differential jet rate plot and the variation of the qcut parameter to check that everything is under control.

Finally, you need to check that you don’t have additional resonances in the 1j/2j samples (like gluino decaying in squark-quark)
if such resonances is possible you need to subtract it via the $ syntax:
i.e. add process p p > sq sq j j $ go

Cheers,

Olivier

On Jul 7, 2014, at 8:07 PM, Marc Thomas <email address hidden> wrote:

> New question #251324 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/251324
>
> Is it possible to have a large cut on the leading jet pt at the ME level, and still have matching in a consistent way (with a meaningful cross-section)?
>
> I'm performing post-Delphes cuts of pt_j1 > 550 GeV, and as expected, only a very small number of generated
> events survive cuts. Therefore, I'm trying to generate events with a cut already at the ME level (e.g. xptj > 100 GeV).
>
> My problem is that I'm not sure if the post-matching cross section is accurate when the cut has been applied?
> (which I plan to use in CX_after_post_Delphes_cuts = CX_post_matching * (Number_pass_post_Delphes_cuts / Number_generated_post_matching)
>
> If I generate,
> import model mssm-full
> define sq = ul ul~
> generate p p > sq sq QED=0 QCD=99 @0
> add process p p > sq sq j QED=0 QCD=99 @1
> add process p p > sq sq j j QED=0 QCD=99 @2
> with ickkw = 1, xptj > 100, xqcut = 30
> then I miss the area of phase space with 30 < xqcut < 100, but xqcut < 30 is still filled in by Pythia.
>
> If I use xptj > 100, xqcut > 100
> then it's as if there is no cut, as Pythia fills in the phase space with xqcut < 100.
>
> However, if I instead miss out the zero jet diagram, and just do
> generate p p > sq sq j QED=0 QCD=99 @1
> add process p p > sq sq j j QED=0 QCD=99 @2
> then the cross-section is highly dependent on xqcut.
>
> Any advice would be appreciated. Thanks.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Marc Thomas (mct1g11) said : 		#2

Thanks Olivier,

I've done this and it works well and consistently thanks.
In fact this is what I was already doing, but I have now found an error in my analysis code which incorrectly made me think that it was not the correct way!

One follow on question regarding the possibility of additional resonances,
"if such resonances is possible you need to subtract it via the $ syntax:
i.e. add process p p > sq sq j j $ go"

Just to check that I understand this correctly;
In my case I'm ONLY producing
generate p p > sq sq j QED=0 QCD=99 @1
add process p p > sq sq j j QED=0 QCD=99 @2
where sq = ul ul~
which means that while this does involve some diagrams containing go > sq q,
I don't need to subtract these as I'm not also including
e.g. p p > sq go at the ME level (in which case I would have double counting and need to use $ go).

Cheers,
Marc

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#3

Hi Marc,

> I've done this and it works well and consistently thanks.
> In fact this is what I was already doing, but I have now found an error in my analysis code which incorrectly made me think that it was not the correct way!

great that you find the problem.

> Just to check that I understand this correctly;
> In my case I'm ONLY producing
> generate p p > sq sq j QED=0 QCD=99 @1
> add process p p > sq sq j j QED=0 QCD=99 @2
> where sq = ul ul~
> which means that while this does involve some diagrams containing go > sq q,
> I don't need to subtract these as I'm not also including
> e.g. p p > sq go at the ME level (in which case I would have double counting and need to use $ go).

In itself you are right. I’m just worry about the dependencies of your generation in the gluino mass and the to fact that your
sample/analysis can probably not be combine with the standard gluino searches.

Cheers,

Olivier

On Jul 10, 2014, at 2:26 PM, Marc Thomas <email address hidden> wrote:

> Question #251324 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/251324
>
> Marc Thomas posted a new comment:
> Thanks Olivier,
>
> I've done this and it works well and consistently thanks.
> In fact this is what I was already doing, but I have now found an error in my analysis code which incorrectly made me think that it was not the correct way!
>
> One follow on question regarding the possibility of additional resonances,
> "if such resonances is possible you need to subtract it via the $ syntax:
> i.e. add process p p > sq sq j j $ go"
>
> Just to check that I understand this correctly;
> In my case I'm ONLY producing
> generate p p > sq sq j QED=0 QCD=99 @1
> add process p p > sq sq j j QED=0 QCD=99 @2
> where sq = ul ul~
> which means that while this does involve some diagrams containing go > sq q,
> I don't need to subtract these as I'm not also including
> e.g. p p > sq go at the ME level (in which case I would have double counting and need to use $ go).
>
> Cheers,
> Marc
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Marc Thomas (mct1g11) said : 		#4

Hi Olivier,

In the current analysis, we aren't worrying about the standard gluino searches, and in fact we've fixed the mass at 10 TeV, but thanks for clarifying this for future reference.

Cheers,
Marc

Revision history for this message
Marc Thomas (mct1g11) said : 		#5

Thanks Olivier Mattelaer, that solved my question.