MadGraph5_aMC@NLO

filter born level diagrams of NLO process

Asked by Matthias Komm

Hi,

using the model 'sm_loop-ckm' I like to generate off shell single top production. Running the command:
generate p p > u~ b b~ w+ $$ w+ w- t t~ / a z h QCD<=1 QED<=4 [QCD]
yields the desired diagrams but it produces also diagrams with no top quarks. I tried to use the diagram filter function with
generate p p > u~ b b~ w+ $$ w+ w- t t~ / a z h QCD<=1 QED<=4 --diagram_filter [QCD]
yet this causes MG to fall back to LO production. How can these diagrams be removed?

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Matthias Komm
Solved:
Last query:
Last reply:
Revision history for this message
Matthias Komm (mkomm) said : 		#1

any news ? I am also interested in just applying some hack if this cannot be solved using the filter option at the moment ...
So far I've found the function user_filter in madgraph/loop/loop_diagram_generation.py where multiple examples are given to remove loops. But it is not clear where one could remove born level diagrams before the loops are generated.

BTW, I hope what I am trying to do makes even sense; let me know if generating on- and off-shell tops separately and then combining the files breaks some observables (e.g. due to missing interference/gauge invariance)

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#2

When running
> generate p p > u~ b b~ w+ $$ w+ w- t t~ / a z h QCD<=1 QED<=4 --diagram_filter [QCD]
 I get:
InvalidCmd : No particle --diagram_filter in model

As far as I know, the diagram filtering is not possible at NLO.
It is already something quite dangerous at LO, and likely even more crazy at NLO accuracy.

> BTW, I hope what I am trying to do makes even sense; let me know if
> generating on- and off-shell tops separately and then combining the
> files breaks some observables (e.g. due to missing interference/gauge
> invariance)

If you break gauge invariance, you are likely breaking NLO accuracy and the cancelation of the vritual by the real this is quite gauge sensitive. breaking gauge invariance (even at LO) is a bad idea since then none of the result of madgraph makes sense. All the comparison will be very diffcult to do since you have a lot of spurious dependency.
The two most striking one are
1) you can not use the default dynamical scale of MG5 (easy to handle)
2) you have to force madgraph to work in a specific frame since breaking gauge invariance will also break lorentz invariance and then you have to pick of frame to allow a fair comparison.
(This in itself prevent to use NLO computation since the computation are done in multiple frame)

Cheers,

Olivier

> On 26 Jul 2020, at 18:35, Matthias Komm <email address hidden> wrote:
>
> Question #691906 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/691906
>
> Matthias Komm posted a new comment:
> any news ? I am also interested in just applying some hack if this cannot be solved using the filter option at the moment ...
> So far I've found the function user_filter in madgraph/loop/loop_diagram_generation.py where multiple examples are given to remove loops. But it is not clear where one could remove born level diagrams before the loops are generated.
>
> BTW, I hope what I am trying to do makes even sense; let me know if
> generating on- and off-shell tops separately and then combining the
> files breaks some observables (e.g. due to missing interference/gauge
> invariance)
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Matthias Komm (mkomm) said : 		#3

Hi, sorry but you did not answer the questions. My main question is how to generate off-shell single top quark production (ideally in 4FS). My follow up question is IF generating on- and off-shell tops separately and then combining the files breaks something, i.e is different from generating both together.

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#4

> y main question is how
> to generate off-shell single top quark production (ideally in 4FS).

This is not supported.
You can hack the code to try. But I'm not sure how deep you need to go in top of changing all the file for all the type of amplitude (born, loop, real, color connected born, ...)
I'm also worried that you would need to change the phase-space integration security,...
This is uncharted territory.

> My
> follow up question is IF generating on- and off-shell tops separately
> and then combining the files breaks something, i.e is different from
> generating both together.

Tthat I answer, yes you will break gauge invariance, lorentz invariance and NLO accuracies.

Cheers,

Olivier

> On 31 Jul 2020, at 02:50, Matthias Komm <email address hidden> wrote:
>
> Question #691906 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/691906
>
> Status: Answered => Open
>
> Matthias Komm is still having a problem:
> Hi, sorry but you did not answer the questions. My main question is how
> to generate off-shell single top quark production (ideally in 4FS). My
> follow up question is IF generating on- and off-shell tops separately
> and then combining the files breaks something, i.e is different from
> generating both together.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Matthias Komm (mkomm) said : 		#5

ah too bad. I thought that MG first generates the born diagrams and one can remove the unwanted ones. Then the loop and real would just be created from the selected born ones in a 2nd step.

I may then just do this in 5FS then since on and off-shell can just be separate by enforcing top and b quark charge combinations. I am surprised that 4FS causes problems.