Higgs QCD interference in tt final state

Asked by jmli

Dear MG5 team,

I am trying to generate the tt~ production at NLO in SM:

import model loop_sm
generate p p > t t~ QED=2 [QCD]
output sm-ttnlo
launch

The following error occurs in run_01_tag_1_debug.log:
launch auto
Traceback (most recent call last):
  File "/Users/jinmianli/Mywork/MG5_aMC_v2_4_3/madgraph/interface/extended_cmd.py", line 1011, in onecmd
    return self.onecmd_orig(line, **opt)
  File "/Users/jinmianli/Mywork/MG5_aMC_v2_4_3/madgraph/interface/extended_cmd.py", line 966, in onecmd_orig
    return func(arg, **opt)
  File "/Users/jinmianli/Mywork/MG5_aMC_v2_4_3/madgraph/interface/amcatnlo_run_interface.py", line 1208, in do_launch
    self.compile(mode, options)
  File "/Users/jinmianli/Mywork/MG5_aMC_v2_4_3/madgraph/interface/amcatnlo_run_interface.py", line 4317, in compile
    self.check_tests(test, this_dir)
  File "/Users/jinmianli/Mywork/MG5_aMC_v2_4_3/madgraph/interface/amcatnlo_run_interface.py", line 4328, in check_tests
    return self.parse_test_mx_log(pjoin(dir, '%s.log' % test))
  File "/Users/jinmianli/Mywork/MG5_aMC_v2_4_3/madgraph/interface/amcatnlo_run_interface.py", line 4340, in parse_test_mx_log
    ' set to 0 in the param_card.dat.')
aMCatNLOError: Some tests failed, run cannot continue.
Please check that widths of final state particles (e.g. top) have been set to 0 in the param_card.dat.

I am quite sure that I have set the width of top quark to zero. The problem still exist.

Thank you very much!
Jinmian

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
marco zaro Edit question
Solved by:
Valentin Hirschi
Solved:
Last query:
Last reply:
Revision history for this message
marco zaro (marco-zaro) said :
#1

Dear Jinmian,
what are you exactly interested in?
If I generate your process, I see these warnings which appear
>> WARNING: All Born diagrams do not factorize the same sum of power(s) of the the perturbed order(s) QCD.
>> This is potentially dangerous as the real-emission diagrams from aMC@NLO will not be consistent with these virtual contributions.
>> WARNING: Some loop diagrams contributing to this process are discarded because they are not pure (QCD)-perturbation.
>> Make sure you did not want to include them.
because with that syntax you include all born diagrams up to order alpha_ew^2 alpha_s^2 (orders which are not specified are unconstrained), that is both diagrams with gluons in the s-channel and with a gamma/Z.
In any case, the current code is capable to compute NLO QCD corrections only if born diagrams factor the maximum allowed power of the QCD coupling.
If you tell us more in details what you are interested in, we can see if /how we can help you...

Cheers,

Marco

Revision history for this message
jmli (phyljm) said :
#2

Dear Marco,

Thanks a lot for your reply.

I want to use the MG5 to study the interference between g g > h > t t~ and QCD tt~ production.
I think I can set the Higgs mass and width freely for this purpose?

Cheers,
Jinmian

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

Dear Jimli,

This is actually not a trivial computation.
g g > h > t t~ is a loop-induced process.
g g > t t~ QED=0 is a tree-level process.
We are not able to mix those two type of diagram and (therefore select only the interference term)

The approach I would suggest is to contract the loop to replace it by an effective vertex (such that the code believes that this is a tree-level process).
Using the heft model, is however not a good idea since the invariant mass for your higgs is at least twice the top quark (i.e. outside the validity region of the heft approximation).
I would therefore suggest to use a model like this one:
https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/Models/ggHFullLoop
I’m not an author of this model so I cannot provide help on it (please contact the author of the model in case of problem.)
but from the description, it should be able to run what you want.

If you want only the interference term, the syntax should be
generate p p > t t~ / h1 HIG^2==1

Cheers,

Olivier

PS: For interference computation, it is better to not use the default dynamical scale of MG.

> On Aug 29, 2016, at 13:22, jmli <email address hidden> wrote:
>
> Question #369922 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/369922
>
> Status: Needs information => Open
>
> jmli gave more information on the question:
> Dear Marco,
>
> Thanks a lot for your reply.
>
> I want to use the MG5 to study the interference between g g > h > t t~ and QCD tt~ production.
> I think I can set the Higgs mass and width freely for this purpose?
>
> Cheers,
> Jinmian
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Best Valentin Hirschi (valentin-hirschi) said :
#4

Dear Jimli,

Olivier's suggestion would work for this particular case, however not in
general since you cannot expect to always find a model that contains the
exact loop-induced form factor as interaction.

The more generic way to do such 'tree x loop-induced' computation is to
remain in the framework of MadEvent for loop-induced computations (since
after all your process is still finite, i.e. does not involve any IR
divergence). Also the loop-induced framework would offer a more efficient
integration.
Finally, you would not need the support for squared ordered amplitude
constraints in MadFKS (i.e. the NLO integrator) which is not available in
the current MG5aMC public version.

So a better approach consists in treating this computation as a regular
loop-induced process, but with additional fake loops which are in fact the
tree-level diagrams you want to interfere against.
This can be done reasonably simply by defining new R2 interactions in the
UFO model with a coupling order 'TreeLevelInterference' and then asking to
compute the loop-induced process 'g g > t t~ [noborn=QCD]
TreeLevelInterference^2==1' with a loop-filter selecting only the Higgs
loop.

Achieving the above requires some (minimal) technical hacking of MG5aMC
which I refrain from reporting here as we cannot guarantee that such
modification would be stable across future MG5aMC versions. However, this
approach would allow for generic 'loop-induced x tree' computations. If
there is enough demand for this, we might think of developing a native
stable implementation of such trick. In the meantime, if you want to go
down this path, contact me directly (valentin.hirschi[at]gmail.com) and
I'll give you detailed instructions.

Cheers,

On Mon, Aug 29, 2016 at 5:57 AM, Olivier Mattelaer <
<email address hidden>> wrote:

> Question #369922 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/369922
>
> Status: Open => Answered
>
> Olivier Mattelaer proposed the following answer:
> Dear Jimli,
>
> This is actually not a trivial computation.
> g g > h > t t~ is a loop-induced process.
> g g > t t~ QED=0 is a tree-level process.
> We are not able to mix those two type of diagram and (therefore select
> only the interference term)
>
> The approach I would suggest is to contract the loop to replace it by an
> effective vertex (such that the code believes that this is a tree-level
> process).
> Using the heft model, is however not a good idea since the invariant mass
> for your higgs is at least twice the top quark (i.e. outside the validity
> region of the heft approximation).
> I would therefore suggest to use a model like this one:
> https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/Models/ggHFullLoop
> I’m not an author of this model so I cannot provide help on it (please
> contact the author of the model in case of problem.)
> but from the description, it should be able to run what you want.
>
> If you want only the interference term, the syntax should be
> generate p p > t t~ / h1 HIG^2==1
>
> Cheers,
>
> Olivier
>
> PS: For interference computation, it is better to not use the default
> dynamical scale of MG.
>
>
> > On Aug 29, 2016, at 13:22, jmli <email address hidden>
> wrote:
> >
> > Question #369922 on MadGraph5_aMC@NLO changed:
> > https://answers.launchpad.net/mg5amcnlo/+question/369922
> >
> > Status: Needs information => Open
> >
> > jmli gave more information on the question:
> > Dear Marco,
> >
> > Thanks a lot for your reply.
> >
> > I want to use the MG5 to study the interference between g g > h > t t~
> and QCD tt~ production.
> > I think I can set the Higgs mass and width freely for this purpose?
> >
> > Cheers,
> > Jinmian
> >
> > --
> > You received this question notification because you are an answer
> > contact for MadGraph5_aMC@NLO.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.
>

--
Valentin

Revision history for this message
Valentin Hirschi (valentin-hirschi) said :
#5

I had misread Olivier's proposal, and since he suggests to it within the MadEvent (LO) integration framework, it will work with the current public version of MG5aMC, since squared order amplitude constraints are supported there.
Sorry for the confusion.

However, it remains true that Oliver's suggestion only applies when you have a model that has all relevant loop-induced loop factors coded as interactions, which is of course basically only the case for g g > h.
Therefore, my proposal is likely the only way to proceed to compute for any 'loop-induced x tree' interference in any more complicated process.

Revision history for this message
Valentin Hirschi (valentin-hirschi) said :
#6

For reference, I have listed the instructions to simulate the interference of loop-induced and tree-level diagrams on this wiki page:

https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/LoopInducedTimesTree

I stress that we do not have official support for such computation yet, so employing the procedure detailed above requires care and cross-checks.

Revision history for this message
jmli (phyljm) said :
#7

Thanks Valentin Hirschi, that solved my question.