Small decay width treatment

Asked by Anastasia on 2019-05-02

Dear community,

I am trying to generate events for a process which includes particles with very small decay widths (I am also setting these widths to AUTO in order to let MadGraph calculate them), but I'm getting the following error:

ERROR: The width of particle 9000006 is too small for an s-channel resonance (3.67072e-18). If you have this particle in an s-channel, this is likely to create numerical instabilities.

The problem is that I indeed produce a pair of such particles on-shell and then let them decay, so it seems that the error directly applies to my case.

I've tried to improve the precision adding the following line to my run card:
1e-25 = small_width_treatment

but it didn't change the output. Could you please help me with this issue? I'm using version 2.6.5.

Cheers,
Anastasia

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
2019-05-02
Last reply:
2019-05-02

Hi,

What happens if you keep the "small_width_treatment" parameter to the default value?
You should then have a warning stating the following:

Particle 9000006 will use a fake width ( XXX instead of YYY ).
Cross-section will be rescaled according to NWA if needed.
To force exact treatment reduce the value of 'small_width_treatment' parameter of the run_card

This should allows to have a nice handling of the resonances by avoiding any numerical issues.
As stated the code will correct the cross-section according to the NWA.

In the unlikely case where you have other non-resonant diagram that contribute significantly to the cross-section, the scaling does not occur on such contribution of the cross-section.
Now this assumes that interference term between the onshell and off-shell diagram are negligeable (since otherwise NWA does not make sense).

So expect for some of the problematic case. This mode is equivalent to
1) make the computation with a "large" width (large means here 1e-8 times the mass)
2) correct the cross-section by the ratio of the fake width by the real width
Obviously the shape of the Breit-wigner will be much too large (in your case by 10 order of magnitude) but this should not be a problem for any experience.

This was actually the recomendation before (asking the user to do that manually), the small_width_treatment is simply an automation of this procedure.

Cheers,

Olivier

Anastasia (assimptota) said : #2

Dear Oliver,

Thanks for your reply. Without adding small_width_treatment I get both the warning and the error:

WARNING: Particle 9000006 will use a fake width ( 0.000324 instead of 3.67072e-18 ).
Cross-section will be rescaled according to NWA if needed.To force exact treatment reduce the value of 'small_width_treatment' parameter of the run_card
INFO: Update the dependent parameter of the param_card.dat
ERROR: The width of particle 9000006 is too small for an s-channel resonance (3.67072e-18). If you have this particle in an s-channel, this is likely to create numerical instabilities .

I indeed think that at least in my particular case there is no other non-resonant diagram that interferes with the present one.

Do I understand correctly that you suggest to remove small_width_treatment parameter and that in this case I can trust the computations of the cross section and its error bars?

Hi,

> Thanks for your reply. Without adding small_width_treatment I get both
> the warning and the error:
>
> WARNING: Particle 9000006 will use a fake width ( 0.000324 instead of 3.67072e-18 ).
> Cross-section will be rescaled according to NWA if needed.To force exact treatment reduce the value of 'small_width_treatment' parameter of the run_card
> INFO: Update the dependent parameter of the param_card.dat
> ERROR: The width of particle 9000006 is too small for an s-channel resonance (3.67072e-18). If you have this particle in an s-channel, this is likely to create numerical instabilities .

Ok this should be fine, can you send me your model/ process/...
such that I can check why you have both message and see if I can remove the printing of the error in that case (that should be the case). You can send that to <email address hidden>

> Do I understand correctly that you suggest to remove
> small_width_treatment parameter and that in this case I can trust the
> computations of the cross section and its error bars?

Cross-section and error-bars should be fine.
The invariant mass distribution will be too large (with a width of 0.000324 instead of 3.67072e-18)
and therefore all observables correlated with that mass might be technically impacted but the impact should be very small.

Cheers,

Olivier

> On 2 May 2019, at 14:47, Anastasia <email address hidden> wrote:
>
> Question #680594 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/680594
>
> Status: Answered => Open
>
> Anastasia is still having a problem:
> Dear Oliver,
>
> Thanks for your reply. Without adding small_width_treatment I get both
> the warning and the error:
>
> WARNING: Particle 9000006 will use a fake width ( 0.000324 instead of 3.67072e-18 ).
> Cross-section will be rescaled according to NWA if needed.To force exact treatment reduce the value of 'small_width_treatment' parameter of the run_card
> INFO: Update the dependent parameter of the param_card.dat
> ERROR: The width of particle 9000006 is too small for an s-channel resonance (3.67072e-18). If you have this particle in an s-channel, this is likely to create numerical instabilities .
>
> I indeed think that at least in my particular case there is no other
> non-resonant diagram that interferes with the present one.
>
> Do I understand correctly that you suggest to remove
> small_width_treatment parameter and that in this case I can trust the
> computations of the cross section and its error bars?
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Can you help with this problem?

Provide an answer of your own, or ask Anastasia for more information if necessary.

To post a message you must log in.