number of event generated hit CRITICAL error

[Marzieh Bahmani]

Hi

I have problem generating events:

import model SM_HeavyN_Meson_NLO
define ee = e+ e-
define mm = mu+ mu-
define ll = ta+ ta-
define vv = ve ve~
define pix = pi+ pi-

## ###########################
## p p > W > N e, N > mu pi ##
## ###########################
generate p p > n1 mm [QCD]
output out_meson -f

launch out_meson
order=nlo
shower=py8
madspin=onshell
set VeN1 1
set VmuN1 1
set VtaN1 0

set mn1 3
set wn1 1e-15
set nevents 20k #-40k

decay n1 > pix mm
done

 I get this error: CRITICAL:
The number of event generated is only 30763/56881. This typically
indicates that you need specify cut on the decay process.
CRITICAL: We strongly suggest that you cancel/discard this run.

and it doesn't matter if I change the number of events to 80k, i still get this error.

Thanks

[Olivier Mattelaer (olivier-mattelaer)]

What happens if you do

> import model SM_HeavyN_Meson_NLO
> define ee = e+ e-
> define mm = mu+ mu-
> define ll = ta+ ta-
> define vv = ve ve~
> define pix = pi+ pi-
>
>
> ## ###########################
> ## p p > W > N e, N > mu pi ##
> ## ###########################
> generate n1 > pix mm
> output
>
> launch
> set VeN1 1
> set VmuN1 1
> set VtaN1 0
>
> set mn1 3
> set wn1 1e-15
> set nevents 60k
>

The problem is not the production events generation, but the generation of the decay.
Is it something "unusual" about such decay?

Cheers,

Olivier

> On 1 Jun 2023, at 14:46, Marzieh Bahmani <email address hidden> wrote:
>
> Public bug reported:
>
> Hi
>
> I have problem generating events:
>
> import model SM_HeavyN_Meson_NLO
> define ee = e+ e-
> define mm = mu+ mu-
> define ll = ta+ ta-
> define vv = ve ve~
> define pix = pi+ pi-
>
>
> ## ###########################
> ## p p > W > N e, N > mu pi ##
> ## ###########################
> generate p p > n1 mm [QCD]
> output out_meson -f
>
> launch out_meson
> order=nlo
> shower=py8
> madspin=onshell
> set VeN1 1
> set VmuN1 1
> set VtaN1 0
>
> set mn1 3
> set wn1 1e-15
> set nevents 20k #-40k
>
> decay n1 > pix mm
> done
>
>
> I get this error: CRITICAL:
> The number of event generated is only 30763/56881. This typically
> indicates that you need specify cut on the decay process.
> CRITICAL: We strongly suggest that you cancel/discard this run.
>
> and it doesn't matter if I change the number of events to 80k, i still
> get this error.
>
> Thanks
>
> ** Affects: mg5amcnlo
> Importance: Undecided
> Status: New
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/2022069
>
> Title:
> number of event generated hit CRITICAL error
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi
>
> I have problem generating events:
>
> import model SM_HeavyN_Meson_NLO
> define ee = e+ e-
> define mm = mu+ mu-
> define ll = ta+ ta-
> define vv = ve ve~
> define pix = pi+ pi-
>
>
> ## ###########################
> ## p p > W > N e, N > mu pi ##
> ## ###########################
> generate p p > n1 mm [QCD]
> output out_meson -f
>
> launch out_meson
> order=nlo
> shower=py8
> madspin=onshell
> set VeN1 1
> set VmuN1 1
> set VtaN1 0
>
> set mn1 3
> set wn1 1e-15
> set nevents 20k #-40k
>
> decay n1 > pix mm
> done
>
>
> I get this error: CRITICAL:
> The number of event generated is only 30763/56881. This typically
> indicates that you need specify cut on the decay process.
> CRITICAL: We strongly suggest that you cancel/discard this run.
>
> and it doesn't matter if I change the number of events to 80k, i still
> get this error.
>
> Thanks
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/2022069/+subscriptions
>

[Marzieh Bahmani (marziehbahmani)]

Hi Olivier

the only unusual thing I can think of is very small width of the n1 particle in the order of 10^-15.
because we need to force n1 to be long lived.

Cheers,

Marzieh

[Olivier Mattelaer (olivier-mattelaer)]

Could you send me your UFO model at
<email address hidden>
(please put a link to this thread)

Cheers,

Olivier

[Olivier Mattelaer (olivier-mattelaer)]

So the real problem here is that the efficiency of the method is super low.
I only have the one estimated after 10 events so far and it is critically low
INFO: decaying event number 10. Efficiency: 102.44444444444444 [2.391008138656616 s]
(so this means that we do generate ~100 unweighted events (of the decay type) to get one valid for each of the production.

So if you want to do 80k production event, we need to generate 80k*100 = 80M decay events.
We do split such generation by batch of ~60k which means that we do have many batches and in rare occasion some batches fail to reach target. I would not worry too much about that as long as it is exceptional.

On the other hand, I would worry on why the efficiency is so small. This likely means that the method is breaking (you do expect to have the efficiency between 2 and 8 in general for this method) and the most typical reason is that NWA is not valid.

I guess that the issue here should be related to the presence of identical particle.
Indeed the onshell mode rely on the re-weighting module which does not support correctly identical particle between decay and production (impossible to identify correctly which one is which). This has been improved
in version 3.4.0 where the reweighting mode allow to average over all combination. And this averaging mode should be the default for onshell mode in madspin in 3.4.0. I have tried 3.5.2 and efficiency is also bad so this is likely not solving the issue in this case (maybe because the width are so small that we do have numerical issue when evaluating the matrix-element even for the correct combination).

In any case, which version of MG5aMC did you use?

Cheers,

Olivier

[Olivier Mattelaer (olivier-mattelaer)]

So here the solution is really to use madspin=full
and not madspin=onshell

Cheers,

Olivier

[Marzieh Bahmani (marziehbahmani)]

Hi Olivier

many thanks for your reply,

I was using madgraph version 3.3.1, and 3.5.0
however to be exact I had this problem only with version 3.3.1, I jus tried the version 3.4.2 now i dont have this error, just different issue with Pythia test, apparently in around 3% of the events energy momentum is not conserved.
I changed the threshold for E{beams} - total final state particles energy from 1GeV to 2Gev and finally I got one sample without any other problem. I am not sure if I am allowed to change this threshold, but this is another issue I have to discuss with Pythia and physics group.
I still have problem with the portion of lepton number violating and lepton number conserving, they are 35%/65% (should be 50/50).

I did not tried it for the above sample. I only tired to generate 1000 events now which was successful and the fraction LNC/LNV now is 53/47 which is getting better. I will generate more events and let you know. many thanks again for your help.

Cheers,

Marzieh

[Olivier Mattelaer (olivier-mattelaer)]

Clearly this "onshell" method does not work.
And I would really discard all the events generated for your processes.
Use the "full" (default) method, it is going to generate events much more quickly.
And you should likely have less issue (both in term in fraction and in term of accuracy)

Cheers,

Olivier

> On 28 Aug 2023, at 11:31, Marzieh Bahmani <email address hidden> wrote:
>
> Hi Olivier
>
> many thanks for your reply,
>
> I was using madgraph version 3.3.1, and 3.5.0
> however to be exact I had this problem only with version 3.3.1, I jus tried the version 3.4.2 now i dont have this error, just different issue with Pythia test, apparently in around 3% of the events energy momentum is not conserved.
> I changed the threshold for E{beams} - total final state particles energy from 1GeV to 2Gev and finally I got one sample without any other problem. I am not sure if I am allowed to change this threshold, but this is another issue I have to discuss with Pythia and physics group.
> I still have problem with the portion of lepton number violating and lepton number conserving, they are 35%/65% (should be 50/50).
>
> I did not tried it for the above sample. I only tired to generate 1000
> events now which was successful and the fraction LNC/LNV now is 53/47
> which is getting better. I will generate more events and let you know.
> many thanks again for your help.
>
> Cheers,
>
> Marzieh
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/2022069
>
> Title:
> number of event generated hit CRITICAL error
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi
>
> I have problem generating events:
>
> import model SM_HeavyN_Meson_NLO
> define ee = e+ e-
> define mm = mu+ mu-
> define ll = ta+ ta-
> define vv = ve ve~
> define pix = pi+ pi-
>
>
> ## ###########################
> ## p p > W > N e, N > mu pi ##
> ## ###########################
> generate p p > n1 mm [QCD]
> output out_meson -f
>
> launch out_meson
> order=nlo
> shower=py8
> madspin=onshell
> set VeN1 1
> set VmuN1 1
> set VtaN1 0
>
> set mn1 3
> set wn1 1e-15
> set nevents 20k #-40k
>
> decay n1 > pix mm
> done
>
>
> I get this error: CRITICAL:
> The number of event generated is only 30763/56881. This typically
> indicates that you need specify cut on the decay process.
> CRITICAL: We strongly suggest that you cancel/discard this run.
>
> and it doesn't matter if I change the number of events to 80k, i still
> get this error.
>
> Thanks
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/2022069/+subscriptions
>

[Marzieh Bahmani (marziehbahmani)]

Hi

with spinmode=full 20k events, the sample LNC/LNC is 50/50,
however the spin correlation Mandelstam variable s12 (one is the prompt lepton (from W) and 2 is the lepton from n1, for LNC and LNV events shows similar trend. that should not be the case.

[Marzieh Bahmani (marziehbahmani)]

Hi

I have noticed that beside the problem mentioned above, I also get wrong lifetime if I use spinmode=full
I am using version 3.4.2.

Cheers,

Marzieh

[Olivier Mattelaer (olivier-mattelaer)]

Sorry to come back only too late on this, doing some cleaning on old issue marked as unsolved.
I guess that your issue is here that you hardcode your width to a value which is not consistent with the LO width computed within the model, MadSpin does recompute the width by default if the BR are not defined in the param_card, if the recomputed width is not in agreement with the one provided a warning is raised.

Was that your issue?

Olivier