Auto width calculation throws key Error: 3

[Mohamed Aly]

Hi,

I am running a generation for a process p p > mu mu j j on MG5 V2.8.1 which will contain :

s-channel diagrams with W2 -> N mu
t-channel diagrams with N exchanged between 2 W2 particles or 1 W2 and 1 W particle

where W2 is a heavy right-handed gauge boson, N are heavy neutrinos and W are SM gauge bosons.

Running the above process with MN= 90 TeV and MW2 =10 TeV gives the following error during the automatic width calculation for W2 and N:

Command "generate_events run_01" interrupted in sub-command:
"set zerowidth_tchannel True" with error:
KeyError : 3

Is this a bug or is this a result of the extreme ratio of masses of W2 and N? The MG5_debug file is attached.

My generation commands are:

import model MLRSM
define bsmhiggs = h h2 h02 h03 h+ hp2 hl++ hr++ h3 a02 h- hm2 hl-- hr--
define j = g u c d s b t u c d s b t
generate p p > mu+ mu+ j j / bsmhiggs @1
add process p p > mu- mu- j j / bsmhiggs @2
output MW_10.0_MN_90.0_WRWR
launch
shower=pythia8
done
set MW2=10000.0
set MN4=90000.0
set MN5=90000.0
set MN6=90000.0
set WW2 Auto
set WN4 Auto
set WN5 Auto
set WN6 Auto
set zerowidth_tchannel False
set nevents=100
set mmjj=20
set ebeam1 50000
set ebeam2 50000
done

[Olivier Mattelaer (olivier-mattelaer)]

This sounds a bug (I guess model related).
Could you attach the model?
Without that I will not be able to help.

Thanks,

Olivier

[Mohamed Aly (moaly)]

hi Olivier,

Thanks for the quick response. Please find the model here:

https://drive.google.com/drive/folders/1H4lMk29tJ9rDjUtdY8T2ene84Ve42RI8?usp=sharing

Thanks,

Mohamed

[Olivier Mattelaer (olivier-mattelaer)]

Thanks for the model however I failed to reproduce your error.
Which version of python did you use. (python3 --version)
I have tested with 3.8 and with 2.7 and no issue on either of them.

However not that
1) set zerowidth_tchannel False
has no effect at that level of the code. To have this in effect, you have to put that command BEFORE the
" generate p p > mu+ mu+ j j / bsmhiggs @1" line

2) You seem to have singularity with your setup since the code returns the following cross-section
- Current estimate of cross-section: 2.4847437821776575e-14 +- 4.0022952788422104e-14

The error is larger than the cross-section which seems really bad (and typical of singularity) --even the fact that the cross-section is so small is a bit weird for singularity--).

Cheers,

Olivier

> On 29 Oct 2020, at 12:28, Mohamed Aly <email address hidden> wrote:
>
> hi Olivier,
>
> Thanks for the quick response. Please find the model here:
>
> https://drive.google.com/drive/folders/1H4lMk29tJ9rDjUtdY8T2ene84Ve42RI8?usp=sharing
>
> Thanks,
>
> Mohamed
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1901953
>
> Title:
> Auto width calculation throws key Error: 3
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> I am running a generation for a process p p > mu mu j j on MG5 V2.8.1
> which will contain :
>
> s-channel diagrams with W2 -> N mu
> t-channel diagrams with N exchanged between 2 W2 particles or 1 W2 and 1 W particle
>
> where W2 is a heavy right-handed gauge boson, N are heavy neutrinos
> and W are SM gauge bosons.
>
> Running the above process with MN= 90 TeV and MW2 =10 TeV gives the
> following error during the automatic width calculation for W2 and N:
>
> Command "generate_events run_01" interrupted in sub-command:
> "set zerowidth_tchannel True" with error:
> KeyError : 3
>
>
> Is this a bug or is this a result of the extreme ratio of masses of W2 and N? The MG5_debug file is attached.
>
>
> My generation commands are:
>
> import model MLRSM
> define bsmhiggs = h h2 h02 h03 h+ hp2 hl++ hr++ h3 a02 h- hm2 hl-- hr--
> define j = g u c d s b t u c d s b t
> generate p p > mu+ mu+ j j / bsmhiggs @1
> add process p p > mu- mu- j j / bsmhiggs @2
> output MW_10.0_MN_90.0_WRWR
> launch
> shower=pythia8
> done
> set MW2=10000.0
> set MN4=90000.0
> set MN5=90000.0
> set MN6=90000.0
> set WW2 Auto
> set WN4 Auto
> set WN5 Auto
> set WN6 Auto
> set zerowidth_tchannel False
> set nevents=100
> set mmjj=20
> set ebeam1 50000
> set ebeam2 50000
> done
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1901953/+subscriptions

[Mohamed Aly (moaly)]

Hi Olivier,

The error seems to occur when I use Python 3.6, 3.8, 2.7 with MG5 2.8.1 on 2 different machines.... When I use MG5 2.6.7 It does not seem to crash (however it is still running numerical integration for the width). Any idea why the error can just be on my side?

We are aware of the singularity which we believe arises from a lack of proper cuts!

[Mohamed Aly (moaly)]

Actually, I just noted that I had put the generation command missing a couple of lines to set 4 model parameters which when set to our preferred values the code breaks with the above bug. Does running the attached file reproduce the issue? I am really sorry about this oversight on my part.

Thanks,
Mohamed

[Olivier Mattelaer (olivier-mattelaer)]

I'm confused we are not supporting 3.6 neither 2.6 with 2.8.1

Cheers,

Olivierr

> On 29 Oct 2020, at 17:07, Mohamed Aly <email address hidden> wrote:
>
> Hi Olivier,
>
> The error seems to occur when I use Python 3.6, 3.8, 2.7 with MG5 2.8.1
> on 2 different machines.... When I use MG5 2.6.7 It does not seem to
> crash (however it is still running numerical integration for the width).
> Any idea why the error can just be on my side?
>
> We are aware of the singularity which we believe arises from a lack of
> proper cuts!
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1901953
>
> Title:
> Auto width calculation throws key Error: 3
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> I am running a generation for a process p p > mu mu j j on MG5 V2.8.1
> which will contain :
>
> s-channel diagrams with W2 -> N mu
> t-channel diagrams with N exchanged between 2 W2 particles or 1 W2 and 1 W particle
>
> where W2 is a heavy right-handed gauge boson, N are heavy neutrinos
> and W are SM gauge bosons.
>
> Running the above process with MN= 90 TeV and MW2 =10 TeV gives the
> following error during the automatic width calculation for W2 and N:
>
> Command "generate_events run_01" interrupted in sub-command:
> "set zerowidth_tchannel True" with error:
> KeyError : 3
>
>
> Is this a bug or is this a result of the extreme ratio of masses of W2 and N? The MG5_debug file is attached.
>
>
> My generation commands are:
>
> import model MLRSM
> define bsmhiggs = h h2 h02 h03 h+ hp2 hl++ hr++ h3 a02 h- hm2 hl-- hr--
> define j = g u c d s b t u c d s b t
> generate p p > mu+ mu+ j j / bsmhiggs @1
> add process p p > mu- mu- j j / bsmhiggs @2
> output MW_10.0_MN_90.0_WRWR
> launch
> shower=pythia8
> done
> set MW2=10000.0
> set MN4=90000.0
> set MN5=90000.0
> set MN6=90000.0
> set WW2 Auto
> set WN4 Auto
> set WN5 Auto
> set WN6 Auto
> set zerowidth_tchannel False
> set nevents=100
> set mmjj=20
> set ebeam1 50000
> set ebeam2 50000
> done
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1901953/+subscriptions

[Mohamed Aly (moaly)]

Okay let me clarify what I have done:

On CERN lxplus, the version of python3 I have is 3.6.8 (from python3 --version). When I then go to madgraph 2.8.1 directory and run python3 ./bin/mg5_aMC <attaced_txt>, then MG runs but throws the error in question during width calculation.

The same happens when I run MG from my local machine where I have python3 being version 3.8.3 and the same error still occurs during width calculation.

When I remove the following lines from the generation command

set k1=174.09
set VKe=0.03
set VKmu=0.03
set VKta=0.03

The error goes away on LXPLUS (Python 3.6) and on my local machine (Python 3.8).

I hope this clarifies the situation so far :) Sorry about any confusion.

Cheers,
Mohamed

[Olivier Mattelaer (olivier-mattelaer)]

Hi,

Do you use a CMS version of MG5aMC?

I know that they have removed the limitation of python version from requesting 3.7.
But we do not advise to run with python3.6, we have never tested our code with that version of the code. If I remember correctly we have one small segment which is not supporting python3.6 (but I do not remember which one) and when continuing our development we do not check/ensure compatibility with python3.6

For the bug, this is actually another thread related to that issue:
https://bugs.launchpad.net/mg5amcnlo/+bug/1902788
They are no solution available for that for the moment (and version without that bug are likely impacted by another bug returning a wrong value for the total width. So this means you can not use the auto-width feature for such model for the moment.

Cheers (and sorry),

Olivier

[Mohamed Aly (moaly)]

Hi Olivier,

Thanks for that. I will look into the Python version issue and try to avoid 3.6.

About the bug, from the linked thread, do I understand correctly that this bug will only affect auto width calculations for 1 -> 3 and 1 -> 4 decays with identical particles in the final states of the decay?

Cheers,
Mohamed

[Olivier Mattelaer (olivier-mattelaer)]

HI,

> this bug will only affect auto width calculations for 1 -> 3 and 1 -> 4
> decays with identical particles in the final states of the decay?

I do not know yet the exact spread of the current key error issue.
This issue was created by fixing another bug for 1 -> 3 and 1 -> 4
decays with identical particles in the final states of the decay.

So it would make sense that the key error issue is for a subset of
 1 -> 3 and 1 -> 4 decays with identical particles in the final states of the decay.
But this is not 100% sure.

Cheers,

Olivier

> On 9 Nov 2020, at 10:53, Mohamed Aly <email address hidden> wrote:
>
> Hi Olivier,
>
> Thanks for that. I will look into the Python version issue and try to
> avoid 3.6.
>
> About the bug, from the linked thread, do I understand correctly that
> this bug will only affect auto width calculations for 1 -> 3 and 1 -> 4
> decays with identical particles in the final states of the decay?
>
> Cheers,
> Mohamed
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1901953
>
> Title:
> Auto width calculation throws key Error: 3
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> I am running a generation for a process p p > mu mu j j on MG5 V2.8.1
> which will contain :
>
> s-channel diagrams with W2 -> N mu
> t-channel diagrams with N exchanged between 2 W2 particles or 1 W2 and 1 W particle
>
> where W2 is a heavy right-handed gauge boson, N are heavy neutrinos
> and W are SM gauge bosons.
>
> Running the above process with MN= 90 TeV and MW2 =10 TeV gives the
> following error during the automatic width calculation for W2 and N:
>
> Command "generate_events run_01" interrupted in sub-command:
> "set zerowidth_tchannel True" with error:
> KeyError : 3
>
>
> Is this a bug or is this a result of the extreme ratio of masses of W2 and N? The MG5_debug file is attached.
>
>
> My generation commands are:
>
> import model MLRSM
> define bsmhiggs = h h2 h02 h03 h+ hp2 hl++ hr++ h3 a02 h- hm2 hl-- hr--
> define j = g u c d s b t u c d s b t
> generate p p > mu+ mu+ j j / bsmhiggs @1
> add process p p > mu- mu- j j / bsmhiggs @2
> output MW_10.0_MN_90.0_WRWR
> launch
> shower=pythia8
> done
> set MW2=10000.0
> set MN4=90000.0
> set MN5=90000.0
> set MN6=90000.0
> set WW2 Auto
> set WN4 Auto
> set WN5 Auto
> set WN6 Auto
> set zerowidth_tchannel False
> set nevents=100
> set mmjj=20
> set ebeam1 50000
> set ebeam2 50000
> done
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/mg5amcnlo/+bug/1901953/+subscriptions

[Olivier Mattelaer (olivier-mattelaer)]

This patch/version should fix the issue:
https://bazaar.launchpad.net/~maddevelopers/mg5amcnlo/LTS_dev/revision/327

Thanks,

Olivier