Peculiar identical events from two similar models

Asked by Jonathan Guiang


I’m trying to generate events for two different BSM models: (1) the W+W-H coupling is negative and (2) the ZZH coupling is negative. I will refer to models (1) and (2) as sm_mkw and sm_mkz respectively. For each model, I’ve changed just one line in the SM MadGraph model. In particular, I made the following change to models/sm/ for each model:

GC_72 = Coupling(
    name = 'GC_72',
    value = '-((ee**2*complex(0,1)*vev)/(2.*sw**2))', # changed; added a negative sign
    order = {'QED':1}

GC_81 = Coupling(
    name = 'GC_81',
    value = '-(ee**2*complex(0,1)*vev + ... )', # changed, just added a negative sign, omitting the rest of the string for brevity
    order = {'QED':1}

Then, for either model, I generate 10,000 events using the following proc cards:

import model sm
import model sm_mkw
define w = w+ w-
generate p p > w h j j QCD=0

import model sm
import model sm_mkz
define w = w+ w-
generate p p > w h j j QCD=0

Importantly, I use the same seed to generate both sets of 10,000 events. This all seems to run fine, and I am left with two LHE files. What confuses me, though, is the following: there are 6,014 events that are exactly identical between the mkw and mkz samples. Now, I expect the physics to be identical on paper. However, at this step of simulation, I expect the events to either 100% identical or 0% identical (i.e. the same kinematics but not the exact same values). The fact that 60% of them are identical concerns me--perhaps I did something wrong? In case it’s helpful, I’ve put the files I used to check this here[1].


Question information

English Edit question
MadGraph5_aMC@NLO Edit question
No assignee Edit question
Solved by:
Olivier Mattelaer
Last query:
Last reply:
Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :


Which version of MG5aMC are you using?
and if this is 2.9.0 or higher, what is the value of sde_strategy that you pick in the run_card?

The default should be "2" for which I have hard time to explain this, why for "1" (which is the strategy before 2.9.0)
this can make sense.



> On 4 May 2022, at 03:31, Jonathan Guiang <email address hidden> wrote:
> define w = w+ w-
> generate p p > w h j j QCD=0

Revision history for this message
Jonathan Guiang (jguiang) said :

Hi Olivier,

Thanks for the quick response. I'm using MadGraph v2.6.1, so I don't see sde_strategy in the run card.


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

Ok that's a very old version of the code,
I guess that even if you run the same theory with the same seed you will not get 100% identical event within that version.
The reason is that in some version, the seed was only impacted the fortran side (the event generation) but not the python side (the event un-weighting)



Revision history for this message
Jonathan Guiang (jguiang) said :

Indeed it is an old version of the code. I confirmed that with the same theory (same proc card, same run card), I get only ~60% of identical events again. In any case, it sounds like I shouldn't be worried then. Thanks for the information and the prompt responses!


Revision history for this message
Jonathan Guiang (jguiang) said :

Thanks Olivier Mattelaer, that solved my question.