The difference between decay in MadGraph and decay in MadSpin

Asked by Artur on 2020-05-09

Dear MG5 experts,

I'm generating process p p > z a j j QCD=0 QED=5, z > vl vl~ and get the following cross sections:
0.00714+-0.00002 pb (1) - with default sm model ;
0.0867+-0.0002 pb (2) - with 8-dim EFT model SM_LT012_UFO using f_{T0}=2.9*10^{-12} GeV^{-4} (

Also I can generate this process in an alternative way:
1. Firstly by generating p p > z a j j QCD=0 QED=5, which gives the following results:
0.03555+-0.00008 pb - in sm;
0.4449+-0.0007 pb - in SM_LT012_UFO.
2. Secondly by using MadSpin (./Madspin/madspin) to decay z > vl vl~. After MadSpin I have the following results:
0.00724+-0.00002 pb (3) - in sm;
0.0918+-0.0002 pb (4) - in SM_LT012_UFO.

How can be explained differences in cross sections (1) and (3), (2) and (4)? These differences can not by covered by statistical

With best regards,

Question information

English Edit question
MadGraph5_aMC@NLO Edit question
No assignee Edit question
Solved by:
Olivier Mattelaer
Last query:
Last reply:


MadSpin use the Narrow-widh approximation for the computation of the cross-section (so it just use the Branching ratio information) while the decay chain syntax does not use this approximation for the computation of the cross-section.
You can therefore expect difference of the order of width/mass between the two (when Narrow-widh approximation) and the difference can be larger if you have issue with such approximation.

Other possible reasons are
- not same cut on the final states between the process
- not same dynamical scale between the two computations



Artur (asemushi) said : #2


Thanks for your answer. I have a few more questons.
Which result is more correct: MadSpin or MadGraph?
About other reasons: I have no cut on MET in run_card, so, I think, first reason is not possible. I use dynamical scale choice=3 in MadGraph, how can I use the same dynamical scale in MadSpin?

With the best regards,

If Narrow-width approximation is not valid, then both are wrong.
If narrow-width approximation is valid then both should be equivalent to Gamma/M effect (which should be very small to have the approximation valid)

If "Gamma/M" is not that small then MadEvent is more correct since it can correctly includes offshell contribution that can not be catched by MadSpin. But if this effect is significant, you are in trouble.

One more reasons that you can be wrong in your comparison is the value of the total width/partial width.
if those are not set to the LO (partial) width you can get some differences obviously (which means that your param_card was not physical).



Artur (asemushi) said : #4

Thanks Olivier Mattelaer, that solved my question.