Masses of final particles are different from the on-shell

Asked by Andrew McAdams

Hi! I have simulated the process

generate bs2 > b+ k-, (b+ > n4 d0bar e+, d0bar > pi- k+)

Here, Bs2, B+, D0bar are SM B and D mesons, while N4 is a heavy neutrino. The widths of the mesons that I use correspond to the realistic widths (such that for the B,D mesons, it is ~10^-13 GeV). The output typically contains the following events:

535 -1 0 0 0 0 +0.0000000000e+00 +0.0000000000e+00 +9.9999998245e-15 5.8398600000e+00 5.8398600000e+00 0.0000e+00 0.0000e+00
      521 2 1 0 0 0 -6.5795548681e-01 -8.4416596065e-01 +1.0700204393e+00 4.2479484721e+00 3.9692065819e+00 0.0000e+00 0.0000e+00
    18000 1 2 2 0 0 +3.0777993975e-01 +3.8369413528e-01 +4.8384709175e-01 1.2149311458e+00 1.0000000000e+00 0.0000e+00 -1.0000e+00
     -421 2 2 2 0 0 -1.5780823871e+00 -6.5419001595e-01 +6.1967708172e-01 2.1932620512e+00 1.2280432168e+00 0.0000e+00 0.0000e+00

I noticed a huge difference of the masses reported from the on-shell mass of the particles: 3.96 for the B-meson (should be ~5.3), 1.22 for the D0 meson (should be ~1.9).

Could you please tell me what is the reason and how to fix this issue?

Question information

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


Width set to 1e-13 is something that is numerically unstable (we do everything in double precision) and that we can not handle for sure.

Now we have some security to artificially increase the width to avoid the numerical issue but I do not know if such security are present only for 2>n process or also for 1>n process.

So my bet is that you have numerical issue. If you put your model available online, I would be able to take a deeper look.

Revision history for this message
Andrew McAdams (mcadams1994) said :

Thanks, Olivier! The problem was that during making UFO some decay widths have been set from something like 1e-13 to 1. Restoring correct parameters fixed the issue.