Symmetry factor missing for decay processes with symmetry between diagrams

[Johan Alwall]

Hello,

I have a question about a behavior of Madgraph.

I implemented a model using FeynRules. It is the standard model (the one implemented in the FeynRules website) with two extra states called X83 and X53. When I compute the decay width of the X83, I have the following behaviors:

The decay chain goes has followed X83 > X53 w+ ,(X53 > w+ t). So the final states is t w+ w+ and there is one off shell propagator X53. (I have the same behavior for the decay X53 > w+ t,(t>w+ b) if I set the decay width of the top to 0.

If the decay width of X53 is set to 0 (not really physical I confess) the cross section is exactly two times bigger than a X53 with a width of 0.001 gev. (The mass of the two states is 400 gev)

I checked the matrix element for one point in the phase space using the output standalone command and there are almost the same in the 2 cases(the difference is only due to the 0.001 width.

Having done the computation using Mathematica, the good result seems to be the one with 0 width (if I have not forgotten a factor of two.

I am quite confuse by this behavior. Where do you think it come from?

(if you want I can send you the files)

Thank you for your time

Thibaud

[Johan Alwall (johan-alwall)]

Hello Thibaud,

Many thanks for the report. This was due to a poorly formatted read-in in the file SubProcesses/P.../auto_dsig.f, which only affects decay processes (of the type A > B C ...) when some intermediate resonance is unable to go onshell due to the masses of initial and final state particles AND there are multiple identical particles in the final state, giving symmetric diagrams. So basically, only for exactly the process you are calculating. So indeed the factor 2 should be there, and the zero-width result is the correct one.

You can fix this once and for all in your MG5 directory by replacing the following lines in the file madgraph/iolibs/template_files/auto_dsig_v4.inc:

On line 41, insert the following line:
      DOUBLE PRECISION RCONF

On line 78, replace the line
          READ(LUN,*,ERR=10,END=10) ICONF, IFACT
with
          READ(LUN,*,ERR=10,END=10) RCONF, IFACT
          ICONF=INT(RCONF)

I have committed this fix to the branch 1.5.6. This version should get released soon.

Thanks again, all the best,
Johan

[Thibaud Vantalon (thibaud-vantalon)]

ok thank you very much for your quick answer.