Amegic vs. Comix deviations in WWbb

[Frank Siegert]

The LO result for WWbb with cuts according to arXiv:1012.3975 can be reproduced by Comix with the Run.comix.dat attached, but not with Amegic (Run.amegic.dat):

Generator	mu=mt (LO, LHC)	mu=0.5*mt (LO, LHC)	mu=2*mt (LO, LHC)
arXiv:1012.3975	662.353 fb +- 0.03362 fb	925.8 fb	488.3 fb
Comix	661.687 fb +- 0.990112 fb	923.175 fb +- 1.81227 fb	488.364 fb +- 0.878304 fb
Amegic	655.878 fb +- 0.503296 fb	916.208 fb +- 1.17167 fb	483.39 fb +- 0.544478 fb

Run cards are basically identical except for the generator and Order_EW switches. They will only work with recent trunk (r18704 for the numbers above).

[Frank Siegert]

The difference is not related to any of the complications of the original run cards. I am attaching two more simplified run cards, without complex-mass-scheme etc., which show the same effect. I wonder whether this is related to gauge warnings that I see for the gluon-gluon channel in Amegic:

AMEGIC::Single_Process::Tests for 2_6__G__G__W+[nu_e__e+]__W-[mu-__nu_mub]__b__bb
   Prepare gauge test and init helicity amplitudes. This may take some time.
WARNING:  Gauge test not satisfied: 5.92959e-09 vs. 5.92954e-09 : 0.000968657%

If I integrate qq and gg separately with the simplified run cards, then I do see compatibility for qq and a difference for gg:

generator	qq	gg
amegic	0.198086 pb +- 0.000375282 pb	0.689186 pb +- 0.000231988 pb
comix	0.197797 pb +- 0.000425573 pb	0.695662 pb +- 0.000652262 pb

Leaving out the W decays and setting STABLE[24]=1 and WIDTH[24]=0 the Amegic gauge violations remain there at a similar level:

AMEGIC::Single_Process::Tests for 2_4__G__G__W+__W-__b__bb
   Prepare gauge test and init helicity amplitudes. This may take some time.
WARNING:  Gauge test not satisfied: 6.59091e-11 vs. 6.59088e-11 : 0.000410332%

but I can't reproduce the differences between Amegic and Comix anymore, at least not within the integration accuracy that I have achieved so far.

[Frank Siegert]

Since this seems to be an Amegic issue, do you have any idea (for further checks), Steffen?

[Steffen Schumann]

Naively I have no explanation for that behaviour. I can just guess that it is related to the treatment of width though. I will look into that asap!

[Frank Siegert]

Ok, thanks! Just for the record: I've tested Simplified.amegic.dat with version 1.3.1 and it gives basically the same result (0.689954 +- 0.000137961).

[Frank Siegert]

Another one for the record: I've also tested Amegic's old default gauge choice AMEGIC_DEFAULT_GAUGE=1 and the problem persists (655.7 +- 0.7 fb for Run.amegic.dat).

[Steffen Schumann]

Just a brief update: Seemingly the problem shows up only in the 'DecayOS' treatment for the W decays. Using the full propagator via 'Decay' the results of Comix & Amegic agree nicely. Further, the gauge test is not really pointing to a problem here. Considering WWbb and setting the top width to zero the gauge test is fine. I will now try to figure out what the two codes do differently when using on-shell decays.

[Stefan Hoeche]

Could you please try whether the problem persists with trunk r18737? It might also be related to ticket #201.

[Steffen Schumann]

I don't no yet what makes the difference but the problem seems to be solved in the current version. I will do some more checks before hopefully closing this ticket! Thanks Stefan!

[Stefan Hoeche]

Happy to hear that. The solution is unfortunately not a general one though. For some of the subtraction methods to work, I had to revert Amegic's new gauge choice (10) to the old one (1), which is now the default again. In itself this would not present a problem. But if spin correlations are ever to work, we are faced with the problem of solving ticket #201, i.e. sorting out the polarisation vector issues in Basic_Sfuncs for gauge choice 10.

[Frank Siegert]

Since this issue unfortunately seems to remain open even with the old gauge choice (1)...

I can see a related issue if I try to calculate the R-S NLO contribution. For offshell W decays the integration is stable, i.e. the singularities cancel fine. For onshell W decays the integration is divergent right away. I can provide the corresponding setup in case that helps anything.

[Frank Siegert]

I have one more much simpler issue which I think could very well be related: The gauge test for the following setup fails for Amegic (at the 0.000410432% level), while it seems to work fine with Comix.

(run){
  BEAM_1 = 2212; BEAM_ENERGY_1 = 3500.;
  BEAM_2 = 2212; BEAM_ENERGY_2 = 3500.;

  ME_SIGNAL_GENERATOR = Amegic

  WIDTH[24]=0.0
  WIDTH[5]=0.0
}(run)

(processes){
  Process 21 21 -> 24 -24 5 -5;
  Order_EW 2;
  End process;
}(processes);

(selector){
  NJetFinder 2 30.0  0.0  0.5  -1  2.5
}(selector)

Is there any reason, why this should not be gauge invariant?

Cheers, Frank

PS: Tested with trunk r19289 and r19401

[Steffen Schumann]

The gauge invariance issue solved in r19436. However, in order to preserve gauge invariance the complex mass scheme has to be used, WIDTH_SCHEME=CMS.