flavor sum over different masses passes through unchecked

[fbach]

I observe an awkward instability of the total integrated cross section of LHC single top production in the t-channel from intrinsic bottom and valence quarks (i. e. signal process 'u b -> lbar v d b') when further (insignificant) non-signal processes are included in the process definition (refer to 'foo.sin' and 'foo.log' attached for further description):

1) the signal processes (ubdb/bubd) converge well and have stable x-sec's ~1000 fb

2) the non-signal processes with b -> bbar (ubdbx/bubdx) still converge acceptably, giving ~10 fb

3) summing over these processes in the process definition (ubbdbb1 - ubbdbb4) spoils the convergence and makes the total x-sec (containing the signal!) drop drastically

4) this is very sensitive to the LOWER cut on |eta| of the light jet, i. e. convergence problems and dropping of the x-sec get worse in the forward directions

5) it depends on the initial beam assignment: ubbdbb1 still converges to the expected x-sec while ubbdbb2 - ubbdbb4 show the described instability

6) same observations hold for replacing u -> c and d -> s for the non-signal (bbar) processes, but no problems when crossing the light down- and up-type quarks in the initial and final state

7) for tbar production (signal process 'd bbar -> l vbar u bbar'), this instability under addition of small non-signal processes vanishes entirely (cf. 'foo2.sin' and 'foo2.log')

[Juergen Reuter]

Probably connected to the phase space tickets 389 and 390.

[Juergen Reuter]

One question: are you using the newer SM.mdl files which do no longer contain the Hss and Hcc couplings for the phase-space setup?

[fbach]

Replying to jr_reuter:

One question: are you using the newer SM.mdl files which do no longer contain the Hss and Hcc couplings for the phase-space setup?

I was running on the latest trunk, so I think everything was up to date...

[kilian]

Good example. I have to see how to find the right phase space here.

[kilian]

A second look, less superficial:

(1) All results are crap, look at the efficiency. (Maybe the cross section value is ok nevertheless.)

(2) The processes ubbdbb1 and ubbdbb2 can't possibly work unless the b mass is zero. But they remind me of re-activating the check for illegal flavor sums.

I have to check these with the modified phs module.

[kilian]

There are two problems here:

(1) The check for identical masses in flavor sums is missing (I'll add this)

(2) O'Mega rearranges the final state, with fatal consequences (#404)

[kilian]

(1) fixed in r3110. Immediately showed that two of the self-tests had wrong flavor sums (also fixed). To be considered again and hopefully closed after #404 has been accounted for (ranking down).

[Juergen Reuter]

What about the tau masses in the self test examples? The lepton alias includes them, but the tau mass is not zero.

[Juergen Reuter]

Ok, the alias is only in the cut, not for the process. Understood.

[kilian]

The mass check works, and the flavor-sum ticket is closed (#404).