DIPOLE_ALPHA in event generation broken

[marek]

Using the Examples/NLO_W setup with different DIPOLE_ALPHA parameters (1./0.1/0.01) gives wrong results in the event generation. The integration is fine. This problem occurs in rev 17291 and is unaltered by commit 17292.

[Stefan Hoeche]

I have the strong suspicion that this problem is related to the changes in c17171, especially the removal of these lines in Single_Dipole_Term.C

  m_realevt.m_me   = m_realevt.m_result =
    m_realevt.m_last[0] = m_realevt.m_last[1] = 0.0;

This is what fixed the broken DIPOLE_ALPHA treatment in 1.2.3.

[Stefan Hoeche]

Apparently, c17171 is not responsible for this problem. However, what I observe with r17170 is that for several values of DIPOLE_ALPHA (0.1, 0.01, 0.001) the cross sections in event generation agree within statistical errors (<1% for 0.1 and 0.01, about 10% of the discrepancy). They also agree with the integration results. Only the result from default DIPOLE_ALPHA disagrees with all other values. Setting DIPOLE_ALPHA=0.999 on the other hand gives the same problem.

[marek]

after carefully checking rev. 17292 for a wide set of alpha cuts with 100k events. everything seems to be fine:

DIPOLE_ALPHA XS_Integration(B+VI+RS) XS_Event_Generation

1. 1098+214+ 19 1330 +- 3 (0.2 %) 0.999 1098+214+ 19 1330 +- 3 (0.2 %) 0.9 1098+214+ 19 1330 +- 3 (0.2 %) 0.5 1098+206+ 27 1331 +- 3 (0.2 %) 0.2 1098+144+ 89 1331 +- 3 (0.2 %) 0.1 1098+ 44+190 1336 +- 10 (0.7 %) 0.01 1098-666+899 1383 +- 40 (2.9 %) 0.001 1098-1986+2203 1487 +- 183 (12.3 %)

so no problem there. The W+j setup also works. I do not see any problem with the default DIPOLE_ALPHA=1, or DIPOLE_ALPHA=0.999. Please reopen, if you still see this problem.

[Stefan Hoeche]

Milestone rel-1.3.0 deleted