Generator level study of the "M(ee) problem"

A disagreement was observed in the M(e+e-) distributions at generator level between HZHA and PYTHIA, namely HZHA has a much more important tail at low masses. A third generator, WPHACT, agrees with PYTHIA. The origin of the difference can be attributed to the treatment of the final state radiation.

The fsr in HZHA is introduced explicitely, while in the other programs it is assumed to be performed by JETSET at the parton shower phase. This in fact turned out not to be true, since the parton shower is active for a parton pair and not for leptons: this means that, even setting mstj(41)=2, there is no emission of photons from electrons (or other charged leptons).

In order to check that the HZHA tail is really related to the fsr, the M(ee) mass distribution has been redefined, including the contribution from photons flying within a cone of selectable opening angle around either electron. The first ps file shows the original mass distributions for the HZHA sample, the modified distributions using cones of respectively 5 and 10 degrees, and the WPHACT distribution, both in lin and log scale. It is clear that the disagreement decreases with increasing opening angle, although remaining not negligible even with the 10 degrees cone.

When looking at fully simulated MC data (as well as real data) it is clear that some of the fsr is collected in the electron showers either directly or after a "dressing" procedure, so it is worthwhile to have a look at the difference between generators using the reconstructed electrons of a physical analysis. For this purpose a second .ps file shows HZHA vs WPHACT plots. In this case one can consider the unfitted or the 4C or the 5C fitted masses. Since our standard ntuple does not contain the 4C fit result, only the unfitted and 5C fitted masses are shown. The used events are selected after all our standard cuts and in addition a combined fit probability > 10% is required. While the unfitted distributions are essentially undistinguishable, the 5C shows, in the log plot, a remnant of the original difference, that should not have any practical influence. Note that in case of 5C fit the M(ee) is not a very significant variable, what really matters is the fit probability.

Conclusions: there are significant differences at generator level, and our standard generator (HZHA) seems in a good shape since it does include fsr while other presently don't. The difference is anyway small after full simulation and reconstruction. In that case it is clear that some of the fsr effects are readsorbed, but at the same time the experimental resolution does not allow to apply strong cuts on the unfitted e+e- mass. The conclusion could be somehow different considering the 5C fitted mass, but this is a rather delicate point since the fit procedure using a more or less BW distributed contraint is not yet quite well established.

NEW: Philip Bambade suggested that in order to trigger the fsr another jetset/pythia switch (MSTP(71)=1) could be necessary. This is the default according to the manual, but actually the special JETSET library with DELPHI tuning used for hadronization within other generators does not set that switch. To check its possible influence a separate WPHACT production has been done using the standard JETSET library. As shown in another plot no effect is produced by this switch using JETSET hadronization. Things could be different when using PYTHIA as a self contained generator.

Click here for an update of this study.