To summarize what was said at yesterday's meeting, I found that the MC
production where the ks bug was originally present and then has been
fixed, still showed a nasty discrepancy.
Apparently the spurious pi0 daughters have actually been removed, but
there seems to be an eccess of clusters in an intermediate energy region.
To study the effect, I have considered two sets of runs, a bunch before
the bug fix (but with the Beta patch turned on) and a bunch after it
(which I will call "noBug" as oposed to "Bug" MC).
I have then plotted the energy spectrum of the photons coming from the Ks.
In
http://www.slac.stanford.edu/~rfaccini/phys/vub/soff/ksGam.eps
the "Bug" distribution (crosses) is overlayed to the noBug one normalized
to the same area. It is clearly harder. As a check I have produced the
same plot for all other photon candidates
http://www.slac.stanford.edu/~rfaccini/phys/vub/soff/otherGam.eps
and it is identical.
In order to decide how to fix this, I have normalized the ks daughters
energy spectrum to the number of Ks->Pi0Pi0 present in the sample and I
got http://www.slac.stanford.edu/~rfaccini/phys/vub/soff/ksGamNorm.eps
which shows that the Bug configuration always has too many clusters.
I therefore suggest the following fix: for the SP4bug sample we should
kill daughters of true Ks->pi0pi0 according to the fit to the ratio
between the two histograms normalized to the number of Ks.
As a fitting function I am using a gaussian offset by 1, in order to make
sure that at very high energy no correction is applied.
Note that in the bug fixed MC there are 2.1 reconstructed photons per Ks,
while in the buggy one 2.6. This means that the average killing is about
20%.
As a result the two spectra, normalized to the number of Ks (and with a >2
times bigger statistics) are shown in
http://www.slac.stanford.edu/~rfaccini/phys/vub/soff/ksDauNewNorm.eps
comments are welcome
I will do the standard data-MC agrement checks and then commit the fix
ciao
ric
|