Procedure and results for optimizing some of the cuts for EEmc pi0 reconstruction.

Procedure

The reconstructed pi0s are assigned a type according to this blog.  Afterwards, cuts are applied to both the generated pi0s and the reconstructed pi0 candidates.  The following cuts are considered:

1. min photon energy
2. max preshower 1 energy (for each photon)
3. max preshower 1 energy over tower energy (for each photon)
4. max preshower 2 energy (for each photon)
5. max preshower 2 energy over tower energy (for each photon)
6. max postshower energy (for each photon)
7. max postshower energy over tower energy (for each photon)
8. max SMD energy over tower energy (for each photon)

For all cuts, we consider a high pT range of 10-12 GeV and a low pT range of 5-6 GeV.  For the high pT range, the pT resolution used in setting the type is 1 GeV, while for the lower pT range it is set to 0.5 GeV.  For all cuts after the 1st cut, a photon energy cut is applied (2 GeV for TSIU and IU, 2.5 GeV forTSP).  For all cuts after the second cut, a max preshower 1 energy cut of 40 MeV is applied.  The digamma mass is also required to be within 0.1-0.2 GeV.

Data

The highers partonic pT bin from the QCD background sample is used.   While using one partonic pT bin does not accurately produce the true pT distribution of pi0s, it is sufficient for optimizing the above cuts.

Presentation of Results

One page of results is produced for each cut (1-8), pT region (low and high) , and algorithm (IU, TSIU, TSP).  The page is arranged in two columns and four rows.

The columns represent a different choice of definition for reconstructed "signal" and "background".  For the left column, only the purest PI0_GOOD_PT type is considered signal, while all other types are considered background.  For the right column, signal types include all the pi0 types (PI0_GOOD_PT, PI0_BAD_PT, PI0_1_TO_M, PI0_M_TO_A) and background includes the other two types (PHOTON_CONV, OTHER_PAIR).

The rows represent different criterea that may be used for optimizing.  The first row is the background fraction (reconstructed background / reconstructed signal and background ).   The second row is the efficiency (reconstructed signal over generated signal).  The third row is the counts for the reconstructed signal, and the fourth row represents a quantity proportional to an estimate of the statistical uncertainty on the cross section.  Ignorning uncertainties due to the background fraction and background subtraction, the statistical uncertainty on the cross section is proportional to the inverse of the efficiency divided by the square root of the number of signal counts.

Results

All the various pages are attached at the bottom, grouped by algorithm and pT region.  In most cases there is not an obvious good choice for the cut.  The following are the conclusions, i.e. decision regarding the above cuts:

Photon Energy Cut

IU: 1.5 GeV  (2.0 might be a little better, but not worth changing from the historic choice of 1.5 GeV)

TSIU: 2.0 GeV

TSP: 2.5 GeV

Preshower 1 Max Energy Cut

For all algorithms, the choice of 40 MeV is still OK.  This study doesn't demonstrate that it helps, but does show that it doesn't hurt.  To remove some of the messiest real events, we have decided to keep using this cut.

SMD Energy / Energy Cut

For all algorithms, a cut of 0.006 looks optimal.

All Other Cuts Herein Investigated

Not implemented, as it doesn't appear they would be very benificial.