Comparison of conversion rates in MC to data, take 2

Abstract: We compare the conversion rate for simulated π0 events in the 1st radiator of the endcap to data.  We find a conversion rate of 88% +/- 4% for the simulation, compared with 91.5% +/- 0.3% for the data, where we count any signal w/in R<0.3 of the pion candidate as a conversion.


0. Monte Carlo Data Sample

EEMC Jet/QCD backrgound events from the MIT production.
Sample pT min pT max weight
mit0040 2 3 8.12
mit0041 3 4 7.66
mit0042 4 5 7.00
mit0043 6 9 0.587
mit0044 9 15 0.844
Last partonic pT bin yielded precisely 3 reconstructed pions after CPV, so we ignored it for the current study.
Processed files located here...

... and here...


1. Event Reconstruction

The pi0 maker is configured with low seed thresholds, since we will be performing isolation and charged-particle cuts on the data which will kill off backgrounds.

Trigger simulation

  • Use StTriggerUtilities for trigger simulation
  • Timestamp: dbMk->SetDateTime(20060522, 55000); // timestamp R7142018
  • Simulate trigger ID 137641  (n.b. this is only one of the triggers used in the data... need another pass at this)



2. Event Selection

Reconstruct pi0 candidates, and analyze events within the window 110 MeV to 170 MeV.
  • Isolation cut: Sum energy w/in R<0.3 of the leading photon in the pair.  Require 90% of energy in the tower containing the leading photon.
  • CPV cut: Sum energy deposited in preshower-1 w/in R<0.3.  Veto events with energy > 0
  • Analysis cut: Sum energy deposited in preshower-2 w/in R<0.3.  Increment conversion histogram(s) when energy > 0 (3 sigma > ped).

3. Results

Figure 3.1 below shows the invariant mass spectrum reconstructed for each of the MIT data samples.  Note that each of these samples corresponds to a different partonic pT bin, as given in the table.  The last of the six plots shows the combined spectrum.  The 5 partonic pT samples are combined, weighted appropriate to the luminosity reported by pythia.  The mass spectrum is scaled to represent 1pb-1 of data.
Figure 3.1 -- Isolated, two-photon Invariant mass distribution reconstructed for each of the partonic pT bins in the MIT event sample.  The solid histogram indicates events which pass the isolation cut and the charged-particle veto (CPV).  The dashed lines indicate the events which register energy in the 2nd preshower layer within a radius R<0.3 of the pion candidate.

For each of the above mass spectra, we calculate the conversion efficiency ε. This is the ratio between the number of events which pass the CPV to the number of events which register energy in the 2nd preshower layer. This ratio is plotted in figure 3.2, and compared with the measured conversion rates for pi0 and eta decays in the data.  The most relevant comparison is between the conversion rate for the weighted spectra and the pi0.

Figure 3.2 -- Conversion rate observed by the 2nd preshower layer for isolated pions, w/in the mass window 110 to 170 MeV.  The rate is plotted for the various partonic pT samples, and compared to the values extracted from the data for the pi0 and eta samples.

4. Discussion

The Monte Carlo exhibits a conversion rate which is (nearly) independent of which partonic pT bin it arises from.  This suggests that we are "ok" to ignore trigger simulation for the moment... (will include it in the a future study).  In the data, both the pi0 and eta event samples show similar conversion rates to the simulate pi0s.

Alice has examined the background conversion rates for events which pass gamma ID cuts.  Her study showed a lower conversion rate than what we are seeing for pi0 decays.  While her study used tighter gamma ID cuts than mine, I suspect that these results indicate that π0 decays are not going to be a good surrogate to determine the background conversion rates in prompt photon samples.  That being said, I still need to check the following:

1. When I compare the same set of triggers between data and MC, do I still get consistent conversion rates?
2. Hadronic triggers appear to be greatly suppressed in identified pi0 samples.  Does this adequately explain the dilution of the conversion rates?
3. When consistent cuts are applied to the gamma background sample, do we see consistency with the pi0 conversion rates?