One possible cause of the observed discrepancy between data and simulation in the BEMC is the insufficiency of the nominal tracking parameters in GEANT.  This brief study investigates this hypothesis and shows that there is no indication that tuning the tracking parameters will improve simulation.

Tracking parameters in the BEMC are set in pams/geometry/calbgeo/calbpar.g.  Here a local copy of pams/geometry was checked out and pams/geometry/calbgeo/calbpar.g modified for each tune.  The local library was selected in starsim with the command 

        gexec .sl44_gcc346/lib/geometry.so

        gclose all

Each study utilized the same 100,000 single photons randomly sampled from the phase space

        4 GeV < Et < 7 GeV

        v = 0

        -0.6 < eta < -0.55

        -0.046866 < phi < -0.41866

Note that the eta/phi constraints correspond to the BEMC tower 3932.  Due to technical issues within starsim, low energy electrons were also thrown at the opposite side of the detector.

After reconstruction, the StGammaMaker was run to extract information around each photon.  Clustering parameters were 

        Eseed = 3.0 GeV

        Et = 4.0 GeV

The three tunes are defined as

No difference was seen between the three reconstructed samples.  Differences would manifest in shape variables, especially f = Eseed / E and the energy weighted RMS of the SMD strips,