Below are SMD plots made from 5000 events each of gammas and piminus.  First tower clustering is done by identifying all seed towers with energy > 1 GeV and then finding the 2x2 patch of towers including the seed tower with the highest energy.  If this 2x2 patch has energy > 4 GeV, it is saved as a cluster.  Then for each cluster the SMD eta and phi strips with the highest energy that lie within the eta and phi range of the cluster towers are identified.  If this strip has energy > 10 keV it is saved as the center of a cluster which is defined as the center strip and the three strips on either side.  The SMD cluster energy is then defined as the sum of the energies of these strips, the SMD cluster eta (or phi) as the energy-weighted eta (or phi) value of the cluster, and the cluster 2nd moment in eta (or phi) defined as around the cluster eta (or phi).

Figure 1 below shows the distributions of these quantities, for both SMD eta and phi, for the photon sample.  Figure 2 shows the eta and phi cluster energies and 2nd moments plotted vs. the tower cluster energy for the photon sample.  Figure 3 shows the eta and phi cluster energies and 2nd moments plotted against each other, again for the photon sample.  Figures 4-6 show the same plots, but for the piminus sample.

Requiring Smd cluster energy > ~75 keV seems like a reasonable first approximation at a cut.  More speculatively, it might be possible to make a cut such as log (2nd moment) < 10^-4.5: however, the 2nd moment peaks for gamma and piminus are closer than the cluster energy peaks are, making it more difficult to discriminate.

Figure 1:

Figure 2:

Figure 3:

Figure 4:

Figure 5:

Figure 6: