A Study of Multi-Shower Discrimination in the New Muon $(g-2)$ Calorimeter

We have used a GEANT3 simulation of the New Muon ($g-2$) Calorimeter to study discrimination between single $e^+$ and double $e^+$ electromagnetic showers. The calorimeter was defined as a $20\times 16\times 16$RL cm$^3$ box with alternating layers of plastic scintillator and tungsten. Assumed energy resolution was matched to the measured resolution of a detector prototype. An artificial neural network algorithm (ANN) was trained to discriminate between a single positron shower and the background events. The five ANN input variables were: (1) maximum energy deposited in the single module $E_{\rm M}$, (2) module index with maximum energy deposition $i_{\rm M}$, (3) energy deposition $E_{\rm R1}$ in the ring of modules around $i_{\rm M}$, (4) energy deposition $E_{\rm R2}$ in the next-to-nearest-neighbors around $i_{\rm M}$, and (5) outer energy sum $E_{\rm OUT}$. The efficiency of the ANN algorithm in recognizing a single $e^+$ signal while suppressing false positives (two $e^+$'s misidentified as a single) was studied as a function of energy and the calorimeter segmentation. The results were used as an input design parameter for the 2D tracking calorimeter hodoscope.