EID and fake rate

Electrons are identified by using the following discriminants:

• the ratio of energy deposited in ECL and charged track momentum measured by CDC,
• transverse shower shape at ECL,
• the matching between a cluster at ECL and charged track position extrapolated to ECL,
• $dE/dx$ measured by CDC,
• light yield in ACC, and
• time of flight measured by TOF.

We made probability density functions (PDF) for the discriminants beforehand. Based on each PDF, likelihood probabilities are calculated with track-by-track basis, and unified into a final likelihood output. This likelihood calculation is carried out taking into account the momentum and angular dependence. Fig.  shows the output from the above procedure. Closer to unity the particle is more likely to be an electron. The solid (dashed) histogram shows for $e^{\pm}$ in $e^{+}e^{-} \rightarrow e^{+}e^{-}e^{+}e^{-}$ data ($\pi^{\pm}$ in $K_{S} \rightarrow \pi^{+}\pi^{-}$ decays in data). The clear separation can be seen.

Figure: The distribution of final unified discriminant to identify electrons. The solid histogram is for electrons in $e^{+}e^{-} \rightarrow e^{+}e^{-}e^{+}e^{-}$ events and the dashed one for charged pions.

Figure: Electron identification efficiency (circles) and fake rate for charged pions (squares). Note the different scales for the efficiency and fake rate.

The efficiency and fake rate are displayed in Fig.  using electrons in real $e^{+}e^{-} \rightarrow e^{+}e^{-}e^{+}e^{-}$ events for the efficiency measurement, and $K_{S} \rightarrow \pi^{+}\pi^{-}$ decays in real data for the fake rate evaluation. For momentum greater than 1 GeV/$c$, the electron identification efficiency is maintained to be above 90% while the fake rate to be around 0.2 to 0.3%.