This study was performed using Gradphi [2], a Monte Carlo simulation of the Radphi apparatus based upon the GEANT package. The simulation begins with an electron from the primary beam entering the bremsstrahlung radiator. After the radiator, the electron is removed and the simulation continues to follow the photon produced. Bremsstrahlung is generated between a lower cutoff of 1MeV and the end point, and has the proper angular distribution. The photon beam passes through the primary collimator, a pair of sweeping magnets and a secondary collimator before arriving at the CLAS target. Interactions are simulated whenever photons pass through matter and interact. Downstream of the CLAS target, the remaining photons pass through an exit window into a helium tube which passes through the forward aperture in CLAS and down to the alcove area. At the entrance to the alcove the helium tube passes through the lead ``doughnut'' that defines the beam aperture for this experiment, through the lead shielding wall and on to within a few cm of the target. There the photons remaining in the beam exit through a window into air and then pass into the target if they are within its radius. Downstream of the target, particles are tracked until they either enter a calorimeter block, reach the back wall of the alcove, or fall below the kinetic energy cutoff of 1MeV. Any secondaries generated by interactions or decays are also followed by the simulation. All of the interactions that GEANT knows about are simulated, including for photons electron pair production, Compton scattering and the photoelectric effect, and for electrons ionization energy loss, multiple scattering, bremsstrahlung and annihilation (for positrons).
Hadronic interactions for all particles are included using the GHEISHA package. The hadronic interactions are parameterizations of cross sections with a range of validity usually limited to high energies or averages over an ensemble of low-energy particles such as characterizes the response of a hadronic calorimeter. The electromagnetic part of the GEANT simulation is quite detailed, however, and generally reliable at energies above 1MeV down to the level of a few percent in its cross sections and angular distributions. Care was taken in the geometrical description of the experimental setup that is used by Gradphi to put in any elements that might affect the simulation at the percent level. The dimensions and distances between detector elements in the simulation are accurate to 5% or better. Based upon this and the observation that the detector rates are dominated by electromagnetic background, the systematic error for the simulation may be estimated to be at the level of 5%.