Design of the next generation of ion engines can benefit from detailed computer simulations of the plasma in the discharge chamber. In this work a complete particle based approach has been taken to model the discharge chamber plasma.This is the first time that simplifying continuum assumptions on the particle motion have not been made in a discharge chamber model. Because of the long mean free paths of the particles in the discharge chamber continuum models are questionable. The PIC-MCC model developed in this work tracks following particles: neutrals, singly charged ions, doubly charged ions, secondary electrons, and primary electrons. The trajectories of these particles are determined using the Newton-Lorentz's equation of motion including the effects of magnetic and electric fields. Particle collisions are determined using an MCC statistical technique. A large number of collision processes and particle wall interactions are included in the model. The magnetic fields produced by the permanent magnets are determined using Maxwell's equations. The electric fields are determined using an approximate input electric field coupled with a dynamic determination of the electric fields caused by the charged particles. In this work inclusion of the dynamic electric field calculation is made possible by using an inflated plasma permittivity value in the Poisson solver. This allows dynamic electric field calculation with minimal computational requirements in terms of both computer memory and run time. In addition, a number of other numerical procedures such as parallel processing have been implemented to shorten the computational time. The primary results are those modeling the discharge chamber of NASA's NSTAR ion engine at its full operating power. Convergence of numerical results such as total number of particles inside the discharge chamber, average energy of the plasma particles, discharge current, beam current and beam efficiency are obtained. Steady state results for the particle number density distributions and particle loss rates to the walls are presented. Comparisons of numerical results with experimental measurements such as currents and the particle number density distributions are made. Results from a parametric study and from an alternative magnetic field design are also given.