vanish. The normal component of B on the outer boundary is also assumed fixed.2.3 Boundary conditions
The boundary fields specify the tangential components of the electric field and the normal component of the magnetic field. The radial component of the magnetic field at the Earth’s surface is assumed fixed to its dipole value. Along the polar axes boundary, the normal component of the magnetic field is determined by the condition that the y-component of vanish. The normal component of B on the outer boundary is also assumed fixed.
The simulation is driven by the y-, dawn-dusk, component of the electric field. The run that is featured in this paper was driven by the constant Ey0 = 0.45, which is the same value that was used to initialize the plasma flow in the interior. Another run with Ey0 = 0.0 will also be discussed. This field was applied along the entire boundary, except at the Earth’s surface, where the y-component was taken to be zero. The other tangential component of the electric field was determined by the condition that the y-component of 
along the entire boundary, except at the Earth’s surface, where the latitudinal component was set to zero. These choices of the boundary electric field minimized the generation of spurious fields. One consequence of the choice of the Ey boundary condition is that all the flow is into the simulation domain with no outflow. In a three-dimensional model, plasma driven by convection toward the Earth could be diverted to flow eastward and westward around the Earth. In two dimensions there is no way plasma can flow around to the days side. As a result, plasma will pile up in a stagnation region at about four or five RE .
Considerable programming effort went into specifying the particle boundary conditions. The simplest boundary was at the Earth’s surface. Any particle that got there was simply removed. The next level of complexity was the outer boundary. After the particle boost and prior to the particle move, a layer of particles is initialized to the density and temperature of the adjacent region inside the boundary. The flow velocity is initialized to the drift consistent with the specified Ey0. The boundary subroutine then moves the particles one time step. Those particles that find themselves inside the simulation domain are then added to the active particle array. Following this procedure, the subroutine which moves the active particles is called. Particles which wander outside the simulation domain are removed from the active particle array.
The polar axes constitute the complicated boundary. This is because there is a strong magnetic field tangential to the boundary. There particles would want to gyrate around the field lines back and forth across the boundary. The boundary conditions described above would result in interruption of particle orbits and spurious boundary currents [Naitou et al., 1979]. To overcome this problem, buffer regions of about three gyroradii thick were established along the north and south polar boundaries. The positions and velocities of these particles were advanced assuming a constant convection electric field and a magnetic field that was extrapolated from just inside the boundary. These particles could therefore gyrate back and forth between the buffer and simulation domain. They did not, however, contribute to the field sources. The buffer domain boundaries opposite the simulation/buffer domain boundaries were supplied with new particles in the same way that the outer boundary was supplied.