Discretization methods for one-dimensional Fokker-Planck operators (Q1067761)

Two numerical methods of solving the Fokker-Planck kinetic equation are introduced. These use a rewriting of the equation in the form \[ (\partial u/\partial t)(x,t)=(1/A(x))(\partial /dx)\{D(x,t,u)(\partial /\partial x)E(x,t,u)u\} \] together with derivative boundary conditions at \(x=0\) and \(x=\infty\). The functions D and E are positive and sufficiently smooth functions of x,t and u. The first method is a simple difference discretization of the equations which is semi-implicit. The second method assumes certain forms for D and E before the discretization of the equations. If D and E are only functions of x then it is shown that the methods have first order convergence in time and second order convergence in space. The methods are applied to the non-linear problem of Compton and inverse Compton scattering and the results compared with the Chang-Cooper method.