FDTD is a fundamentally coherent simulation method. Below, we discuss three types of incoherence that are relevant for FDTD simulations, and the difficulties in direct simulation of these effects. In many physical processes, we have a combination of all types of incoherence.
The following sections provide three examples that show how to model various incoherent effects with FDTD and MODE' propagator.

The phase, φ, of the light shifts randomly over time, on a time scale τc. For light near visible wavelengths, we have
It is not practical to simulate the statistics of temporal phase incoherence because τc >> T.
A spatially incoherent source could be, for example, an ensemble of dipole emitters spread over a given volume. Each dipole is independent of its neighbors, and emits radiation with a random phase that varies on a time scale on the order of τc. At any given instant, the relative phases of all the dipoles are fixed, however, on a time scale of τc the relative phases of the dipoles change in a completely random fashion. The direct simulation of spatial incoherence requires simulations for very long periods of time or a large amount of ensemble averaging with one FDTD simulation per ensemble. Therefore it is not practical to simulate spatial incoherence directly in most cases.
In a similar manner to temporal phase incoherence, the polarization of a beam or the orientation of a dipole source depends on time. In the case of an unpolarized beam we have
where the unit vector u(t) defines the beam polarization and varies on a time scale τc >> T. In the case of a dipole source, we have
where the unit vector u(t) defines the dipole source polarization and varies on a time scale τc >> T.
In both cases, the time scale for the variation of the polarization is much larger then the optical cycle, making it unpractical to simulate the statistics of temporal polarization incoherence with FDTD.
Direct simulation of incoherent effects requires simulations that last many times the correlation time (τc) which is typically many orders of magnitude larger than the optical cycle. In some cases, ensemble averaging can be used but this requires a very large number of simulations. Therefore to simulate the physical conditions required for incoherence means simulation times that are far too long to be practical in most cases.
In general, a more efficient approach is to calculate the statistical averaging analytically rather than performing lengthy or numerous simulations. In general, this requires a small number of simulations and some postprocessing of the data. In the following sections, we explain how to perform this postprocessing in a number of examples that can be applied to most simulations of incoherent processes.
See the following pages for more information.