Diffusion coefficient in a finite domain from exit times and application to the tokamak magnetic structure

The standard definition of the diffusion coefficient D uses the limit of the mean square excursion; in a finite (bounded) domain this definition is not relevant, but most physical problems deal with bounded systems. A generalisation of D, useful to describe transport in a finite region which is a subset of a bounded system can be obtained from the exit times τi of test particles. We give an analytical relation between D and the τi for diffusive processes. We apply this method to compute an averaged diffusion coefficient for collisional particles in a domain between overlapping islands as they exist in a tokamak magnetic configuration, and discuss whether this process is of a purely diffusive nature. (orig.)