As already mentioned the convergence of the presented method seems
to be very slow. To improve it we need either to have a ``better''
initial guess, and/or to accelerate the convergence by applying some form of
relaxation [10]. The relaxation parameter 
(specified)
is used to modify the solution vector [x] at iteration i
as follows:
replaces 
when determining the saw-tooth correctors.
Figure 15 shows the resulting improvement in convergence
when 
(over-relaxation) is used.
Values of greater than 1.7
produced oscillating results. This topic requires further
attention.
Figure 15: The convergence rate with number of iterations
for the saw-tooth method and
for different values of relaxation coefficients .