In order to solve the PB equation with the FAS technique,
the charge density of Eq. 40 is inserted as part of
the driving term fk. In addition, a slight modification
of the defect correction
must be added to adjust
for the differences of the nonlinear terms on the two
scales. Once these changes are made, the solution proceeds
just as for the standard elliptic solver, and the same linear
scaling is maintained.[42]
In the calculations presented here,
the total positive and negative charges were maintained
constant in the calculation domain. It was found in numerical
experiments that the FAS solver obtained the solution
with the action converged
to a small fraction of kB T at room temperature within roughly
12 fine scale iterations. Using the same
relaxation weighting parameter
,
an SOR solver on the finest scale alone
required on the order of 103 iterations to obtain the same
level of convergence in the action,
both methods starting from an initial
guess of
everywhere.
A cut through the three dimensional potential
field is presented in Figure 3. These grid results compare well
with an exact numerical solution of the radial
PB Eq. for these conditions.
The results show that there is
promise that the PB fields can be updated quickly enough (roughly 10
lattice sweeps) that the PB solver can be incorporated in
simulations of large scale colloid or polyelectrolyte systems.