Enhancement to "volume correction"

[GoogleCodeExporter]

Currently "volume correction" feature is available to slightly adjust the
dipole grid to better fit the particle shape. However, it is limited to
scaling the dipole model as it was generated assuming size of the dipole
grid equal to size of the particle. 

More detailed optimization can be performed if dipoles are added or removed
together with scaling. For instance, Draine (1988) claims the optimal (in a
certain sense) sphere model with grid=8 to contains 304 dipoles. But ADDA
produces the shape with 280 dipoles (when given '-grid 8'). Going from the
latter to the former requires first extra shrinking of dipole grid, so that
centers of remaining 16 dipoles get inside the sphere. After the addition
of these dipoles, volume equivalence can again be satisfied.

There are two possible ways of addressing this issue.
1) Allow input of non-integer grid, so that sizes of the computational grid
and of the particle are originally slightly different. Then user may play
with grid value to manually optimize the shape.

2) (harder to implement) Perform such optimization of non-integer grid
value automatically inside ADDA. This will require regenerating the shape
with different values of grid in some small interval. As a measure of
optimality (additionally to volume equivalence) the Draine's criterion may
be used (that is the ratio of gyration radius of the dipole grid to that of
the ideal shape with the same volume). This will also require analytical
expressions for gyration radii of predefined shapes in ADDA.

I do not know what algorithm is currently used in DDSCAT (worth investigating).

Finally, this is mostly relevant for small values of grid (e.g. 8) and
hence is not very interesting for practical applications. Hence, a low
priority is given to this issue.

Original issue reported on code.google.com by yurkin on 16 Apr 2010 at 9:03