added:

Now I have got an idea a repulsive, force that first get greater with distanc ebefore becoming weaker.

for example.

G*m1*m2 *(r/20)^2/(r/20)^3

That's an interesting idea. Maybe some sort of way to describe forces in terms of an equation. Custom forces could be really neat.

A computational note: r^3 will slow the program down dramatically. r^2 is easy to calculate sicne you don't take a sqr root. That is, you can just do r^2 = a^2 + b^2 (pythagorous theorem).

r^3 on the other hand is sort of a mess.

Negative G values work jsut as well, but I'm not sure they'd be terribly interesting or not. I mean, eventually you'd get a kind of uniform distriubtion.