You are on page 1of 2

Obviously, whatever i write below must be symmetrical in x and X (as long as we change

EVERYTHING to be wrt to the other variable if we choose to flip it about).

DEFINE the Greens function to be the solution of


in V, and G = 0 on S,
and I also DEFINE that the
acts wrt x, not . The definitions of V and S are not yet quantified.
Then when I find the Greens function for any problem (fundamental solution, the disk in the plane
etc) then I am solving with this in mind. So my solution for the disk allows
to differentiate wrt to x
but not wrt and so the solution if
differentiates wrt will NOT be the same. This is obvious in
the expression for the Greens function for the disk in the plane, as we have two image charges and
their contribution, and so cant just switch all of the xs and s without switching what we
differentiate with respect to. I know the Greens function is symmetrical in the two variables, but not
if we dont change the subscript in the equation above.
In particular, if I get an expression (say in polars) for G, then I will only get
for
differentiating wrt the correct variable. I know Im labouring the point, but it makes it clear for
me, and is worth keeping in mind.
So now apply Greens identity, using the definition above, that

acts wrt x:

where again S and V are not really quantified. We also see here another reason why our definition of
is important. It acts upon
in the surface integral, which would give 0 if it did not differentiate
wrt x. Of course, we could flip all of the variables no problem, and change the subscript on to
again ensure a non identically zero answer from the gradient, but we also have to change the
variables in the Greens function as mentioned above.
Now we can choose to integrate wrt to either or x (I think...is this true?) and get either
or
from the integral over the delta function that is
in the second term of the first
integrand.
Either way, it appears that remains acting upon x. This is due to the definition of the Greens
function I have found, which goes back to the point that if I had found the Greens function for the
disk the other way around so to speak, then I would have this expression instead. As in, I can easily
change the variables around, and in the case of the disk, end up with two image charges but
expressed with the other variable. I realise this is a slightly unclear paragraph, but I think it works!
Anyway, the integrals appear to work into the solution:

using G vanishing on the Surface, with the integrals varying , denoted by the subscripts

The only things to clear up now are what the Surface and Volumes are, which should be clear from
the question. In Q11, (gravity), im not sure why the surface should be the disk. As in, why must G be
0 on this disk. Should I not just say that G is not on the surface at infinity. I get why the volume
integral is only over the disk, because that is the only place (=
is defined. Actually, this
is the case, isnt it. In that question, we did indeed throw away the surface integral, so I think Im
happy again!
One final thing. I think the argument about how G and the equation defining it must change under
the switching of the variables works, but I know G should be symmetric in both variables. There is a
slight issue about whether these two ideas are consistent...if I do calculate the solution to
(so the other way around) i think Im clear as to how the form of G is
subtly different, but im not quite convinced that it is symmetrical in the sense I was thinking
before.
Thanks for looking and thinking about this in your free time. Please feel free to point out any holes in
the argument!
James

You might also like