2
$\begingroup$

I am looking for the Green's function necessary to solve the following PDE: $$ \nabla^2\Phi-\epsilon M\Phi=M, $$ where $\Phi,M$ are scalar fields in 3D, $\nabla^2$ is the 3D Laplace operator and $\epsilon>0$ is a positive parameter which, if necessary, can be taken to be small. Boundary conditions can be specified to be that $\nabla\Phi$ vanishes at infinity.

What confuses me is that the source is coupled to the field itself, which makes it not clear at all on how I should go about finding a Green's function.

Many thanks.

EDIT: $\Phi$ and $M$ could be taken to be spherically symmetric if necessary

$\endgroup$
3
  • $\begingroup$ Try to have a look at Helmholtz equation $\endgroup$
    – basics
    Commented May 30, 2023 at 9:27
  • $\begingroup$ @basics The problem is that Helmholtz equation has M constant, but here it is a spatially dependent field. $\endgroup$
    – space-guy
    Commented May 30, 2023 at 9:31
  • $\begingroup$ I doubt that it is possible to find a non-singular Green's function for such a equation. I would try to search for it as follows: let $M(\vec{r}) = \delta(\vec{r} - \vec{r}_0)$. Then, the equation reduces to $\nabla^2 \Phi = (1 + \epsilon\Phi(\vec{r}_0))\delta(\vec{r} - \vec{r}_0)$, which is a Laplace equation. We get $\Phi(\vec{r}) = -(1 + \epsilon\Phi(\vec{r}_0))/(4\pi|\vec{r} -\vec{r}_0|)$, which yields a singular value of $\Phi(\vec{r}_0)$. $\endgroup$
    – E. Anikin
    Commented May 30, 2023 at 11:57

1 Answer 1

2
$\begingroup$

You have to view Green's functions as inverse operators. In your case, you want $\Phi$, so the operator you wish to invert is the RHS: $$ \Delta-\epsilon M $$ so the Green's function you are looking for is: $$ (\Delta_x-\epsilon M(x))G(x,y)=\delta(x-y) $$ Note that since $M$ breaks the translation invariance, $G$ takes two spatial arguments, and does not depend only on the difference. $\Phi$ can therefore be expressed as: $$ \Phi(x) = \int d^Dy G(x,y)j(y) $$ so that: $$ (\Delta-\epsilon M)\Phi = j $$ in particular, you can take $j=M$.

Since you are assuming that $\epsilon$ is a small parameter, you can write $G$ as a perturbative series in $\epsilon$. This is the Born approximation if you truncate at leading order and interpret $M$ as a potential energy and $\Phi$ a first quantisation wave function.

Formally, using the unperturbed case: $$ \begin{align} G_0 &= \Delta^{-1} \\ &= -\frac{1}{4\pi |x-y|} \end{align} $$ you would write using linear algebra: $$ \begin{align} G &= \frac{1}{\Delta-\epsilon M} \\ &= \frac{1}{\Delta}\frac{1}{1-\epsilon \Delta^{-1}M} \\ &= \frac{1}{1-\epsilon \Delta^{-1}M}\frac{1}{\Delta} \\ &= (1+\epsilon \Delta^{-1}M)\frac{1}{\Delta} \\ &= G_0 +\epsilon G_0 MG_0 \\ G(x,y) &= G_0(x,y)+\epsilon\int d^D z G_0(x,z)M(z)G_0(z,y) \\ &= -\frac{1}{4\pi |x-y|}+\frac{\epsilon}{(4\pi)^2}\int d^D z \frac{M(z)}{|x-z||y-z|} \end{align} $$ with the associate diagrammatic representation.

Hope this helps.

$\endgroup$
6
  • $\begingroup$ This helps a lot. Many thanks. What if M did not break the translational invariance and we wouldn't like to take $\epsilon$ small? Do you think there could be a green function still? $\endgroup$
    – space-guy
    Commented May 30, 2023 at 15:13
  • 1
    $\begingroup$ If $M$ is translational invariant, then you get Helmholtz’ equation and you even have a closed form expression. $\endgroup$
    – LPZ
    Commented May 30, 2023 at 16:03
  • 1
    $\begingroup$ If you are not interested in the limit $\epsilon\to0$, then perturbative expansion isn’t valid and you’ll need to exploit other specificities to calculate the Green’s function (large $N$ expansion…) $\endgroup$
    – LPZ
    Commented May 30, 2023 at 16:05
  • 2
    $\begingroup$ In general, the mathematical existence of the Green’s function is not so much an issue. It is rather its computability. $\endgroup$
    – LPZ
    Commented May 30, 2023 at 16:05
  • $\begingroup$ Many many thanks. Last question. I would have Helmholtz with a different sign (according to en.wikipedia.org/wiki/Green%27s_function ) it seems that is more like the Yukawa/Feynman Green's function with $k=\sqrt{\lambda M(r)}$ And yet this seems weird because M is not constant. $\endgroup$
    – space-guy
    Commented May 30, 2023 at 17:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.