# Poisson equation with a dirac delta source.

1. May 23, 2015

### Coffee_

Consider:

$\nabla^{2} V(\vec{r})= \delta(\vec{r})$

By taking the Fourier transform, the differential equation dissapears. Then by transforming that expression back I find something like $V(r) \sim \frac{1}{r}$.

I seem to have lost the homogeneous solutions in this process. Where does this happen?

2. May 23, 2015

### Orodruin

Staff Emeritus
Taking the Fourier transform you are essentially saying that your domain is infinite and that your function ($V$) is square integrable. The homogeneous solution is not square integrable.

3. May 23, 2015

### Coffee_

Thanks. Maybe you could get this one as well since you already bothered to answer so I don't have to make another post:

What's the difference in the evolution of $f$ for $t>0$ between saying

1) $\partial_{x}^{2} f - \frac{1}{c^{2}} \partial_{t}^{2}f = -\delta(x) \delta(t)$ where nothing is happening before $t=0$

2) $\partial_{x}^{2} f - \frac{1}{c^{2}} \partial_{t}^{2}f = 0$ and to impose a delta peak as initial condition?

4. May 23, 2015

### Orodruin

Staff Emeritus
Essentially nothing, but depending a bit on which initial condition you put a delta in (the wave equation is second order in t and so needs two initial conditions).

5. May 23, 2015

### Coffee_

I was talking about the condition on $f$, the derivative condition is zero everywhere.

Is it correct to say that the first expression is more general. It can encompass the state of the system at $t<0$. For example, if an oscillatory motion already exists before this delta peak, this expression can still be in agreement with it.

The second one is only a description for $t>0$ of a special case of the above description (namely, when everything is zero before the delta peak).

6. May 23, 2015

### Orodruin

Staff Emeritus
Well, if you have initial conditions, you are encoding anything that happened before that time in those conditions. So for solving it for future $t$ it does not really matter. But the domain of the other problem is larger, yes.

7. May 29, 2015

### vanhees71

Well, it's all correct with your calculation of the Green function of the Laplace operator. The solution is not square integrable, because it's a distribution and not a function since the source is already a distribution. The Green's function in fact is
$$G(\vec{x})=-\frac{1}{4\pi |\vec{x}|} \; \Leftrightarrow \; \Delta G(\vec{x})=\delta^{(3)}(\vec{x}).$$