# Filling in some steps in Jackson

1. Sep 4, 2009

### Old Guy

1. The problem statement, all variables and given/known data
If you have a copy of Jackson handy, this is in the middle of Section 1.7 (p.35 in my book). I'm trying to understand how he gets from Equation 1.30 to the next step (which is an unnumbered equation). Specifically, how is the term in square brackets a Taylor expansion of $$\rho(\textbf{x'})$$? For instance, I would expect to see a term with $$\nabla$$ before the $$\nabla^{2}$$ term, and shouldn't r$$^{2}$$ be r$$^{3}$$?

2. Relevant equations
They're long, it would take me forever to enter them. If you don't have a copy of Jackson handy, feel free to just skip this.

3. The attempt at a solution

2. Sep 5, 2009

### gabbagabbahey

If you define the vector $\textbf{r}=\textbf{x}-\textbf{x}'$ (this is the separation vector that Griffiths calls script-r), then expanding $\rho(\textbf{x}')$ about the point $\textbf{x}=\textbf{x}'$ is the same as expanding $\rho(\textbf{x}-\textbf{r})$ about the point $\textbf{r}=0$.

The 3D-Taylor expansion of a function is given by:

$$f(\textbf{x}+\textbf{c})=\sum_{j=0}^{\infty}\left[\frac{1}{j!}\left(\textbf{c}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)^jf(\overline{\textbf{x}})\right]_{\overline{\textbf{x}}=\textbf{x}}=f(\textbf{x})+\left(\textbf{c}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)f(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}+\frac{1}{2}\left(\textbf{c}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)^2 f(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}+\ldots$$

Where $\mathbf{\nabla}_{\overline{\textbf{x}}}$ means that the differentiation is done with respect to the dummy coordinates $\overline{\textbf{x}}$.

So, to second order in $\textbf{r}$, you have:

$$\rho(\textbf{x}')=\rho(\textbf{x}-\textbf{r})\approx \rho(\textbf{x})-\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}+\frac{1}{2}\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)^2 \rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}$$

I suspect that if you do the differentiation in spherical coordinates (I'd recommend using Mathematica or Matlab for this, unless you are a masochist) and then do the angular integrals (over a spherical surface), the only surviving terms will be

$$\int_{0}^{\pi} \int_{0}^{2\pi}\rho(\textbf{x}')\sin\theta d\theta d\phi\approx4\pi\rho(\textbf{x})+\frac{2\pi r^2}{3}\nabla_{\overline{\textbf{x}}}^2\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}$$

Last edited: Sep 5, 2009
3. Sep 5, 2009

### Old Guy

Thank you, G; I'll play around with this and report back.

4. Sep 5, 2009

### gabbagabbahey

I'm not sure why Jackson uses such a complicated method to show that the integral definition of $\Phi$ satisfies Poisson's equation, the well known identity $\nabla^2 \left(\frac{1}{\textbf{r}}\right)=-4\pi\delta^3(\textbf{r})$ will yield the same result without any complicated Taylor expansions.

5. Sep 5, 2009

### Ben Niehoff

The reason is that Jackson is trying to prove that identity in the first place.

6. Sep 12, 2009

### Bill Foster

I'm studying Jackson as well.

How does he go from Eq. (1.30) to the next equation in the book:

$$\nabla^2 \Phi_a\left({\textbf{x}}\right)=-\frac{1}{\epsilon_0}\int_{0}^{R} \frac{3a^2}{\left(r^2+a^2\right)^{\frac{5}{2}}} \left[\rho\left(\textbf{x}\right)+\frac{r^2}{6}\nabla^2\rho+...\right]r^2dr+O\left(a^2\right)$$

7. Sep 12, 2009

### Bill Foster

I know there's a Taylor expansion. But I just can't get it worked out. Even our professor gave up on it.

8. Sep 12, 2009

### gabbagabbahey

9. Sep 13, 2009

### Bill Foster

I'm not sure how that works out. When I do it, I get the following:

$$\frac{1}{4\pi\epsilon_0}\int\rho\left(\textbf{x}'\right)\left(\frac{3a^2}{\left(r^2+a^2\right)^{\frac{5}{2}}}\right)d^3x'$$

$$=\frac{1}{4\pi\epsilon_0}\int_0^\pi\int_0^{2\pi}\int_0^R\rho\left(\textbf{x}'\right)\left(\frac{3a^2}{\left(r^2+a^2\right)^{\frac{5}{2}}}\right)r^2sin\theta dr d\phi d\theta$$

$$=\frac{1}{4\pi\epsilon_0}\int_0^\pi\int_0^{2\pi}\int_0^R\left(\rho\left(\textbf{x}\right)-\left(\textbf{r}\cdot\nabla\right)\rho\left(\textbf{x}\right)+\frac{1}{2}\left(\textbf{r}\cdot\nabla\right)^2\rho\left(\textbf{x}\right)+\cdot\cdot\cdot\right)\left(\frac{3a^2}{\left(r^2+a^2\right)^{\frac{5}{2}}}\right)r^2sin\theta dr d\phi d\theta$$

$$=-\frac{1}{\epsilon_0}\int_0^R\left(\rho\left(\textbf{x}\right)-\left(\textbf{r}\cdot\nabla\right)\rho\left(\textbf{x}\right)+\frac{1}{2}\left(\textbf{r}\cdot\nabla\right)^2\rho\left(\textbf{x}\right)+\cdot\cdot\cdot\right)\left(\frac{3a^2}{\left(r^2+a^2\right)^{\frac{5}{2}}}\right)r^2 dr$$

I don't see how this term drops out:

$$-\left(\textbf{r}\cdot\nabla\right)\rho\left(\textbf{x}\right)$$

And I don't see how this term:

$$\frac{1}{2}\left(\textbf{r}\cdot\nabla\right)^2\rho\left(\textbf{x}\right)$$

becomes:

$$\frac{r^2}{6}\nabla^2\rho\left(\textbf{x}\right)$$

10. Sep 13, 2009

### gabbagabbahey

Compute those derivatives in spherical coordinates and then do the angular integration. Remember, a term doesn't have to be zero to integrate to zero over some specific interval.

11. Sep 14, 2009

### Bill Foster

$$\nabla=\hat{\textbf{r}}\frac{\partial}{\partial{r}}+\hat{\theta}\frac{1}{r}\frac{\partial}{\partial{\theta}}+\hat{\phi}\frac{1}{r\sin{\theta}}\frac{\partial}{\partial{\phi}}$$

$$\textbf{r}=r\hat{\textbf{r}}+\theta\hat{\theta}+\phi\hat{\phi}$$

$$\rho\left(\textbf{x}\right)=\rho\left(r,\theta,\phi\right)$$

Now taking into consideration just the middle term (the one that should cancel out), and ignoring the constants:

$$\int_0^{2\pi}\int_0^\pi \left(\textbf{r}\cdot\nabla\right)\rho\left(\textbf{x}\right)r^2\sin\theta d\theta d\phi$$

Making the substitutions:

$$\int_0^{2\pi}\int_0^\pi \left(\left(r\hat{\textbf{r}}+\theta\hat{\theta}+\phi\hat{\phi}\right)\cdot\left(\hat{\textbf{r}}\frac{\partial}{\partial{r}}+\hat{\theta}\frac{1}{r}\frac{\partial}{\partial{\theta}}+\hat{\phi}\frac{1}{r\sin{\theta}}\frac{\partial}{\partial{\phi}}\right)\right)\rho\left(\textbf{x}\right)r^2\sin\theta d\theta d\phi$$

Simplifying:

$$\int_0^{2\pi}\int_0^\pi \left(r\frac{\partial}{\partial{r}}+\theta\frac{1}{r}\frac{\partial}{\partial{\theta}}+\phi\frac{1}{r\sin{\theta}}\frac{\partial}{\partial{\phi}}\right)\right)\rho\left(\textbf{x}\right)r^2\sin\theta d\theta d\phi$$

$$\int_0^{2\pi}\int_0^\pi \left(r\frac{\partial\rho\left(\textbf{x}\right)}{\partial{r}}+\theta\frac{1}{r}\frac{\partial\rho\left(\textbf{x}\right)}{\partial{\theta}}+\phi\frac{1}{r\sin{\theta}}\frac{\partial\rho\left(\textbf{x}\right)}{\partial{\phi}}\right)\right)r^2\sin\theta d\theta d\phi$$

Break that integral up:

$$\int_0^{2\pi}\int_0^\pi \left(r\frac{\partial\rho\left(\textbf{x}\right)}{\partial{r}}\right)r^2\sin\theta d\theta d\phi+\int_0^{2\pi}\int_0^\pi \left(\theta\frac{1}{r}\frac{\partial\rho\left(\textbf{x}\right)}{\partial{\theta}}\right)r^2\sin\theta d\theta d\phi+\int_0^{2\pi}\int_0^\pi \left(\phi\frac{1}{r\sin{\theta}}\frac{\partial\rho\left(\textbf{x}\right)}{\partial{\phi}}\right)\right)r^2\sin\theta d\theta d\phi$$

Now I seem to be at an impasse because I don't know what the derivative are:

$$\frac{\partial\rho\left(\textbf{x}\right)}{\partial{r}}=?$$
$$\frac{\partial\rho\left(\textbf{x}\right)}{\partial{\theta}}=?$$
$$\frac{\partial\rho\left(\textbf{x}\right)}{\partial{\phi}}=?$$

I can assume the charge density is constant, which would give me zero. But that would also give zero for the third term, too, wouldn't it?

12. Sep 14, 2009

### gabbagabbahey

There are several problems here.

First, it is very important to distinguish $\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}$ and $$\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)^2\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}$$from $\left(\textbf{r}\cdot\nabla\right)\rho\left(\textbf{x}\right)$ and $$\left(\textbf{r}\cdot\nabla\right)^2\rho\left(\textbf{x}\right)$$. In the first two, differentiation is done with respect to the dummy variable $\overline{\textbf{x}}$, and so $\textbf{r}$ is treated as a constant vector. In the latter two, $\textbf{r}$ is not treated as a constant vector, and so your results will be quite different.

Second, by defintion, $\textbf{r}=r\mathbf{\hat{r}}\neq r\mathbf{\hat{r}}+\theta\mathbf{\hat{\theta}}+\phi\mathbf{\hat{\phi}}$...the units on your expression don't even make sense... $r$ has unit of distance, while $\theta$ and $\phi$ are angles, with units of radians.

Third, unit vectors in curvilinear coordinates are position dependent. So, for example $\textbf{r}\cdot\textbf{r}'=\left(r\mathbf{\hat{r}}\right)\cdot \left(r'\mathbf{\hat{r}'}\right)=r r' (\mathbf{\hat{r}}\cdot\mathbf{\hat{r}}') \neq rr'$ in general.

Finally, you only require that

$$\int_{0}^{\pi} \int_{0}^{2\pi}\left[-\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}+\frac{1}{2}\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)^2 \rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}\right]\sin\theta d\theta d\phi=\frac{2\pi r^2}{3}\nabla_{\overline{\textbf{x}}}^2\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}$$

This does not necessarily mean that

$$\int_{0}^{\pi} \int_{0}^{2\pi}\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}\sin\theta d\theta d\phi=0$$

and

$$\int_{0}^{\pi} \int_{0}^{2\pi}\frac{1}{2}\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)^2 \rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}\right]\sin\theta d\theta d\phi=\frac{2\pi r^2}{3}\nabla_{\overline{\textbf{x}}}^2\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}$$

13. Sep 16, 2009

### Bill Foster

The integration variables are $$\theta$$ and $$\phi$$. Where do they show up in the integrand?

If nowhere, then that whole integrand is just multiplied by $$4\pi$$ as a result of the integration. And that, evidently, isn't the case.

14. Sep 16, 2009

### gabbagabbahey

The $$\theta$$ and $$\phi$$ dependence is hidden in $\textbf{r}$ and hence in

$$\left[-\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)\rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}+\frac{1}{2}\left(\textbf{r}\cdot\mathbf{\nabla}_{\overline{\textbf{x}}}\right)^2 \rho(\overline{\textbf{x}})|_{\overline{\textbf{x}}=\textbf{x}}\right]$$

as-well, since $$\textbf{r}=r\mathbf{\hat{r}}=r\sin\theta\cos\phi\mathbf{\hat{i}}+r\sin\theta\sin\phi\mathbf{\hat{j}}+r\cos\theta\mathbf{\hat{k}}$$

15. Sep 16, 2009

### CFDFEAGURU

Is this in the Blue cover or Red cover version of Jackson's book?

Thanks
Matt

16. Sep 16, 2009

### Bill Foster

OK, now maybe we're getting somewhere. I'll have to work on this later. I have real homework to do.

One more thing - what is the dummy variable? What coordinates is it in? I'm assuming Cartesian.

17. Sep 16, 2009

### Bill Foster

Blue

18. Sep 16, 2009

### gabbagabbahey

It is in whatever coordinate system you want it to be in...

In spherical coordinates,

$$\overline{\textbf{x}}=\overline{r}\mathbf{\hat{\overline{r}}}=\overline{r}\sin\overline{\theta}\cos\overline{\phi}\mathbf{\hat{i}}+\overline{r}\sin\overline{\theta}\sin\overline{\phi}\mathbf{\hat{j}} +\overline{r}\cos\overline{\theta}\mathbf{\hat{k}}$$

In Cartesian coordinates,

$$\overline{\textbf{x}}=\overline{x}\mathbf{\hat{i}}+\overline{y}\mathbf{\hat{j}} +\overline{z}\mathbf{\hat{k}}$$

19. Sep 18, 2009

### Bill Foster

I would say those are both in Cartesian coordinates since the basis vectors are in Cartesian coordinates.

20. Sep 26, 2009

### Bill Foster

I'll come back to equation 1.30 later, when I get time (if I ever get time).

Here's another one from Jackson that has me a little confused: An unnumbered equation after (1.53):

$$W=-\frac{\epsilon_0}{2}\int{\Phi\nabla^2\Phi d^3x}$$

Jackson says "Integration by parts leads to the result:

$$W=\frac{\epsilon_0}{2}\int{|\nabla\Phi|^2d^3x}=\frac{\epsilon_0}{2}\int{|\textbf{E}|^2d^3x}$$ (1.54)

Now, performing integration by parts, I use the following:

$$\int{udv}=uv-\int{vdu}$$

$$u=\Phi$$
$$du=\nabla\Phi d^3x$$
$$dv=\nabla^2\Phi d^3x$$
$$v=\nabla\Phi$$

$$W=-\frac{\epsilon_0}{2}\int{\Phi\nabla^2\Phi d^3x}=-\frac{\epsilon_0}{2}\left(\Phi\nabla\Phi-\int{\left(\nabla\Phi\right)^2d^3 x}\right)$$

What happened to $$\Phi\nabla\Phi$$ ?