I am trying to understand a derivation for finding the Green's function of Laplace's eq in cylindrical coordinates.

Let the Green's function be written as: $G(r,\theta,z,r',\theta',z') = G(\mathbf{r},\mathbf{r}')$

We know that in cylindrical coordinates


Using the cylindrical Laplacian we can then write:

$$\frac{1}{r}\frac{\partial}{\partial r}\left(r\frac{\partial G}{\partial r}\right)+\frac{1}{r^2}\frac{\partial^2 G}{\partial \theta^2} + \frac{\partial^2 G}{\partial z^2} = \frac{1}{r}\delta(r-r')\delta(\theta-\theta')\delta(z-z')$$

Using the identities:

$$\delta(\theta-\theta') = \frac{1}{2\pi}\sum^{\infty}_{m=-\infty}e^{im(\theta-\theta')}$$

$$\delta(z-z') = \frac{1}{2\pi}\int^{\infty}_{-\infty}e^{ik(z-z')}dk$$

We find that

$$G(\mathbf{r},\mathbf{r}') = \frac{1}{4\pi^2}\sum^{\infty}_{m=-\infty}\int^{\infty}_{-\infty}g_{m}(r,r')e^{ik(z-z')}e^{im(\theta-\theta')}dk$$

$$\frac{1}{r}\frac{\partial}{\partial r}\left(r\frac{\partial g_{m}}{\partial r}\right) - \left(k^2+\frac{m^2}{r^2} \right)g_{m} = \frac{1}{r}\delta(r-r')$$

I'm getting confused on the last step. It seems that the Green's function is assumed to be $G(r,\theta,z,r',\theta',z') = R(r)Q(\theta)Z(z)$ and this is plugged into the cylindrical Laplacian but I'm not seeing how to get $G(r,\theta,z,r',\theta',z')$ in the form shown or how to get that ODE in terms of $g_{m}$. Are there any ideas on how to do this intermediate step?

  • $\begingroup$ You little g_m is written as a function r and r', but it also depends need to depend on k as a continuous parameter. $\endgroup$
    – hft
    Mar 4, 2022 at 1:19

2 Answers 2


The expression

$$G(\mathbf{r},\mathbf{r}') = \frac{1}{4\pi^2}\sum^{\infty}_{m=-\infty}\int^{\infty}_{-\infty}g_{m}(r,r')e^{ik(z-z')}e^{im(\theta-\theta')}dk\tag{∗}$$

is nothing but a Fourier expansion of some function of $z$ and $\theta$, with $g_m(r,r')$ being the expansion coefficients. It's not something you should get: it's an ansatz.

Now, to get your ODE you can go the following way. Use your identities to express the whole RHS of your PDE (the product of deltas) in the form


Then insert $G(\mathbf r, \mathbf r')$ in the form $(*)$ into the PDE. Apply each differential operator termwise, and then move all the summands-integrands under the single sum-integral sign.

You now get the equation


where $f_\mathrm{ODE}(k,m,r)$ is the $\mathrm{LHS}-\mathrm{RHS}$ of your ODE.

The LHS of $(2)$ is a Fourier expansion of some function of $z-z'$ and $\theta-\theta'$, where $f_\mathrm{ODE}(k,m,r)$ denotes the Fourier coefficients. By uniqueness of Fourier expansion, the only way for the whole LHS of $(2)$ to be zero is for $f_\mathrm{ODE}$ to be zero for all $k$ and $m$. Since we also want this equation to hold for all $r$, we finally get

$$f_\mathrm{ODE}=0\;\forall\, k,m,r,\tag3$$

which is equivalen to your ODE.


To make the typing easier I will set $r'$, $\theta'$, and $z'$ to zero (wlog). So, now G is considered to be a function of just $r$, $\theta$, and $z$ (wlog).

Any[1] function of $r$, $\theta$, and $z$ can be expanded in a double fourier series like: $$ G(r,\theta, z) = \frac{1}{2\pi}\sum_m \frac{1}{2\pi}\int dk g_{mk}(r)e^{ikz}e^{im\theta}\;, $$ where $k$ is a continuous parameter and $m$ is a discrete parameter.

Plug this expansion into your equation for $G$ and find: $$ \frac{1}{2\pi}\sum_m \frac{1}{2\pi}\int dk e^{ikz}e^{im\theta} \left[\frac{1}{r}\frac{d}{dr}r\frac{dg_{mk}(r)}{dr} -\frac{n^2}{r^2}g_{mk}(r) - k^2 g_{mk}(r)\right] =\frac{1}{r}\delta(r)\frac{1}{2\pi}\sum_{m'} \frac{1}{2\pi}\int d{k'} e^{ik'z}e^{im'\theta} $$

Because the $e^{im\theta}$ and $e^{ikz}$ are a complete orthonormal set, each side of the equation holds term by term for all $m$ and $k$. So we can "drop the sums" over the dummy variable on each side and we find: $$ \left[\frac{1}{r}\frac{d}{dr}r\frac{dg_{mk}(r)}{dr} -\frac{m^2}{r^2}g_{mk}(r) - k^2 g_{mk}(r)\right] =\frac{1}{r}\delta(r) $$

[1] Some caveats may apply, but we don't care.


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.