# Separation of Variables

1. Feb 16, 2014

### vikasagartha

I am studying Laplace's equation in my electrodynamics course (using griffiths intro to electrodynamics). I am watching a youtube video stepping through the separation of variables method for solving the PDE. It seems to be a common PDE that comes up repeatedly in physics (Helmholtz eqn, Poisson eqn) and I thought it would be worthwhile to understand where the solution comes from.

In the text Griffiths also touches on two additional points:
- "A solution to Laplace's equation has the property that the average value over a spherical surface is equal to the value at the center of the sphere"
- A uniqueness theorem

A first question that came to my mind: why do we assume that separation of variables works? How can we assume that the solution is of the form $f(x)g(y)$? What if it had an expression like $x^y$?

2. Feb 16, 2014

### SteamKing

Staff Emeritus
The functions chosen as solutions using separation of variables must be shown to be solutions of the particular PDE and to satisfy whatever boundary conditions apply. These requirements eliminate many elementary functions from consideration, because they cannot satisfy one or both of these requirements.

3. Feb 16, 2014

### vikasagartha

Okay, a second question: I watched Chris Tisdell's (youtube) derivation of the solution in two dimensions, and it involves a summation of a sine function (a fourier series of sorts). In electrostatics, the solution of the equation generates a potential function V, which we get by taking a path integral $\int \vec{E}\cdot\vec{dl}$. Ive done many such problems and never see any sinusoidal functions in there...why is that? Does the solution look different in 3D or am I misunderstanding the relation between the derivation of the solution and its application to electrostatics?

4. Feb 16, 2014

### dauto

What you have to understand is that the separation method gives you a complete set of solutions (there are theorems that prove that for a fairly large group of PDE problems called Sturm–Liouville equations.) What that means is that even though there are solutions that cannot be obtained by the separation of variables method, these solutions can be written as an expansion in terms of the solutions obtained by the separation of variables method. This expansion is a generalization of the Fourier expansion. So, even though the final solution may not be sinusoidal at all, it can still be expanded in terms of sine functions because the set of sine functions form a complete set according with Sturm–Liouville theory.