College Math Teaching

August 19, 2011

Partial Differential Equations, Differential Equations and the Eigenvalue/Eigenfunction problem

Suppose we are trying to solve the following partial differential equation:
\frac{\partial \psi}{\partial t} = 3 \frac{\partial ^2 \phi}{\partial x^2} subject to boundary conditions:
\psi(0) = \psi(\pi) = 0, \psi(x,0) = x(x-\pi)

It turns out that we will be using techniques from ordinary differential equations and concepts from linear algebra; these might be confusing at first.

The first thing to note is that this differential equation (the so-called heat equation) is known to satisfy a “uniqueness property” in that if one obtains a solution that meets the boundary criteria, the solution is unique. Hence we can attempt to find a solution in any way we choose; if we find it, we don’t have to wonder if there is another one lurking out there.

So one technique that is often useful is to try: let \psi = XT where X is a function of x alone and T is a function of t alone. Then when we substitute into the partial differential equation we obtain:
XT^{\prime} = 3X^{\prime\prime}T which leads to \frac{T^{\prime}}{T} = 3\frac{X^{\prime\prime}}{X}

The next step is to note that the left hand side does NOT depend on x ; it is a function of t alone. The right hand side does not depend on t as it is a function of x alone. But the two sides are equal; hence neither side can depend on x or t ; they must be constant.

Hence we have \frac{T^{\prime}}{T} = 3\frac{X^{\prime\prime}}{X} = \lambda

So far, so good. But then you are told that \lambda is an eigenvalue. What is that about?

The thing to notice is that T^{\prime} - \lambda T = 0 and X^{\prime\prime} - \frac{\lambda}{3}X = 0
First, the equation in T can be written as D(T) = \lambda T with the operator D denoting the first derivative. Then the second can be written as D^2(X) = 3\lambda X where D^2 denotes the second derivative operator. Recall from linear algebra that these operators meet the requirements for a linear transformation if the vector space is the set of all functions that are “differentiable enough”. So what we are doing, in effect, are trying to find eigenvectors for these operators.

So in this sense, solving a homogeneous differential equation is really solving an eigenvector problem; often this is termed the “eigenfucntion” problem.

Note that the differential equations are not difficult to solve:
T = a exp(\lambda T) X  = b exp(\sqrt{\frac{\lambda}{3}} x) + cexp(-\sqrt{\frac{\lambda}{3}} x) ; the real valued form of the equation in x depends on whether \lambda is positive, zero or negative.

But the point is that we are merely solving a constant coefficient differential equation just as we did in our elementary differential equations course with one important difference: we don’t know what the constant (the eigenvalue) is.

Now if we turn to the boundary conditions on x we see that a solution of the form A e^{bx} + Be^{-bx} cannot meet the zero at the boundaries conditions; we can rule out the \lambda = 0 condition as well.
Hence we know that \lambda is negative and we get X = a cos(\sqrt{\frac{\lambda}{3}} x) + b sin(\sqrt{\frac{\lambda}{3}} x) solution and then T = d e^{\lambda t } solution.

But now we notice that these solutions have a \lambda in them; this is what makes these ordinary differential equations into an “eigenvalue/eigenfucntion” problem.

So what values of \lambda will work? We know it is negative so we say \lambda = -w^2 If we look at the end conditions and note that T is never zero, we see that the cosine term must vanish (a = 0 ) and we can ensure that \sqrt{\frac{w}{3}}\pi = k \pi which implies that w = 3k^2 So we get a whole host of functions: \psi_k = a_k e^{-3k^2 t}sin(kx) .

Now we still need to meet the last condition (set at t = 0 ) and that is where Fourier analysis comes in. Because the equation was linear, we can add the solutions and get another solution; hence the X term is just obtained by taking the Fourier expansion for the function x(x-\pi) in terms of sines.

The coefficients are b_k = \frac{1}{\pi} \int^{\pi}_{-\pi} (x)(x-\pi) sin(kx) dx and the solution is:
\psi(x,t) =   \sum_{k=1}^{\infty}  e^{-3k^2 t} b_k sin(kx)

Advertisements

1 Comment »

  1. […] close with a link on how these eigenfunctions and eigenvalues are calculated (in the context of solving a partial…. Like this:LikeBe the first to like this post. Leave a […]

    Pingback by Eigenvalues, Eigenvectors, Eigenfunctions and all that…. « College Math Teaching — May 26, 2012 @ 10:35 pm


RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Create a free website or blog at WordPress.com.

%d bloggers like this: