Applying Fourier Series to Hot RingsIt is now time to pay our scholarly dues. We will discuss Fourier's initial application of Fouirer Series to the problem of heat transfer on a ring. As Prof. Osgood said, the hot-ring problem has become an important part of our intellectual heritage, and we will find it useful to go through this famous example. In Paris, there's lots of big heavy anchor rings that you can find in the city. As the story goes, Fourier stumbled across one of these rings one day, and wanted to figure out the dynamics of heat flow on one of these rings. I really have no idea behind whether this is true or not, but hopefully this story provides some (?) motivation for why hot rings are a problem in the first case. Anyways, Fourier considered the following problem: Fourier's problem
Suppose we heated up a ring to have a particular temperature distribution at different points along the ring. If we waited for the ring to cool down, how would this temperature distribution evolve over time? To translate this physical question into a mathematical statement, suppose we write down the initial temperature distribution as some function The reason why we use Fourier Series to solve this problem is that the function The heat equationThe time evolution of temperature on the ring is governed by a partial differential equation known as the heat equation. It turns out that the heat equation isn't terribly difficult to derive from simple physical considerations; if I have time later this quarter, I might come back to this page and work out the derivation here. But for our purposes, we can just take the heat equation for what it is: it tells us how quickly the temperature at any particular point changes over time, and relates this to the spatial variation in temperature near that point. Mathematically, the heat equation is written as ![]() where The form of the heat equation actually tells us qualitative behavior about how we expect solutions to behave, but for now we'll just move on to the solution. (Honestly, I might just come back and redo this problem as a physicist, so that we could interpret the units of various variables and use dimensions to help us out…) The heat equation is a partial differential equation (PDE), which isn't quite as simple to solve as an ordinary differential equation (ODE). In this problem, the main trick is to use Fourier Series to help us reframe the (PDE) in terms of an ODE. (Note that depending on your level of rigor, you might interpret differently how exactly Fourier Series help us here.) A Fourier Series AnsatzMotivated by the periodicity of ![]() where the Fourier coefficients Notice that I've put in time-dependence into the Fourier coefficients In a sense, we've reduced the problem to finding the set of functions Solving for the CoefficientsTo solve for the coefficients ![]() Let's consider what happens when these derivatives hit the ansatz. On the LHS, the only time-dependence is in the coefficients ![]() For this equality to hold, each term on the left must equal its corresponding term on the right. (There's a more subtle argument, but thankfully the naive approach works.) If we match the coefficients of the exponentials, we find that ![]() or ![]() which is an ordinary differential equation in ![]() which says that each of the Fourier components start of with some initial value Matching Initial ConditionsOur final task is to figure out the initial values of the coefficients Upon a bit of thought, we realize that the initial values of the coefficients So Putting it togetherPutting in our solution for the time-dependence of the coefficients ![]() where ![]() This final result is quite magical and beautiful: we have an explicit formula for the solution of the heat equation! There's plenty to be said about this solution, but I'm getting hungry so I'm going to eat some lunch now. Okay back from lunch now. Notice that the Fourier coefficients of the solution ![]() decay exponentially in time, and that the bigger the In particular, note that the A Taste of ConvolutionThe next day (Mon Week 2), we made a few more concluding remarks about the heat equation; in particular, the fact the the solution could be written as a convolution integral. We'll be discussing convolutions in their full glory later on, but it's a nice sneak peak into what's to come. If we expand the solution to the heat equation by plugging in the formula for the Fourier coefficients ![]() and by grouping together and combining all the terms that depend on ![]() Our solution ![]() where the function ![]() We'll discuss these concepts in much more detail later: it turns out that convolving two functions is a very general thing we can do, and the fact that solutions of the heat equation take the form of a convolution with the fundamental solution has deeper consequences and properties. But for now, we will take this as foreshadowing for later. A preview of delta ‘functions’Earlier, we figured out already what would happen if we took the What if we look at the other limit, as ![]() i.e., it approaches a Fourier Series where all the coefficients are 1. There is no way to make sense of this series ‘‘classically’’ in terms of functions that we know and love; however, the function does converge into another sense to a comb of |