next up previous
Next: About this document

Chemical Engineering 535
Fall, 1996
B. Bernstein, Instructor
Notes 1: Thermal Conduction in a Cylinder

This is a discussion of the heat equation (diffusion equation) in a cylinder. Let us start by defining the Laplacian operator, tex2html_wrap_inline5957 : In cartesian coordinates, x,y,z, it is defined by

  equation5685

In cylindrical coordinates, tex2html_wrap_inline5961 , the Laplacian is

  equation5694

The heat equation is given by

  equation5709

where t is time and where tex2html_wrap_inline5965 here, the thermal diffusivity, corresponds to tex2html_wrap_inline5967 in our book. The problem we now address is as follows:

A cylinder r<a, 0<z<L, has initial temperature distribution tex2html_wrap_inline5973 at time t=0. Thenceforth, the lateral face r=a is kept at temperature T=0 and the ends, z=0 and z=L are thermally isolated: i.e.  tex2html_wrap_inline5985 is zero at both ends. Find the temperature tex2html_wrap_inline5987 at times t>0.

The method for attacking this problem will be separation of variables: We start by seeking a solution of (3) in the form

  equation5718

Substitution of equation 4 into equations (2) and (3) gives

  equation5724

Dividing both sides of (5) by tex2html_wrap_inline5991 , we get

  equation5734

Now, since the right hand side depends only on t, and the left hand side is independent of t, they must both be equal to a constant, say tex2html_wrap_inline5997 . We get, then

  equation5751

Consider, now, the right hand side of the first equation (7): Its right hand side depends only on z, whereas its left hand side is independent of z. Therefore, both sides must be constant, say tex2html_wrap_inline6003 . We get, after a little algebra,

  equation5765

Once more, the right hand side of equation (8) depends only on tex2html_wrap_inline6005 and the right hand side depends only on r, so that both sides are equal to the same constant, say tex2html_wrap_inline6009 . We get

  equation5775

Let us now summarize where we are, we are to solve

     eqnarray5778

We have added a couple of comments to equations (10) and (13): These comments should be clear, but they are needed.

There are several eigenvalue problems of Sturm-Liouville form which now arise. The first has to do with equation (10). We shall leave it to you to show that the constant tex2html_wrap_inline6009 must be zero or positive. Indeed, we get

  equation5789

We leave it to you, the reader, then, to establish that we must have tex2html_wrap_inline6013 , where tex2html_wrap_inline6015 . These values of tex2html_wrap_inline6009 are eigenvalues of what we call the periodic Sturm-Liouville problem and the eigenfunctions are as follows: For n=0 the eigenfunction is a constant, say tex2html_wrap_inline6021 . For positive n, the eigenfunctions are tex2html_wrap_inline6025 and tex2html_wrap_inline6027 , tex2html_wrap_inline6029 . Note that for each positive eigenvalue, the space of eigenfunctions is two dimensional ( is given by an arbitrary linear combination of two linearly independent functions, tex2html_wrap_inline6025 and tex2html_wrap_inline6027 ) unlike the situation where homogeneous end values are specified: When they are so specified, the space of eigenfunctions corresponding to each eigenvalue is one dimensional.

So, now, equation (13) becomes

  equation5795

Now equation (13) is really the statement of a Sturm-Liouville problem for R(r), except that it has two parameters. However, equation (15) has essentially fixed one of the parameters (or has, at least, incorporated the allowable values of one of the parameters), so we now have an eigenvalue problem for each value of n, where tex2html_wrap_inline6039 has to be an eigenvlue. Now equation (15) looks almost like Bessel's equation: Indeed, a simple transformation will transform it into Bessel's equation: Let tex2html_wrap_inline6041 , so that the chain rule gives

displaymath5949

so that equation (15) becomes

  equation5812

Those solutions of (16) which are continuous at tex2html_wrap_inline6043 are

  equation5820

The end condition R(a)=0 is met by requiring tex2html_wrap_inline6047 to be one of the zeroes of the Bessel function tex2html_wrap_inline6049 . Let these zeroes be denoted by tex2html_wrap_inline6051 : i.e.  tex2html_wrap_inline6053 , tex2html_wrap_inline6055 , tex2html_wrap_inline6057 . So the eigenvalues are tex2html_wrap_inline6059 and the eigenfunctions are

  equation5828

What remains is only to put these separation of variables solutions together with unknown coefficients and to evaluate the coefficients using the initial conditions. That it is possible to do this stems from the result in Sturm-Liouville theory that one may represent a fairly arbitrary function as a linear combination of eigenfunctions, a result which one uses in determining the coefficients from the initial conditions.

We wish, now, to carry through (with your help) an example of such a solution. But we shall scale the example down so as to illustrate the notions without a very excessive amount of work.

To begin with, let us assume that the initial condition is independent of z: tex2html_wrap_inline6063 . Then equation the boundary value problem stated in (11) will have the solution z=1, and we can forget about dependence of T on z.

So, our solution is, then

equation5835

Problems:

  1. The problem involving the Bessel functions is of the Sturm-Liouville type. What are its eigenfunctions and eigenvalues. Write down a formula for evaluationg tex2html_wrap_inline6073 , tex2html_wrap_inline6075 , tex2html_wrap_inline6077 .
  2. Consider the simplest case, namely where the initial value depends only on r and not on tex2html_wrap_inline6005 or z.

    displaymath5950

    In that case, assuming that a=1, one gets

    displaymath5951

    Given that g(r)=r, and that the first zeroes of tex2html_wrap_inline6089 are

    displaymath5952

    (rounded off, of course), evaluate tex2html_wrap_inline6073 for m=1,2,3,4,5.




next up previous
Next: About this document

che535
Wed Nov 20 08:09:51 CST 1996