In problem 7.26, show that if we define a new unknown by setting
the PDE for simplifies into the normal heat equation
for suitable values of the constants and .
Find those values. Show that the initial condition for is the same
as the one for , but that the boundary condition becomes
where is a different function than
. Identify what is.
Solve the problem for using Laplace transformation in time.
Write the solution in terms of and the given function ,
of course. Note that the answer in the book is wrong.
Now solve 7.26 directly, by Laplace transforming the problem as
given in time. This is a good way to practice back transform
methods. Note that one factor in is a simpler function
at a shifted value of coordinate .
Solve 7.35 by Laplace transform in time. Clean up completely;
only the given function may be in your answer, no Heaviside
functions or other weird stuff. There is a minor error in the
book’s answer.
Replot the solution to 7.27, summing only one, two, and three
eigenfunctions and examine how the solution approaches the exact one
if you sum more and more terms. Comment in particularly on the
slope of the profiles at time and at the boundaries.
What should the slope be at those times? What is it?
Derive the (real) Fourier series for the -periodic
function with values
If possible, check your answer versus 24.17 in the math handbook.
Plot the Fourier series for increasing number of terms,
using 360 equally spaced plot points in
the range from 0 to .
Based on these plots, discuss whether the Fourier series really
converges to the given function for high enough number of terms:
When does an actual jump first show up?
The original function is undefined when
and
. What happens with the Fourier series at those
locations?
Are there points where the Fourier series does not converge?
If so, give the values for which it does not converge.
Supposing that the Fourier series converges everywhere,
and it does, the difference between the converged function and the
nonconverged function should become vanishingly small, not? In
particular the maximum difference between the Fourier series and the
converged function should become zero. Or does it? If not, explain
why not. In particular, plot the function for
using 1800 equally spaced plot points to explain your reasoning.