Fourier's Approach to Heat Flow
Consider the infinite rectangular lamina, 2 units wide.
The equation that models heat flow in this lamina is given by
Fourier was interested in the equilibrium solution where and chose the boundary conditions:
and
As well as
bounded as
.
He assumed that Substituting this into the equation gives
Since we can separate this equation into two functions one in x only and one in y only it must be the
case that
a constant.
This gives two separate equations:
|
|
|
|
We solve the equation for G first. We have . Using the boundary conditions
we have
and
For arbitrary values of λ this system will have only the trivial solution and we want a nontrivial solution. Adding these
equations gives: . If
we take
or λ
. Since
does not satisfy the boundary conditions we take We obtain then an infinite number of solutions given by
For the equation we have
If we require that
be
bounded then and so
Therefore we obtain the solutions
This then gives the equilibrium solutions as
Fourier then noted that, the principle of superposition, that the general solution has the form
and from the boundary condition we have to choose the values
to satisfy
(A)
Fourier differentiated this expression and infinite number of times and generated a system for the infinite number of unknowns
. He offered no justification why term-by-term differentiation was possible. Later on he noted the orthogonality of the functions
on the interval [-1, 1]. i.e.
Fourier then multiplies (A) by and integrates term-by-term, again with no justification.
Note that the only nonzero term is when and we arrive at
The solution is then given by the infinite series
Let's consider partial sums of this function.
Graphs of this function for and various values of
are shown below.
How do these partial sums converge to the square wave.
Notice that it overshoots near the enpoints of the interval. This is known and the Gibbs-Wilbraham phenomenom.
For a three dimensional veiw of the solution we plot the temperature variation on the rectangle [0, 2]×[-1,1].
Also