Documente Academic
Documente Profesional
Documente Cultură
Solving Nonhomogeneous
PDEs (Eigenfunction
Expansions)
12.1 Goal
We know how to solve di¤usion problems for which both the PDE and the BCs
are homogeneous using the separation of variables method. Unfortunately, this
method requires that both the PDE and the BCs be homogeneous. We also
learned how to apply certain transformations so that nonhomogeneous BCs are
transformed into homogeneous ones. Unfortunately, these transformations may
in some cases, transform the PDE into a nonhomogeneous one. To complete the
set of tools we have to solve di¤usion problems, we must learn how to handle
nonhomogeneous PDEs. More speci…cally, we will show how to solve the IBVP
8
>
> PDE ut = 2 uxx + f (x; t) 0<x<1 0<t<1
<
u
1 x (0; t) + 1 u (0; t) = 0
BC 0<t<1 (12.1)
>
> 2 ux (1; t) + 2 u (1; t) = 0
:
IC u (x; 0) = (x) 0 x 1
by …nding a series solution of the form
1
X
u (x; y) = Tn (t) Xn (t)
n=1
where Xn (t) are the eigenfunctions we …nd when solving the associated homo-
geneous problem
8
>
> PDE ut = 2 uxx 0<x<1 0<t<1
<
1 ux (0; t) + 1 u (0; t) = 0
BC 0<t<1 (12.2)
>
> 2 ux (1; t) + 2 u (1; t) = 0
:
IC u (x; 0) = (x) 0 x 1
85
86CHAPTER 12. SOLVING NONHOMOGENEOUS PDES (EIGENFUNCTION EXPANSIONS)
and Tn (t) are functions which can be found by solving a sequence of ODEs.
12.2 Outline
Recall that the solution of 12.2 is of the form
1
X
)2 t
u (x; t) = An e ( n
Xn (x)
n=1
Where n and Xn (x) are the eigenvalues and eigenfunctions of the problem
8
< X 00 + 2 X = 0
X 0 (0) + 1 X (0) = 0
: 1 0
2 X (1) + 2 X (1) = 0
For the problem in 12.1, we will look for a solution of the form
1
X
u (x; t) = Tn (t) Xn (x)
n=1
The physical reason for this is that without f (x; t), there is no heat source, so it
is normal to expect the temperature to decrease with time, hence the damping
2
term e( n ) t . With a heat source, temperature will no longer decrease, hence
we might expect the part which depends on t to be di¤erent.
and …nd the response un (x; t) = Tn (t) Xn (x) to each of these individual com-
ponents. The solution to our problem will then be
1
X
u (x; t) = un (x; t)
n=1
We will break the procedure of solving this problem into several steps.
12.3. GENERAL IDEA 87
Step 1 Find the functions Xn (x). It turns out that the functions Xn (x)
are the eigenfunctions of the associated homogeneous problem when we
solve it by separation of variables. We derive this problem one more
time. The associated homogeneous PDE is ut = 2 uxx . If we look for
a solution of the form u (x; t) = T (t) X (x). Replacing in the PDE gives
T 0 (t) X (x) = 2 T (t) X 00 (x). Dividing each side by 2 T (t) X (x) gives
T 0 (t) X 00 (x)
2 T (t)
= . We concluded these had to be equal to a negative
X (x)
2
constant we called . Thus, to …nd X, we solve the second order ODE
00 2
X + X = 0. For X, the boundary condition meant that X (0) = 0
and X (1) = 0. Thus, we see that …nding X amounts to solving the initial
value problem 8
< X 00 + 2 X = 0
X (0) = 0
:
X (1) = 0
When we do so, we say that we are …nding the eigenfunctions of this
problem. You will recall that the solutions are X (x) = A sin x+B cos x.
Using the boundary conditions gives 0 = X (0) = B, so that X (x) =
A sin x. Also, we have 0 = X (1) = A sin . It follows that we must have
sin = 0 which means that = n for n = 1; 2; 3; 4; :::. If we call n = n
for each n = 1; 2; 3; 4; ::: then we have Xn (x) = sin n x. Note that we
omitted the constant, it will be part of the other components fn (t).
Step 2 Find the functions fn (t). So far, we have
f (x; t) = f1 (t) sin x + f2 (t) sin 2 x + ::: + fn (t) sin n x + :::
To …nd fn (t) we simply multiply each side by sin m x and integrate from
0 to 1 with respect to x. We have already used this method. We will have
Z 1 X1 Z 1
f (x; t) sin m xdx = fn (t) sin m xdx sin n xdx
0 n=1 0
1
= fm (t)
2
Thus Z 1
fn (t) = 2 f (x; t) sin n xdx (12.4)
0
Step 3 Find the response un (x; t) = Tn (t) Xn (x). We can replace the
nonhomogeneous term f (x; t) by its decomposition
1
X
f (x; t) = fn (t) sin n x
n=1
So, we have to …nd the functions Tn (t) which solve the IBVP 12.3. If we
replace u in that problem with the expression we have, we obtain
8 1 1 1
> X X X
>
> PDE 0
Tn (t) sin n x = 2 2
(n ) Tn (t) sin n x + fn (t) sin n x 0 < x < 1 0 < t <
>
>
>
> n=1 n=1 n=1
>
> X1
>
>
>
> Tn (t) sin 0 = 0
<
n=1
BC X1 0<t<
>
>
>
> Tn (t) sin n = 0
>
>
>
> n=1
>
> 1
X
>
>
>
: IC Tn (0) sin n x = (x) 0 x 1
n=1
The BCs do not give us any information, they simply say 0 = 0. We are
left with
8 X
1 h i
>
> 2
>
< Tn0 (t) + (n ) Tn (t) fn (t) sin n x = 0
n=1
1
X
>
>
>
: Tn (0) sin n x = (x)
n=1
1
X
u (x; t) = Tn (t) sin n x
n=1
X1 1
X Z t
(n )2 t (n )2 ( t)
= an e sin n x + sin n x fn ( ) e d
n=1 n=1 0
This shows in particular that the temperature in the rod is due to two
parts. One comes from the initial condition. The other one from the heat
source.
2
Tn0 + (n ) Tn = fn
Z 1
= 2 sin 3 x sin n xdx
0
0 if n 6= 3
=
1 if n = 3
And
Z 1
Tn (0) = 2 sin x sin n xdx
0
0 if n 6= 1
=
1 if n = 1
90CHAPTER 12. SOLVING NONHOMOGENEOUS PDES (EIGENFUNCTION EXPANSIONS)
2 1 e (3 ) t .
(3 )
(n )2 t
Solution for n 4 T2 (t) = Ae . Since Tn (0) = 0, it follows that A =
0. Thus, Tn (t) = 0.
Thus, we see that the solution is
( )2 t 1 (3 )2 t
u (x; t) = e sin x + 2 1 e sin 3 x
(3 )
Remark 63 It is important to realize that the eigenfunctions Xn (x) and the
eigenvalues n which appear in the solution of nonhomogeneous problems vary
for each problem. They depend on the PDE used and the BCs.
12.5 Problems
1. In the last example, …nd the solution in the case n = 3.
2. Solve the problem
8
>
> PDE ut = uxx + sin x + sin 2 x 0 < x < 1 0 < t < 1
<
u (0; t) = 0
BC 0<t<1
>
> u (1; t) = 0
:
IC u (x; 0) = 0 0 x 1
where 1 is the …rst root of the equation tan = . What are the
eigenfunctions Xn in this problem?
5. Solve the problem
8
>
> PDE ut = uxx 0<x<1 0<t<1
<
u (0; t) = 0
BC 0<t<1
>
> u (1; t) = cos t
:
IC u (x; 0) = x 0 x 1
by: