Lecture 10

February 21, 2021

We are going to use separation of variables to solve some second order PDEs.
Consider the homogeneous Dirichlet condtions for the wave equation:
2
u − c u = 0 f or 0 < x < l
tt xx (1)
u(0, t) = 0 = u(l, t) (2)
with some initial conditions
u(x, 0) = φ(x) (3)
u (x, 0) = ψ(x).
t (4)
A separated solution is a solution of (1) and (2) of the form
u(x, t) = X(x)T (t).

The equation (1) gives us

X(x)T  (t) = c2 X  (x)T (t).

Or we write into the form

T  (t) X  (x)
− = − = λ.
c2 T (t) X(x)

Note that
λ=−
X  (x)
X(x)
(5)
is independent with t, and
λ=−
T  (t)
c2 T (t)
(6)
does not depend on x. So λ is a constant.
Thus it reduced to solve two second order ODEs.
Theorem. Given a second order linear equation with constant coecients

y  (x) + by  (x) + cy(x) = 0, (7)

1
 Solve its characteristic equation r2 +br+c = 0. The general solution depends
on the type of roots obtained.
Case 1. When b2 − 4c > 0, there are two distinct real roots r1 , r2 . The
general solution to (7) will be
y(x) = c 1 e r1 x + c 2 e r 2 x .

Case 2. When b2 −4c < 0, there are two complex conjugate roots r1 = λ+μi,
r2 = λ − μi. The general solution to (7) will be

y(x) = c1 eλx cos μx + c2 eλx sin μx.

Case 3. When b2 − 4c = 0, there is one repeated real root r. Then

y(x) = c1 erx + c2 xerx .

From the ODE results, if λ = 0, we have from Case 3 that

X(x) = c1 + c2 x.

But X(0) = X(l) = 0 infers only c1 = c2 = 0. It only gives us a trivial

solution.
If λ = −β 2 < 0, we have from Case 1 that
X(x) = c1 eβx + c2 e−βx .

Similarly, X(0) = X(l) = 0 will gives us c1 = c2 = 0 which is also a trivial

solution.
If we assume λ = β 2 > 0 we have from Case 2
X(x) = C cos βx + D sin βx
T (t) = A cos βct + B sin βct.

The second step is to impose the boundary condition (2) on the separated
solution.
u(0, t) = X(0)T (t) = 0
u(l, t) = X(l)T (t) = 0.

So we have
X(0) = C=0
X(l) = D sin βl = 0.

If C = D = 0, this is a trivial solution which we are not interested. So we

must have β = nπl .
That is λn = ( nπl )2 , Xn (x) = sin nπx
l .

2
 So we have many solutions
nπct nπct nπx
un (x, t) = (An cos + Bn sin ) sin
l l l
where An and Bn are arbitrary constants (which can be determined by φ and
ψ ).
The sum of solutions is again a solution to (1) and (2)
 nπct nπct nπx
u(x, t) = (An cos + Bn sin ) sin . (8)
n
l l l

If formula (8) also solves (3) and (4), it has to provide that
 nπx
φ(x) = An sin
n
l

and
 nπc nπx
ψ(x) = Bn sin .
n
l l

1 Review second order linear ODE.

We are going to review how to get the ODE Theorem 1. Let us consider the
constant coecient second linear ODE
y  (t) + py  (t) + qy(t) = 0 (9)
There is an existence and uniqueness theorem due to PicardLindelof.
Theorem 2. Consider the initial value problem
y  (t) + p(t)y  (t) + q(t)y = g(t) a≤t≤b
y(t0 ) = y0 (10)

y (t0 ) = y0 .

If the functions p, q and g are continuous on the interval [a, b] which con-
taining the point t0 . Then there exists a unique solution y(t) to the initial value
problem.
This theorem guarantee us there are solutions to the equation (9). We begin
to nd the solution with some simple examples.
Example. y(t) = C sin t is a solution to the equation y (t)+y(t) = 0. From the
solution y(t) = C sin t we know that if we impose initial or boundary conditions
on dierent points there may have no uniqueness. For instance, for any C the
solutions y(t) = C sin t will satisfy the following condition 1: y(0) = 0 and
y  ( π2 ) = 0; or condition 2: y(0) = 0 and y(π) = 0.

3
Example. And y(t) = Cet is a solution to the equation y (t) − y(t) = 0.
So we may guess solutions to the equation (9) will be in the form y(t) = ert .
From the equation (9) we have
r2 ert + prert + qert = 0.

Thus we need solve the characteristic equation r2 + pr + q = 0 of Equation

(9). √
Case √
1. p2 − 4q > 0, we have two distinct real roots r1 = −p+ 2p −4q ,
2

. Then we get two solutions

2
−p− p −4q
r2 = 2

X1 (t) = e r1 t

and
X2 (t) = e r2 t .

Claim. In this case, the general solution to (9) must be in the form
y(t) = c1 X1 (t) + c2 X2 (t) = c1 er1 t + c2 er2 t .

Proof. In order to prove this claim, we need to introduce the Wronskian which
is dened
W (t) = X1 (t)X2 (t) − X2 (t)X1 (t).

If W (t) = 0, we say that X1 and X2 are linearly independent solutions to

the equation (9). In particularly W (t0 ) = 0. One advantage is that suppose the
solution is in the form of y(t) = c1 X1 (t)+c2 X2 (t), we can uniquely determine the
constant c1 and c2 from the initial value conditions y(t0 ) = y0 and y (t0 ) = y0 .
So we have only two parametric freedoms.
One important property of the Wronskian is that
W  (t) = [X1 (t)X2 (t) − X2 (t)X1 (t)]
= X1 (t)X2 (t) + X1 (t)X2 (t) − X2 (t)X1 (t) − X2 (t)X1 (t)
= X1 (t)X2 (t) − X2 (t)X1 (t)
f rom (9) = X1 (t)[−pX1 (t) − qX1 (t)] − X2 (t)[−pX2 (t) − qX2 (t)]
= −p[X1 (t)X1 (t) − X2 (t)X1 (t)]
= −pW (t). (11)
From this we can prove uniquess of Equation (9) with given initial conditions
y(t0 ) = y0 and y  (t0 ) = y0 . Suppose that y(t) and y(t) are the solutions satisfy
the initial value problem (10) then
W (t0 ) = y  (t0 ) − y  (t0 )
y(t0 ) y (t0 ) = y0 y0 − y0 y0 = 0.

4
 So from the equation (11) for W , we have W (t) ≡ 0. Which means y(t) =
y (t) then from the same initial condition y(t0 ) = y0 = y(t0 ) we have y(t) ≡ y(t).
c
This gives another proof of the uniqueness of Theorem 2.
Combining the uniqueness and the linear independent solutions X1 and X2 ,
we have proved the general solution to Equation (9) must be in the form of
y(t) = c1 X1 (t) + c2 X2 (t).

Case 3. p2 − 4q = 0, there is one repeated real root r = − p2 . So one

independent solution X1 (t) = ert . We nd another independent solution X2 (t)
from Equation (11). From Equation (11), we have
 
−pdt 2rdt
W (t) = e =e .
On the other hand by the denition
W (t) = X1 (t)X2 (t) − X2 (t)X1 (t).
Let X1 (t) = ert , we have
ert X2 (t) − rert X2 (t) = e2rt .
Solving this rst order linear inhomogeneous ODE, we get the other inde-
pendent solution
X2 (t) = tert .
By a similar reason as before, the general solution to Equation (9) is
y(t) = c1 ert + c2 tert .
Case 2. p2 − 4q < 0, there are two complex conjugate roots r1 = λ + μi,
r2 = λ − μi. So the fundamental solutions may be in the form
1 (t)
X = eλt+μti = eλt (cos μt + i sin μt)
and
2 (t)
X = eλt−μti = eλt (cos μt − i sin μt).
So in the real form the fundamental solution may be
X1 (t) = eλt cos μt
and
X2 (t) = eλt sin μt.
We can check the Wronskian W = μe2λt is nonzero. So X1 and X2 are
independent solutions.
By a similar reason as before, the general solution to Equation (9) in this
case is
y(t) = c1 eλt cos μt + c2 eλt sin μt.