Strauss PDEs 2e: Section 5.

6 - Exercise 6 Page 1 of 6

Exercise 6
Solve utt = c2 uxx + g(x) sin ωt for 0 < x < l, with u = 0 at both ends and u = ut = 0 when t = 0.
For which values of ω can resonance occur? (Resonance means growth in time.)

Solution

Since the PDE is linear and inhomogeneous, we choose to apply the method of eigenfunction
expansion to solve it. Consider the eigenvalue problem of the differential operator involving the
spatial variable x
d2
φ = λφ (1)
dx2
with the same boundary conditions as u.

φ(0) = 0
φ(l) = 0

Values of λ for which the boundary conditions are satisfied are known as the eigenvalues, and the
nontrivial solutions associated with them are called the eigenfunctions. Equation (1) is known as
the one-dimensional Helmholtz equation; the eigenfunctions for it are known to be orthogonal and
form a complete set, which will prove useful later.

Determination of Positive Eigenvalues: λ = µ2

Suppose that λ is positive. Then equation (1) becomes

d2 φ
= µ2 φ.
dx2
Its solution can be written in terms of hyperbolic sine and hyperbolic cosine.

φ(x) = C1 cosh µx + C2 sinh µx

Apply the boundary conditions to determine C1 and C2 .

φ(0) = C1 = 0
φ(l) = C1 cosh µl + C2 sinh µl = 0

Since C1 = 0, the second equation reduces to C2 sinh µl = 0. Hyperbolic sine is not oscillatory, so
the only way this equation is satisfied is if C2 = 0. The trivial solution is obtained, so there are no
positive eigenvalues.

Determination of the Zero Eigenvalue: λ = 0

Suppose that λ is zero. Then equation (1) becomes

d2 φ
= 0.
dx2
The general solution is obtained by integrating both sides with respect to x twice.

φ(x) = C3 x + C4

www.stemjock.com
Strauss PDEs 2e: Section 5.6 - Exercise 6 Page 2 of 6

Apply the boundary conditions to determine C3 and C4 .

φ(0) = C4 = 0
φ(l) = C3 l + C4 = 0

Since C4 = 0, the second equation reduces to C3 = 0. The trivial solution is obtained, so zero is
not an eigenvalue.

Determination of Negative Eigenvalues: λ = −γ 2

Suppose that λ is negative. Then equation (1) becomes

d2 φ
= −γ 2 φ.
dx2
Its solution can be written in terms of sine and cosine.

φ(x) = C5 cos γx + C6 sin γx

Apply the boundary conditions to determine C5 and C6 .

φ(0) = C5 = 0
φ(l) = C5 cos γl + C6 sin γl = 0

Since C5 = 0, the second equation reduces to C6 sin γl = 0. To avoid getting the trivial solution,
we insist that C6 6= 0. Then

sin γl = 0
γl = nπ, n = 1, 2, . . .
nπ
γn = , n = 1, 2, . . . .
l
The eigenfunctions associated with these eigenvalues for λ are
nπx
φ(x) = C6 sin γx → φn (x) = sin , n = 1, 2, . . . .
l
Method 1 - Using Term-by-Term Differentiation

The eigenfunctions of the Helmholtz equation form a complete set, so the unknown function u can
be expanded in terms of them.
∞
X nπx
u(x, t) = an (t) sin
l
n=1

To determine the generalized Fourier coefficients an (t), substitute this expansion into the PDE.

utt = c2 uxx + g(x) sin ωt

∞ ∞
∂2 X nπx 2 ∂
2 X nπx
2
a n (t) sin = c 2
an (t) sin + g(x) sin ωt
∂t l ∂x l
n=1 n=1

www.stemjock.com
Strauss PDEs 2e: Section 5.6 - Exercise 6 Page 3 of 6

Because u satisfies homogeneous boundary conditions and u, ∂u/∂x, and ∂u/∂t are continuous
(reasonable assumptions for the displacement of a homogeneous elastic string), the two series can
in fact be differentiated term by term.
∞ ∞
X d2 an nπx X d2 nπx
sin = c2 an (t) 2 sin + g(x) sin ωt
dt2 l dx l
n=1 n=1

The operator applied to the eigenfunction is equal to the eigenvalue times the eigenfunction.
∞ ∞
X d2 an nπx X nπx
sin = c2 an (t)λn sin + g(x) sin ωt
dt2 l l
n=1 n=1

Bring both series to the left side and combine them.

∞  2 
X d an 2 nπx
− c λn an (t) sin = g(x) sin ωt
dt2 l
n=1

The left side is essentially a Fourier sine series expansion of g(x) sin ωt. To solve for the term in
square brackets, multiply both sides by sin(mπx/l), where m is an integer,
∞  2 
X d an 2 nπx mπx mπx
− c λn an (t) sin sin = g(x) sin sin ωt
dt2 l l l
n=1

and then integrate both sides with respect to x from 0 to l.

ˆ lX
∞  2  ˆ l
d an 2 nπx mπx mπx
2
− c λn an (t) sin sin dx = g(x) sin sin ωt dx
0 dt l l 0 l
n=1

g(x) is assumed not to be orthogonal to sin(mπx/l) so that the integral on the right side is
nonzero. Bring the functions of t in front of the integrals.
∞  2 ˆ l ˆ l
X d an 2 nπx mπx mπx
− c λn an (t) sin sin dx = sin ωt g(x) sin dx
dt2 0 l l 0 l
n=1

Since the eigenfunctions are orthogonal, the integral on the left side is zero if n 6= m. As a result,
every term in the infinite series vanishes except for one: n = m.
 2 ˆ l ˆ l
d an 2 2 nπx nπx
2
− c λn an (t) sin dx = sin ωt g(x) sin dx
dt 0 l 0 l

Evaluate the integral on the left side.

 2  ˆ l
d an 2 l nπx
− c λ a
n n (t) · = sin ωt g(x) sin dx
dt2 2 0 l

Multiply both sides by 2/l and replace λn with −(nπ/l)2 .

 ˆ l
d2 an 2 2

2n π 2 nπx
+c an = g(x) sin dx sin ωt
dt2 l2 l 0 l

www.stemjock.com
Strauss PDEs 2e: Section 5.6 - Exercise 6 Page 4 of 6

With the help of the method of eigenfunction expansion, the PDE has been reduced to a
second-order inhomogeneous ODE. Because the ODE is linear, the general solution is the sum of
a complementary solution and a particular solution.

an = ac + ap

The complementary solution satisfies the associated homogeneous equation.

d2 ac 2 2
2n π
+ c ac = 0
dt2 l2
Its general solution can be written in terms of sine and cosine.
cnπt cnπt
ac (t) = C7 cos + C8 sin
l l
Since the inhomogeneous term is sine and there are no odd derivatives, the particular solution is
of the form ap = C0 sin ωt. Substitute it into the ODE to find C0 .
 ˆ l
d2 ap 2 2

2n π 2 nπx
+c ap = g(x) sin dx sin ωt
dt2 l2 l 0 l
2 2
 ˆ l 
2 2n π 2 nπx
−ω C0 sin ωt + c C0 sin ωt = g(x) sin dx sin ωt
l2 l 0 l
ˆ
2 2
2n π 2 2 l nπx
c 2
C0 − ω C0 = g(x) sin dx
l l 0 l
ˆ
c2 n2 π 2 − l2 ω 2 2 l nπx
C0 = g(x) sin dx
l2 l 0 l
ˆ l
2l nπx
C0 = 2 2 2 2 2
g(x) sin dx
c n π −l ω 0 l

Hence, the general solution for an is

ˆ l
cnπt cnπt 2l sin ωt nπx
an (t) = C7 cos + C8 sin + 2 2 2 g(x) sin dx.
l l c n π − l2 ω 2 0 l

Use the initial conditions for u in combination with the eigenfunction expansion to determine
those for an .
∞
X nπx
u(x, 0) = an (0) sin =0 ⇒ an (0) = 0
l
n=1
∞
X dan nπx dan
ut (x, 0) = (0) sin =0 ⇒ (0) = 0
dt l dt
n=1

Apply them both to obtain a system of equations for C7 and C8 .

a(0) = C7 = 0
ˆ l
dan cnπ 2lω nπx
(0) = (C8 ) + 2 2 2 g(x) sin dx = 0
dt l c n π − l2 ω 2 0 l

www.stemjock.com
Strauss PDEs 2e: Section 5.6 - Exercise 6 Page 5 of 6

Solving the second equation for C8 gives

  ˆ l
ωl 2l nπx
C8 = − 2 2 2 2 2
g(x) sin dx.
cnπ c n π − l ω 0 l
So then
  ˆ l  ˆ l
ωl 2l nπx cnπt 2l sin ωt nπx
an (t) = − g(x) sin dx sin + 2 2 2 g(x) sin dx
cnπ c2 n2 π 2 − l2 ω 2 0 l l c n π − l2 ω 2 0 l
ˆ l  
2l nπx ωl cnπt
= 2 2 2 g(x) sin dx sin ωt − sin .
c n π − l2 ω 2 0 l cnπ l
Therefore,
∞ ˆ l  
X 2l nπs ωl cnπt nπx
u(x, t) = g(s) sin ds sin ωt − sin sin .
c n π − l2 ω 2
2 2 2
0 l cnπ l l
n=1

The dummy integration variable has been changed to s to distinguish it from x. Resonance occurs
when the solution blows up: c2 n2 π 2 − l2 ω 2 = 0 (that is, if ω is a positive integer multiple of cπ/l).

Method 2 - Without Using Term-by-Term Differentiation

The eigenfunctions of the Helmholtz equation are known to form a complete set, so all of the
functions in the PDE can be expanded in terms of them.
∞ ∞ ˆ l ˆ l
X X l
u(x, t) = An (t)φn (x) → uφm = An φ n φ m → uφn dx = An φ2n dx = An ·
0 0 2
n=1 n=1
∞ ∞ ˆ l 2 ˆ l
∂2u X ∂2u X ∂ u l
2
= B n (t)φ n (x) → 2
φ m = B φ
n n mφ → 2
φ n dx = B n φ2n dx = Bn ·
∂t ∂t 0 ∂t 0 2
n=1 n=1
∞ ∞ ˆ l ˆ l
X X l
g(x) sin ωt = Dn (t)φn (x) → g(x) sin ωtφm = Dn φn φm → sin ωt gφn dx = Dn φ2n dx = Dn ·
0 0 2
n=1 n=1
∞ ∞ ˆ l 2 ˆ l
∂2u X ∂2u X ∂ u l
2
= En (t)φ n (x) → 2
φ m = E φ
n n mφ → 2
φ n dx = En φ2n dx = En ·
∂x ∂x 0 ∂x 0 2
n=1 n=1

It should be emphasized that these are generalized Fourier series expansions for the functions, not
product solutions that come about from using the method of separation of variables. Solve the
latter equations for the generalized Fourier coefficients.
ˆ
2 l
An (t) = uφn dx
l 0
ˆ  ˆ
2 l ∂2u d2 2 l d2 An

Bn (t) = φ n dx = uφn dx =
l 0 ∂t2 dt2 l 0 dt2
ˆ l
2 sin ωt
Dn (t) = g(x)φn dx
l 0
ˆ ˆ l ˆ
2 l ∂2u 2 ∂u l 2nπ l ∂u
  
∂u dφn nπx
En (t) = 2
φn dx = φn  − dx = − 2 cos dx
l 0 ∂x l ∂x 0 0 ∂x dx l 0 ∂x l
| {z }
=0

www.stemjock.com
Strauss PDEs 2e: Section 5.6 - Exercise 6 Page 6 of 6

Apply integration by parts once more in order to write En in terms of An .

ˆ l 
nπx l
  
2nπ nπ nπx 
En (t) = − 2 u cos − u − sin dx
l l 0 0 l l
| {z }
=0
 ˆ
n2 π 2 2 l

nπx
=− 2 u sin dx
l l 0 l
n2 π 2
= − 2 An
l
Now that the coefficients are known, substitute the eigenfunction expansions into the PDE.

utt = c2 uxx + g(x) sin ωt

∞
X X∞ ∞
X
Bn (t)φn (x) = c2 En (t)φn (x) + Dn (t)φn (x)
n=1 n=1 n=1
X∞ ∞
X
2
Bn (t)φn (x) = [c En (t) + Dn (t)]φn (x)
n=1 n=1

Thus,
Bn (t) = c2 En (t) + Dn (t).
Substitute the formulas for Bn , En , and Dn to obtain an ODE for An exclusively.
ˆ l
d2 An 2 2
2n π 2 sin ωt
2
= −c 2
An + g(x)φn dx
dt l l 0

Bring the term with An to the left side and replace φn with sin(nπx/l).
 ˆ l
d2 An 2 2

2n π 2 nπx
+ c A n = g(x) sin dx sin ωt
dt2 l2 l 0 l

This is the same ODE that was obtained for an in Method 1. The initial conditions are also the
same as before, so An (t) = an (t).
ˆ l  
2l nπx ωl cnπt
An (t) = 2 2 2 g(x) sin dx sin ωt − sin
c n π − l2 ω 2 0 l cnπ l

Therefore,
∞ ˆ l  
X 2l nπs ωl cnπt nπx
u(x, t) = g(s) sin ds sin ωt − sin sin .
c n π − l2 ω 2
2 2 2
0 l cnπ l l
n=1

The dummy integration variable has been changed to s to distinguish it from x. Resonance occurs
when the solution blows up: c2 n2 π 2 − l2 ω 2 = 0 (that is, if ω is a positive integer multiple of cπ/l).

www.stemjock.com