PARTIAL DIFFERENTIAL EQUATIONS (MATH417)

SOLUTIONS FOR THE FINAL EXAM

Problem 1 (10 pts.) Classify each equation as linear homogeneous, linear

inhomogeneous, or nonlinear:
2
a) ∂u ∂ u
∂t = 2 ∂x 2 + x
3
b) ∂u ∂u
∂t2 + 2 ∂x = 3u
∂2u ∂2u ∂4u
c) ∂x2 + ∂y2 = u ∂y4 d) ∂x∂y = e−x y
∂ u

Solution:
(a) is a linear inhomogeneous equation;
(b) is a linear homogeneous equation; (c) is a nonlinear (because of the right-hand
side) equation; (d) is a linear inhomogeneous equation.

Problem 2 (20 pts.) Let f (x) = x2 .

(1) Find the Fourier series (complex form) of f (x) on the interval (−π, π).

(2) Rewrite the Fourier series in the real form.

(3) Sketch the function to which the Fourier series converges.

∑∞
(4) Use Parseval’s equality to evaluate n=1 n−4 .
∑
Solution: (1) The required series is f (x) ∼ n∈Z cn einπx/L with
∫ π
1
cn = f (x)e−inx dx.
2π −π
If n = 0, then
∫ π
1 1 x3 π π2
c0 = x2 dx = |−π = .
2π −π 2π 3 3
If n ̸= 0, we can integrate by parts twice to get
2(−1)n
cn = .
n2
We can also use the given formula
∫
x2 2x 2i
x2 eiax dx = ( + 2 + 3 )eiax + c, a ̸= 0.
ia a a
According to this integral,
∫ π
1 1 x2 2x 2i 1 2x −inx π e−inπ + einπ
cn = x2 e−inx dx = (− + 2 − 3 )e−inx |π−π = e | −π = .
2π −π 2π in n n 2π n2 n2
Thus
π2 ∑ 2(−1)n inx
x2 ∼ + e
3 n2
n̸=0,n∈Z
1
2 SOLUTION FOR FINAL

-2 Pi -Pi 0 Pi 2 Pi

Figure 1.

(2) Rewrite it in the real form,

π2 ∑ 2(−1)n inx π 2 ∑ 2(−1)n inx −inx π 2 ∑ 4(−1)n
+ e = + (e +e ) = + cos(nx)
3 n2 3 n2 3 n2
n̸=0,n∈Z n≥1 n≥1

Thus
π 2 ∑ 4(−1)n
x2 ∼ + cos(nx).
3 n2
n≥1

(3) Sketch the function to which the Fourier series converges. The series con-
verges to the 2π-periodic function that coincides with f (x) for −π < x < π. The
sum is continuous and piecewise smooth hence the convergence is uniform. see
Figure 1. ∑∞
(4) Use Parseval’s equality to evaluate n=1 n−4 . Recall that
∑ ∫ π
1
f (x) ∼ cn einx , cn = f (x)e−inx dx
2π −π
n∈Z
∫ π ∑
1
|f |2 dx = |cn |2
2π −π n∈Z
Now ∫ π
1 1 x5 π π4
|f |2 dx = |−π = ,
2π −π 2π 5 5
∑ π2 2 ∑ 2(−1)n
|cn |2 = ( ) +2 ( )2
3 n2
n∈Z n≥1

Thus
π4 π4 ∑ 8
= +
5 9 n4
n≥1
 SOLUTION FOR FINAL 3

It follows that
∑ 1 1 π4 π4 π4
4
= ( − )= .
n 8 5 9 90
n≥1

Problem 3 (40 pts.) a) By the method of your choice, solve the wave equation
on the half-line
 ∂2u 2
 ∂t2 = ∂∂xu2 0 < x < ∞, 0 < t < ∞,
∂u
(0, t) = 0 0 < t < ∞,
 ∂x
u(x, 0) = x2 e−x , ∂u
∂t (x, 0) = 0, 0 < x < ∞.
b) Solve the problem
 ∂2u ∂2u
 ∂t2 = ∂x2 0 < x < ∞, 0 < t < ∞,
∂u
∂x (0, t) =
t 0 < t < ∞,
 2 −x ∂u
u(x, 0) = x e , ∂t (x, 0) = 0, 0 < x < ∞.

a) D’Alembert’s method: We use method of characteristic to give the solution

here
i)Let f be the even extension of the function x2 e−x (x > 0), i.e. f (x) = x2 e−|x| .
Then we consider instead the problem on real line.
ii) We know the general solution can be given by
u(x, t) = F (x − t) + G(x + t).
IC gives us that
F (x) + G(x) = f (x),
G′ (x) − F ′ (x) = 0,
which means that we can set
f (x)
F (x) = ,
2
f (x)
G(x) = .
2
iii) thus we can know that
f (x + t) + f (x − t) 1
u(x, t) = = [(x + t)2 e−|x+t| + (x − t)2 e−|x−t| ].
2 2
Fourier’s method. In view of the BC, we use Fourier cosine transform to solve
it.
Recall that
∫ ∞ ∫ ∞
2
C[f ](ω) = f (x) cos ωxdx, f (x) = C[f ](ω) cos ωxdω.
π 0 0

df 2
C[ ] = − f (0) + ωS[f ],
dx π
df
S[ ] = −ωC[f ].
dx
Then
d2 f 2 df df 2 df
C[ ]=− (0) + ωS[ ] = − (0) − ω 2 C[f ].
dx2 π dx dx π dx
4 SOLUTION FOR FINAL

Apply these formula to the PDE, we get

∂ 2 C[u](ω, t) 2 du
2
=− (0, t)) − ω 2 C[u](ω, t) = −ω 2 C[u](ω, t).
∂t π dx
The ODE has general solution for ω ̸= 0,
C[u](ω, t) = a(ω) cos ωt + b(ω) sin ωt.
We also have the IC
∂C[u](ω, t)
C[u](ω, 0) = C[f ](ω), = 0.
∂t
Thus the general solution to this ODE in t for fixed ω > 0 is given by
C[u](ω, t) = C[f ](ω) cos ωt.
Recall
2 cos ωt cos xω = cos(x − t)ω + cos(x + t)ω = cos |x − t|ω + cos |x + t|ω,
then
C −1 [C[u]](x, t)
u(x, t) =
∫ ∞
= C[f ](ω) cos ωt cos xωdω
0
∫ ∞ ∫ ∞
1
= [ C[f ](ω) cos |x − t|ωdω + C[f ](ω) cos |x + t|ωdω]
2 0 0
1
= [f (|x − t|) + f (|x + t|)]
2
1
= [(x + t)2 e−|x+t| + (x − t)2 e−|x−t| ].
2
b) D’Alembert’s method: We use method of characteristic to give the solu-
tion. Let
f (x) = x2 e−x , x > 0
i) We know the general solution can be given by
u(x, t) = F (x − t) + G(x + t).
IC gives us that
F (x) + G(x) = f (x), x > 0
G′ (x) − F ′ (x) = 0, x > 0
which means that we can set
f (x) x2 −x
F (x) = G(x) = = e , x > 0.
2 2
iii) To solve the problem, we still need to know F for negative variables, which
comes from the BC. By BC, we see that
F ′ (−t) + G′ (t) = t, t > 0
then for x < 0,
x2
F ′ (x) = −x − G′ (−x), ⇒ F (x) = c − + G(−x)
2
 SOLUTION FOR FINAL 5

02
we would like to have F continuous at 0, which gives c = 2 − G(−0) = G(0) = 0.
Thus we can know that
{ 2
x −x
F (x) = 2 e x>0
x2 x
2 (e − 1) x<0
{
F (x − t) + G(x + t) = 21 [(x + t)2 e−x+t + (x − t)2 e−x−t ] x ≥ t
u(x, t) = (x−t)2 x−t 2

2 (e − 1) + (x+t)
2 e−x−t x<t
Problem 4 (30 pts+5 bonus) Consider the regular Sturm-Liouville eigenvalue
problem
d du
(p(x) ) + q(x)u + λσ(x)u = 0.
dx dx
for λ ≫ 1 with ϕ′ > 0.
1/2
We try to find the solution of form u(x) = A(x)eiλ ϕ(x)

a)[10 pts] substitute to the DE, we may get

1/2
[λB0 (x) + iλ1/2 B1 (x) + B2 (x)]eiλ ϕ
=0
determine Bi (x).
b)[5 pts] By letting B0 = B1 = 0, solve for A = A0 (x) and ϕ. This will give the
first approximation of the eigenfunction.
c)[5 pts] In general, we can not have B2 = 0 for the choice in 2). Note that we
have B0 = 0 independent of A, the DE for A is iλ1/2 B1 (x) + B2 (x) = 0. Instead,
let A(x) = A0 (x) + λ−1/2 A1 (x) + λ−1 A2 (x) + · · · . Comparing the coefficients of
powers of λ, we can solve for An (x). Use the described method to solve for A1 in
terms of A0 (together with p, q, σ).
d)[10 pts] Consider the particular case p = q = σ = 1, u(0) = ux (π) = 0. By
using the Rayleigh quotient, give an upper bound for λ1 , the smallest eigenvalue.
e) (Bonus 5points) What is the precise value of λ1 ?
1/2
Solution: a) since u(x) = A(x)eiλ ϕ(x) , we have

u′ (x) = A′ eiλ + iλ1/2 ϕ′ Aeiλ

1/2 1/2
ϕ ϕ

u′′ (x) = A′′ eiλ + iλ1/2 ϕ′ A′ eiλ + iλ1/2 (ϕ′ A)′ eiλ − λ(ϕ′ )2 Aeiλ
1/2 1/2 1/2 1/2
ϕ ϕ ϕ ϕ

Substitute in the equation

d du
(p(x) ) + q(x)u + λσ(x)u = pu′′ + p′ u′ + qu + λσu = 0.
dx dx
we get
λ(σ−p(ϕ′ )2 )Aeiλ +iλ1/2 [p(ϕ′ A′ +(ϕ′ A)′ )+p′ ϕ′ A]eiλ +(pA′′ +p′ A′ +qA)eiλ
1/2 1/2 1/2
ϕ ϕ ϕ
=0
We see that
B0 (x) = (σ − p(ϕ′ )2 )A
B1 (x) = p(ϕ′ A′ + (ϕ′ A)′ ) + p′ ϕ′ A = 2A′ pϕ′ + A(pϕ′ )′
B2 (x) = pA′′ + p′ A′ + qA
b) B0 = 0 tells us that
∫ x
′
ϕ = (σ/p) 1/2
⇒ ϕ(x) = (σ(t)/p(t))1/2 dt
6 SOLUTION FOR FINAL

B1 = 0 tells us that
A′ 1 (pϕ′ )′
=− ⇒ A = c(pϕ′ )−1/2 = c(σp)−1/4
A 2 pϕ′
c) Fix the choice of ϕ, we want to have
iλ1/2 B1 (A) + B2 (A) = 0 .
Let A(x) = A0 (x) + λ−1/2 A1 (x) + O(λ−1 ), we have
iλ1/2 B1 (A0 ) + B2 (A0 ) + λ−1/2 (iλ1/2 B1 (A1 ) + B2 (A1 )) + O(λ−1 λ1/2 ) = 0 ,
that is
B2 (A0 ) + iB1 (A1 ) = O(λ−1/2 )
Let λ → ∞, we see that we must have
B2 (A0 ) + iB1 (A1 ) = 0

1
2i[A′1 pϕ′ + A1 (pϕ′ )′ ] + pA′′0 + p′ A′0 + qA0 = 0
2
′ −1/2
Integrating factor (pϕ ) , we see
i
[(pϕ′ )1/2 A1 ]′ = (pϕ′ )−1/2 [qA0 + (pA′0 )′ ]
2
Recall pϕ′ = (σp)1/2 ,
∫ x
i −1/4
A1 = (σp) (σp)−1/4 [qA0 + (pA′0 )′ ]dt
2
d) ∫π ∫ π ′2
−puu′ |π0 + 0 [pu′2 − qu2 ]dx [u − u2 ]dx
RQ[u] = ∫π = 0 ∫π 2
2
u σdx u dx
0 0
Based on the BCs u(0) = ux (π) = 0, we choose a trial function (without interior
zeros) {
x 0 < x < π/2
u(x) =
π/2 π/2 < x < π
then
∫ π/2 ∫π 2
dx − u dx π/2 3
λ1 ≤ RQ[u] = 0 ∫ π 2 0 = ∫ π/2 −1= 2 −1
0
u dx 2
x dx + π/2(π/2) 2 π
0
e) In this case, the SL eigenvalue problem becomes
d2 u
+ u + λu = 0
dx2
which is just the standard problem with (p, q, σ) = (1, 0, 1) with eigenvalue λ + 1
d2 u
+ (λ + 1)u = 0
dx2
For this problem, we know that λ + 1 > 0. u(0) = 0 tells us that
u(x) = sin(kx)
′
u (π) = 0 tells us that the first eigenfunction corresponding to k = 12 . That is
u1 (x) = sin x/2
 SOLUTION FOR FINAL 7

and
λ1 + 1 = (1/2)2
λ1 = −3/4