Math 220A

Practice Final Exam I Solutions - Fall 2002

1. (a) Suppose S(t) is the solution operator associated with the homogeneous equation
(
ut + aux = 0
(∗)
u(x, 0) = φ(x).

In particular, assumeRthe solution of (*) is given by u(x, t) = S(t)φ(x). Show that

t
v(x, t) = S(t)φ(x) + 0 S(t − s)f (x, s) ds solves the inhomogeneous problem
(
ut + aux = f (x, t)
u(x, 0) = 0.

Answer:
½ Z t ¾
[∂t + a∂x ]v = [∂t + a∂x ] S(t)φ(x) + S(t − s)f (x, s) ds
0
Z t
= 0 + S(t − t)f (x, t) + [∂t + a∂x ]S(t − s)f (x, s) ds
0
= S(0)f (x, t) = f (x, t).

In addition,
Z 0
v(x, 0) = S(0)φ(x) + S(0 − s)f (x, s) ds = φ(x).
0

(b) Find the solution operator S(t) for (*).

Answer: The solution of (*) is given by u(x, t) = φ(x − at). Therefore, the
solution operator S(t) is the operator such that

S(t)φ(x) = φ(x − at).

(c) Find a solution of the inhomogeneous initial-value problem

(
ut + aux = f (x, t)
u(x, 0) = φ(x).

Answer: A solution is given by

Z t
v(x, t) = S(t)φ + S(t − s)f (x, s) ds
0
Z t
= φ(x − at) + f (x − a(t − s), s) ds.
0

1
2. (a) Solve the following initial-value problem.
(
u2x ut − 1 = 0
u(x, 0) = x.

Answer: Let
F (p, q, z, x, t) = p2 q − 1.
The set of characteristic equations are given by
dx
ds
= 2pq x(r, 0) = r
dt
ds
= p2 t(r, 0) = 0
dz
ds
=3 z(r, 0) = r
dp
ds
=0 p(r, 0) = ψ1 (r)
dq
ds
=0 q(r, 0) = ψ2 (r)
where ψ1 , ψ2 satisfy
φ0 (r) = ψ1 (r)
ψ12 ψ2 − 1 = 0.
Therefore,
ψ1 (r) = 1 = ψ2 (r).
Solving this system of ODEs, we have
p=1
q=1
x = 2s + r
t=s
z = 3s + r.
Solving for r, s, we find our solution is given by

u(x, t) = z(r(x, t), s(x, t)) = x + t.

(b) Consider the initial-value problem

(
ut + ux = x
u(x, x) = 1.

Explain why there is no solution to this problem.

Answer: The projected characteristic curves for this PDE are given by
dt
=1
ds
dx
= 1.
ds

2
 Therefore, they are the lines x−t = c. Further, du/ds = x along the characteristic
curves. But we are prescribing initial data which is constant along the projected
characteristics. Therefore, du/ds 6= x. Our initial data does not satisfy our
equation.

3. (a) Find the general solution of

utt + 2uxt − 3uxx = 0.

Answer: Factoring as

(∂t − ∂x )(∂t + 3∂x )u = 0,

then we make a change of variables by defining new coordinates ξ, η such that

∂
= ∂t − ∂x
∂ξ
∂
= ∂t + 3∂x .
∂η
In particular, we let
1
ξ = − (x − 3t)
4
1
η = (x + t).
4
Therefore, we have
uξη = 0,
which implies

u(x, t) = f (ξ(x, t)) + g(η(x, t)) = f (x − 3t) + g(x + t).

(b) Find the solution of the initial-value problem,


 utt + 2uxt − 3uxx = 0

u(x, 0) = φ(x)


ut (x, 0) = ψ(x).

Answer: The general solution is given by

u(x, t) = f (x − 3t) + g(x + t).

Therefore, the initial data implies we need

u(x, 0) = f (x) + g(x) = φ(x)

ut (x, 0) = −3f 0 (x) + g 0 (x) = ψ(x).

3
 Solving this system of equations, we have
1
f 0 (x) = [φ0 (x) − ψ(x)]
4
1
g 0 (x) = [3φ0 (x) + ψ(x)].
4
Integrating these equations, we conclude that the solution to our initial-value
problem is given by
Z x+t
1 1
u(x, t) = [φ(x − 3t) + 3φ(x + t)] + ψ(y) dy.
4 4 x−3t

4. Consider the initial-value problem

(
ut + uux = 0
u(x, 0) = φ(x)

where 

 a x≤0
φ(x) = a(1 − x) 0<x<1


0 x≥1
where a > 0. Find the unique, weak solution which satifies the entropy condition.
Answer: The projected characteristics are given by

x(r) = φ(r)t + r.

For r < 0, we have x = at + r. For 0 < r < 1, we have x = a(1 − r)t + r. For
r > 1, we have x = r. We see these curves do not intersect until t = 1/a. Therefore,
for 0 ≤ t ≤ 1/a, our solution is well-defined, and the solution is constant along these
projected characteristics. In particular, for 0 ≤ t ≤ 1/a, our solution is given by


 a x < at

 µ1−x¶
u(x, t) = a at < x < 1

 1 − at


0 x > 1.

For t ≥ 1/a, the projected characteristics intersect. Therefore, we need to introduce a

shock curve. The values of the solution to the left and right of the curve of discontinuity
are given by u− = a and u+ = 0. Our shock curve x = ξ(t) must satisfy
1
[f (u)] (u− )2 − 12 (u+ )2
ξ 0 (t) = = 2
[u] u− − u +
1 2
2
a 1
= = a.
a 2

4
 This curve x = ξ(t) also
¡ contains
¢ the point t = 1/a, x = 1. Therefore, this curve is
given by (x − 1) = 12 a t − a1 . Therefore, for t ≥ 1/a our solution is given by

 1 1
a x < at +
u(x, t) = 2 2

0 1 1
x > at + .
2 2
5. Consider the initial-value problem

 utt + 2uxt − 3uxx = 0

u(x, 0) = φ(x)

 u (x, 0) = ψ(x)
t

(a) Use energy methods to prove the value of the solution u at the point (x0 , t0 )
depends at most on the values of the initial data in the interval (x0 − 3t0 , x0 + t0 ).
Answer: Define an energy for this problem by
Z
1
E(t) = (u2t + 3u2x ) dx.
2 R
Now for a fixed t, define the energy over the interval (x0 − 3(t0 − t), x0 + (t0 − t))
as Z
1 x0 +(t0 −t) 2
e(t) = (u + 3u2x ) dx.
2 x0 −3(t0 −t) t
Suppose the initial data φ, ψ is zero in the interval (x0 −3t0 , x0 +t0 ). We will show
that the solution is zero in the triangle bounded by the lines t = 0, x = x0 −3(t0 −t)
and x = x0 + (t0 − t). We will do so by showing that e0 (t) ≤ 0 and then use the
fact that e(0) = 0 and e(t) ≥ 0 to conclude that e(t) ≡ 0 for all t such that
0 ≤ t ≤ t0 . We proceed as follows.
1 3
e0 (t) = − [u2t + 3u2x ]|x=x0 +(t0 −t) − [u2t + 3u2x ]|x=x0 −3(t0 −t)
2 2
Z x0 +(t0 −t)
1
+ (2ut utt + 6ux uxt ) dx
2 x0 −3(t0 −t)
1 3
= − [u2t + 3u2x ]|x=x0 +(t0 −t) − [u2t + 3u2x ]|x=x0 −3(t0 −t)
2 2
Z x0 +(t0 −t)
1
+ (2ut utt − 6uxx ut ) dx + 3ux ut |x=x0 +(t0 −t) − 3ux ut |x=x0 −3(t0 −t)
2 x0 −3(t0 −t)
1 3
= − [u2t − 6ux ut + 3u2x ]|x=x0 +(t0 −t) − [u2t + 2ux ut + 3u2x ]|x=x0 −3(t0 −t)
2 2
Z x0 +(t0 −t)
− (u2t )x dx
x0 −3(t0 −t)
1 3 1
= − [3u2t − 6ux ut + 3u2x ]|x=x0 +(t0 −t) − [ u2t + 2ux ut + 3u2x ]|x=x0 −3(t0 −t)
2 2 3
3 1
= − [ut − ux ]2 |x=x0 +(t0 −t) − [u2t + 3ux ]2 |x=x0 −3(t0 −t) ≤ 0.
2 2

5
 Therefore, e0 (t) ≤ 0, which implies ut = 0 = ux within the interval (x0 − 3(t0 −
t), x0 + (t0 − t)). Therefore, u ≡ C for some constant C. But, u(x, 0) ≡ 0 in the
interval (x0 − 3t0 , x0 + t0 ) implies u ≡ 0 in that interval.
(b) Use energy methods to prove uniqueness of solutions to this initial-value problem
if the initial data has compact support.
Answer: We define the energy as
Z
1 ∞ 2
E(t) = (u + 3u2x ) dx.
2 −∞ t

Now assume we have two solutions u, v with the same initial data. Let w = u − v.
Therefore, w satisfies the initial-value problem with zero initial data. Now
Z
0 1 ∞
E (t) = (2wt wtt + 6wx wxt ) dx
2 −∞
Z ∞
= wt wtt − 3uxx wt dx + wx wt |x→+∞
x→−∞ .
−∞
Z ∞
= −2 wt wxt dx
−∞
Z ∞
=− (wt2 )x dx = 0,
−∞

using the fact that if the initial data has compact support, then the solution has
compact support. Therefore, E 0 (t) = 0. Therefore,
Z ∞
(wt2 + 3wx2 ) dx = 0,
−∞

which implies wt = 0 = wx . Using the fact that w(x, 0) = 0, we conclude that

w ≡ 0, and, therefore, u ≡ v.

6. Consider the following eigenvalue problem.

 00
 y + λy = 0,
 0<x<l
0
y (0) + y(0) = 0


y(l) = 0.

(a) Show the boundary conditions are symmetric.

Answer: First,
f 0 (l)g(l) − f (l)g 0 (l) = 0
for any functions f, g satisfying the boundary conditions, because f (l) = 0 = g(l).
Second,
f 0 (0)g(0) − f (0)g 0 (0) = −f (0)g(0) + f (0)g(0) = 0
for any functions satisfying the boundary conditions. Therefore, the boundary
conditions are symmetric.

6
(b) State the definition of orthogonality of functions on [0, l].
Answer: The functions f and g are orthogonal on [0, l] if
Z l
f (x)g(x) dx = 0.
0

(c) Use the fact that the boundary conditions are symmetric to prove all eigenfunc-
tions of this operator must be orthogonal.
Answer: Note: I should say to prove that eigenfunctions corresponding to
distinct eigenvalues are orthogonal. Eigenfunctions corresponding to the same
eigenvalue can be chosen to be orthogonal using a Gram-Schmidt orthogonalization
process.
Let Xm , Xn be two eigenfunctions corresponding to distinct eigenvalues λn 6= λm .
Therefore,
Z l Z l
λn Xn Xm dx = − Xn00 Xm dx
0 0
Z l
= Xn0 Xm 0
dx − Xn0 Xm |x=0
x=l
0
Z l
00 0
=− Xn Xm + (Xn Xm − Xn0 Xm )|x=0
x=l
0
Z l
00
= λm Xn Xm dx,
0

using the fact that the boundary conditions are symmetric. Therefore,
Z l
(λn − λm ) Xn Xm dx = 0.
0

But, λn 6= λm . Therefore, Z l
Xn Xm dx = 0,
0
as claimed.
(d) Find all positive eigenvalues and their corresponding eigenfunctions. (Note: You
may not be able to find an explicit formula for these eigenvalues.) Show graph-
ically that there are an infinite number of positive eigenvalues {λn } such that
λn → +∞.
Answer: Look for positive eigenvalues λ = β 2 > 0. Therefore,
 00 2
Y + β Y = 0

Y 0 (0) + Y (0) = 0

 Y (l) = 0.

Now the general solution of this ODE is given by

Y (y) = C cos(βy) + D sin(βy).

7
 Now Y (0) = C and Y 0 (0) = Dβ. Therefore, the first boundary condition implies
C + Dβ = 0. Further, the second boundary condition implies

Y (l) = C cos(βl) + D sin(βl) = 0.

Therefore, by the first condition, we need

−Dβ cos(βl) + D sin(βl) = 0.

We don’t want D = 0. Therefore, we need

sin(βl) = β cos(βl),

or
tan(βl) = β.
Therefore, the eigenvalues and corresponding eigenfunctions are given by

λn = βn2 where tan(βn l) = βn

Yn (y) = −Dn βn cos(βn y) + D sin(βn y).

7. Consider the following initial/boundary value problem,



 utt − 4uxx = 0 0 < x < l, t > 0


 u(x, 0) = 0 0<x<l
ut (x, 0) = 0 0<x<l



 u(0, t) = sin t

u(l, t) = 1.

Define a function U(x, t) such that by letting v(x, t) = u(x, t) − U (x, t), then v(x, t)
will satisfy 

 vtt − 4vxx = f (x, t) 0 < x < l, t > 0

v(x, 0) = φ(x) 0<x<l

 vt (x, 0) = ψ(x) 0<x<l

v(0, t) = 0 = v(l, t) t>0
for some functions f (x, t), φ(x) and ψ(x), thus, reducing the problem with inhomo-
geneous boundary data to an inhomogeneous problem with Dirichlet boundary data.
You do not need to solve the new inhomogeneous problem.
Answer: Let
1
U(x, t) = ((l − x) sin t + x).
l
8. Consider the initial-value problem for the wave equation in n dimensions,

 utt − ∆u = 0 x ∈ Rn , t > 0
u(x, 0) = φ(x)

ut (x, 0) = ψ(x)

8
 (a) If the initial data is supported in the annular region {a < |x| < b}, find where the
solution is definitely zero in
i. R2
Answer:
|x| + t < a and |x| − t > b.
ii. R3 .
Answer:
|x| + t < a and |x| − t > b and t − |x| > b.
(b) Find the value of the solution u of the initial-value problem

 utt − ∆u = 0 x ∈ R3 , t ≥ 0
u(x, 0) = 0

ut (x, 0) = ψ(x)

where (
1, |x| < a
ψ(x) =
0, |x| > a
at a point (x, t) such that |x| + t < a.
Answer: By Kirchoff’s formula, the solution is given by
Z
1
tψ(y) dS(y).
4πt2 ∂B(x,t)

Now, ψ(y) ≡ 1 for |y| < a. Therefore, if |x| + t < a, then ψ ≡ 1. Therefore, the
solution is given by Z
1
u(x, t) = t dS(y)
4πt2 ∂B(x,t)
or
u(x, t) = t.

9. Let Ω = {(x, y) ∈ R2 : 0 < x < π, 0 < y < π}. Solve the following initial/boundary
value problem.


 utt = uxx + uyy + 1 (x, y) ∈ Ω, t > 0

u(x, y, 0) = sin(x) sin(2y)

 ut (x, y, 0) = 0

u(x, y, t) = 0 (x, y) ∈ ∂Ω.

Answer: First, we will solve the homogeneous problem. Then, we will use Duhamel’s
principle. Using separation of variables, we have

T 00 X 00 Y 00
− =− − = λ,
T X Y

9
which leads us to
X 00 Y 00
− =λ+ = µ.
X Y
Now, first, we consider the eigenvalue problem
− X 00 = µX 0 < x < π. X(0) = 0 = X(π).
The solutions of this eigenvalue problem are given by µn = n2 , Xn (x) = sin(nx). Next,
we solve
Y 00
− =λ−µ 0<y<π
Y
Y (0) = 0 = Y (π).
The solutions of this eigenvalue problem are given by λ − µ = m2 . Therefore, we
conclude that λmn = m2 + n2 and Xn (x)Ym (y) = sin(nx) sin(ny). Solving our equation
for Tmn , we have
p p
Tmn (t) = Amn cos( λmn t) + Bmn sin( λmn t).
Therefore, our solution has the form
X p p
u(x, y, t) = [Amn cos( λmn t) + Bmn sin( λmn t)] sin(nx) sin(my).
m,n

Now u(x, y, 0) = sin(x) sin(2y) implies

(
1 n = 1, m = 2
Amn =
0 otherwise.

Now ut (x, y, 0) = 0 implies Bmn = 0. Therefore, the solution of the homogeneous

problem is given by
p √
u(x, y, t) = cos( λ2,1 t) sin(x) sin(2y) = cos( 5t) sin(x) sin(2y).

Using Duhamel’s principle, we conclude that the inhomogeneous part of the solution
is given by X p
Bmn sin( λmn (t − s)) sin(nx) sin(my)
m,n

where
p Z π Z π
h1, sin(nx) sin(my)i 4
λmn Bmn = = 2 sin(nx) sin(my) dx dy.
hsin(nx) sin(my), sin(nx) sin(my)i π 0 0

Therefore, our solution is given by

√ Z tX p
u(x, y, t) = cos( 5t) sin(x) sin(2y) + Bmn (s) sin( λmn (t − s)) sin(nx) sin(my) ds.
0 m,n

where Bmn (s) is defined above.

10
10. Use Green’s Theorem to show that the value of the solution u at the point (0, t0 ) of
the wave equation on the half-line with Neumann boundary conditions


 utt − c2 uxx = 0 0 < x < ∞, t > 0

 u(x, 0) = φ(x)

 ut (x, 0) = ψ(x)


ux (0, t) = 0
is given by Z ZZ
1 ct0 1
u(0, t0 ) = φ(ct0 ) + ψ(y) dy + f (y, s) dy ds
c 0 c ∆
where ∆ is the triangle in the xt-plane bounded by the lines x = 0, t = 0 and x =
c(t0 − t).
Answer: Note: This should be the inhomogeneous problem! Integrating over ∆, we
have
ZZ ZZ
2
(utt − c uxx ) dx dt = f (x, t) dx dt.
∆

By Green’s Theorem, we have

ZZ Z
2
− [(c ux )x − (ut )t ] dx dt = − [ut dx + c2 ux dt]
∆ Z∂∆ Z
2
=− [ut dx + c ux dt] − [ut dx + c2 ux dt]
L1 L2
Z
− [ut dx + c2 ux dt],
L3

where L1 is the line segment t = 0 from x = 0 to x = ct0 , L2 is the line segment

x = c(t0 − t) from (ct0 , 0) to (0, t0 ) and L3 is the line segment x = 0 from (0, t0 ) to
(0, 0).
Now Z Z ct0 Z ct0
2
− [ut dx + c ux dt] = − ut (x, 0) dx = − ψ(x) dx.
L1 0 0

Z Z t0
2
− [ut dx + c ux dt] = − [−cut (c(t0 − t), t) + c2 ux (c(t0 − t), t)] dt
L2 0
Z t0
=c [ut − cux ] dt
0
Z t0
dx
=c [ut + ux ] dt
dt
Z0 t 0
=c du
0
= c[u(0, t0 ) − u(x0 , 0)]
= cu(0, t0 ) − cφ(ct0 ).

11
Lastly, Z Z 0
2
− [ut dx + c ux dt] = − c2 ux (0, t) dt = 0.
L3 t0

Therefore, we conclude that

Z ct0 ZZ
cu(0, t0 ) = cφ(ct0 ) + ψ(x) dx + f (x, t) dx dt,
0 ∆

which implies
Z ct0 ZZ
1 1
u(0, t0 ) = φ(ct0 ) + ψ(x) dx + f (x, t) dx dt,
c 0 c ∆

as claimed.

12