MATH 4220 (2012-2013) Partial differential equations CUHK

Suggested Answer to Assignment 4

Suggested Answer to Exercise 4.1
Exercise 4.1
2. Consider a metal rod (0 < x < l), insulated along its sides but not at its ends, which is
initially at temperature = 1. Suddenly both ends are plunged into a bath of temperature = 0.
Write the differential equation, boundary conditions, and initial condition. Write the formula
for the temperature u(x, t) at later times. In this problem, assume the infinite series expansion
4 πx 1 3πx 1 5πx
1= (sin + sin + sin + · · · ).
π l 3 l 5 l
Answer: It is not difficult to see the equation to this problem satisfies the following DE

ut = kuxx , (0 < x < l, 0 < t < ∞)

u(0, t) = u(l, t) = 0,
u(x, 0) = 1

Following the process on Page. 85, we get the solution formula is the following
∞
X nπ 2 nπx
u(x, t) = An e−( l
) kt
sin .
n=1
l

By the initial condition, we get

∞
X nπx
1= An sin ,
n=1
l
4 1 −( nπ 2
hence An = π4 n1 for n odd and An = 0 for n even. Then u(x, t) = n=odd e l ) kt sin nπx
P
πn l
.
4. Consider waves in a resistant medium that satisfy the problem

utt = c2 uxx − rut for 0 < x < l

u = 0 at both ends
u(x, 0) = φ(x) ut (x, 0) = ψ(x),

where r is a constant, 0 < r < 2πc/l. Write down the series expansion of the solution.
Answer: Let u(x, t) = T (t)X(x), we get

T 00 + rT 0 X 00
= = −λ.
c2 T X
Hence
nπ 2 nπx
λn = ( ) , X(x) = sin , n = 1, 2, ....
l l
Since 0 < r < 2πc/l, we get
p p
Tn (t) = An cos( −∆n t/2) + Bn sin( −∆n t/2), n = 1, 2, ...,

1
MATH 4220 (2012-2013) Partial differential equations CUHK

where ∆n = r2 − ( 2nπc
l
)2 relative to the equation
nπc 2
λ2 + rλ + ( ) = 0.
l
Therefore ∞
X p p nπx
u(x, t) = [An cos( −∆n t/2) + Bn sin( −∆n t/2)] sin . 
n=1
l
5. Do the same for 2πc/l < r < 4πc/l.
Answer: Let u(x, t) = T (t)X(x). As above, we get

T 00 + rT 0 X 00
= = −λ.
c2 T X
Hence
nπ 2 nπx
λn = (
) , X(x) = sin , n = 1, 2, ....
l l
For n = 1, since 2πc/l < r < 4πc/l,
+ −
T1 (t) = A1 eλ1 t + B1 eλ1 t ,

where q
−r ± r2 − ( 2πc
l
)2
λ±
1 =
2
be the roots of the equation
πc 2
λ2 + rλ + ( ) = 0.
l
For n ≥ 2, as before,
p p
Tn (t) = An cos( −∆n t/2) + Bn sin( −∆n t/2), n = 2, 3, ...,

where ∆n = r2 − ( 2nπc
l
)2 relative to the equation
nπc 2
λ2 + rλ + ( ) = 0.
l
Therefore
∞
λ+
1 t λ−
1 t
πx X p p nπx
u(x, t) = [A1 e + B1 e ] sin + [An cos( −∆n t/2) + Bn sin( −∆n t/2)] sin . 
l n=2
l

6. Separate variables for the equation tut = uxx + 2u with the boundary conditions u(0, t) =
u(π, t) = 0. Show that there are an infinite number of solutions that satisfy the initial condition
u(x, 0) = 0. So uniqueness is false for this equation!
Answer: Let u(x, t) = T (t)X(x), we get

tT 0 − 2T X 00
= = −λ.
T X

2
MATH 4220 (2012-2013) Partial differential equations CUHK

As before we have
λn = n2 , Xn (x) = sin nx, n = 1, 2, 3, ....
By the initial condition, we have

tT 0 − 2T = −λT, T (0) = 0.

Therefore
u(x, t) = ct sin x, for any constant c,
are solutions (You can check them directly). So uniqueness is false for this equation! 
Suggested Answer to Exercise 4.2
1. Solve the diffusion problem ut = kuxx in 0 < x < l, with the mixed boundary conditions
u(0, t) = ux (l, t) = 0.
Answer: Let u(x, t) = T (t)X(x), we get

T0 X 00
= = −λ.
kT X
By the boundary condition, we have the following eigenvalue problem

−X 00 = λX, X(0) = X 0 (l) = 0.

Since the eigenvalues be [(n + 12 )π]2 /l2 and the eigenfunctions be sin[(n + 12 )πx/l] for n =
0, 1, 2, ..., then we have
∞
X
−[
1 )π
(n+ 2
]2 kt (n + 12 )πx
u(x, t) = An e l sin . 
n=0
l

2. Consider the equation utt = c2 uxx for 0 < x < l, with the boundary conditions ux (0, t) =
0, u(l, t) = 0 (Neumann at the left, Dirichlet at the right).
(a) Show that the eigenfunctions are cos[(n + 21 )πx/l].
(b) Write the series expansion for a solution u(x, t).
Answer: (a) This can be proved as before (Please see it in the textbook). Here we give
another proof.
Since X 0 (0) = 0, by the even expansion X(−x) = X(x) for −l ≤ x < 0, then X satisfies

−X 00 = λX, X(−l) = X(l) = 0.

Hence
1 1
λn = [(n + )π]2 /l2 , Xn (x) = cos[(n + )πx/l], n = 0, 1, 2, ....
2 2
(b) As before, using (a) we have
∞ h
X (n + 12 )πct (n + 12 )πct i (n + 12 )πx
u(x, t) = An cos + Bn sin cos .
n=0
l l l

3
MATH 4220 (2012-2013) Partial differential equations CUHK

3. Consider diffusion inside an enclosed circular tube. Let its length (circumference) be
2l. Let x denote the arc length parameter where −l ≤ x ≤ l. Then the concentration of the
diffusing substance satisfies

ut = kuxx for − l ≤ x ≤ l
u(−l, t) = u(l, t) and ux (−l, t) = ux (l, t).

These are called periodic boundary conditions.

(a) Show that the eigenvalues are λ = (nπ/l)2 for n = 0, 1, 2, 3, . . . .
(b) Show that the concentration is
∞ 
1 X nπx nπx  −n2 π2 kt/l2
u(x, t) = a0 + an cos + bn sin e .
2 n=1
l l

Answer: (a) As before, let λ be any complex number and γ be either one of the two square
roots of −λ; the other one is −γ. Then

X(x) = Ceγx + De−γx .

The boundary conditions yield

Ce−γl + Deγl = Ceγl + De−γl , and Cγe−γl − Dγeγl = Cγeγl − Dγe−γl .

Hence e2γl = 1 and then

γ = ±nπi/l and λ = −γ 2 = (nπ/l)2 , n = 0, 1, 2, . . . .

(b) As before, by (a) we get

∞ h
1 X nπx nπx i −( nπ )2 kt
u(x, t) = a0 + an cos + bn sin e l .
2 n=1
l l

Suggested Answer to Exercise 4.3

1. Find the eigenvalues graphically for the boundary conditions

X(0) = 0, X 0 (l) + aX(l) = 0.

Assume that a 6= 0.
Answer: Firstly, let’s look for the positive eigenvalues λ = β 2 > 0. As usual, the general
solution of the ODE is
X(x) = C cos βx + D sin βx.
By the boundary conditions,

C = 0, and Dβ cos βl + aD sin βl = 0.

Hence tan βl = − βa . The graph is omitted.

4
MATH 4220 (2012-2013) Partial differential equations CUHK

Secondly, let’s look for the zero eigenvalue, i.e., X(x) = Ax+B, by the boundary conditions,
al + 1 = 0. Hence λ = 0 is an eigenvalue if and only if al + 1 = 0.
Thirdly, let’s look for the negative eigenvalues λ = −γ 2 < 0. As usual, the solution of the
ODE is
X(x) = C cosh γx + D sinh γx.
By the boundary conditions,

C = 0, and Dγ cosh γl + aD sinh γl = 0.

Hence tanh γl = − γa and the graph is also omitted. 

2. Consider the eigenvalue problem with Robin BCs at both ends:

−X 00 = λX
X 0 (0) − a0 X(0) = 0, X 0 (l) + al X(l) = 0.

(a) Show that λ = 0 is an eigenvalue if and only if a0 + al = −a0 al l.

(b) Find the eigenfunctions corresponding to the zero eigenvalue. (Hint: First solve the
ODE for X(x). The solutions are not sines or cosines.)
Answer: (a) If λ = 0, then X(x) = Ax + B. By the boundary conditions, we get

A − a0 B = 0, and A + al (Al + B) = 0,

which is equivalent to
a0 + al = −a0 al l.
Therefore λ = 0 is an eigenvalue if and only if a0 + al = −a0 al l.
(b) By (a), we have X(x) = B(a0 x + 1), here B is constant. 
3. Derive the eigenvalue equation (16) for the negative eigenvalues λ = −γ 2 and the formula
(17) for the eigenfunctions.
Answer: If λ = −γ 2 < 0, we have

X(x) = C cosh γx + D sinh γx.

Thus
X 0 (x) = Cγ sinh γx + Dγ cosh γx.
Hence by the boundary conditions,

Dγ − a0 C = 0, and Cγ sinh γl + Dγ cosh γl + al [C cosh γl + D sinh γl] = 0,

which implies
(a0 + al )γ
tanh γl = − .
γ 2 + a0 al
And the corresponding eigenfunction
a0
X(x) = C cosh γx + C sinh γx,
γ
where C is constant. 

5
MATH 4220 (2012-2013) Partial differential equations CUHK

4. Consider the Robin eigenvalue problem. If

a0 < 0, al < 0 and − a0 − al < a0 al l,
show that there are two negative eigenvalues. This case may be called “substantial absorption
at both ends.” (Hint: Show that the rational curve y = (a0 + al )γ/(γ 2 + a0 al ) has a single
maximum and crosses the line y = 1 in two places. Deduce that it crosses the tanh curve in
two places.)
Answer: It is easily known that the rational curve y = − (a 0 +al )γ
γ 2 +a
has a single maximum at
√ √ √0 al
γ = a0 al and is monotonic in the two intervals [0, a0 al ) and [ a0 al , ∞). Furthermore,
(a0 + al ) a0 + al
max y(γ) = − √ > 1, lim y = 0, and y 0 (0) = − .
γ∈[0,∞) 2 a0 al γ→∞ a0 al
Note that tanh γl is monotonic in [0, ∞),
a0 + al
tanh(γl) < 1 in [0, ∞), lim tanh γl = 1, and(tanh γl)0 |γ=0 = l > − .
γ→∞ a0 al
Therefore the rational curve y = − (a 0 +al )γ
γ 2 +a0 al
have two common points with the curve y = tanh γl
and then there are two negative eigenvalues. 
5. In Exercise 4 (substantial absorption at both ends) show graphically that there are an
infinite number of positive eigenvalues. Show graphically that they satisfy (11) and (12).
Answer: Since
(a0 + al )β [(−a0 ) + (−al )]β
= −
β 2 − a0 al β 2 − (−a0 )(−al )
and −a0 > 0, −al > 0, −a0 − al < a0 al l, the graph is similar to the Figure 1 in Section 4.3 in
the textbook. 
6. If a0 = al = a in the Robin problem, show that:
(a) There are no negative eigenvalues if a ≥ 0, there is one if −2/l ≤ a < 0, and there are
two if a < −2/l.
(b) Zero is an eigenvalue if and only if a = 0 or a = −2/l.
Answer: (a) If a > 0, this is case 1 and there is only positive eigenvalue. And if a = 0,
this is the Neumann condition. Hence there are only zero and positive eigenvalues if a ≥ 0.
By the result of exercise (4), we known that there is one negative eigenvalue if −2/l ≤ a < 0,
and there are two if a < −2/l.
(b) By the result of exercise (2), we known that λ = 0 is an eigenvalue if and only if
a0 + al = −a0 al l, which is equivalent to a = 0 or a = −2/l. 
7. If a0 = al = a, show that as a → +∞, the eigenvalues tend to the eigenvalues of the
Dirichlet problem. That is,
(n + 1)π 2
lim {λn (a) − [ ] } = 0,
a→∞ l
where λn (a) is the (n + 1)st eigenvalue.
Answer: Under the condition of a0 = al = a, the eigenvalue equation
(a0 + al )β
tan βl = ,
β 2 − a0 al

6
MATH 4220 (2012-2013) Partial differential equations CUHK

is changed into the equation

2aβ
tan βl = .
β 2 − a2
Thus
βl → nπ as a → +∞.
That is, (  2 )
(n + 1)π
lim λn (a) − = 0. 
a→∞ l
9. On the interval 0 ≤ x ≤ 1 of length one, consider the eigenvalue problem

−X 00 = λX, X 0 (0) + X(0) = 0 and X(1) = 0

(absorption at one end and zero at the other).

(a) Find an eigenfunction with eigenvalue zero. Call it X0 (x).
(b) Find an equation for the positive eigenvalues λ = β 2 .
(c) show graphically from part (b) that there are an infinite number of positive eigenvalues.
(d) Is there a negative eigenvalue?
Answer: (a) If λ = 0, then X(x) = ax + b for some constants a, b, and we can easily obtain
a + b = 0 by X 0 (0) + X(0) = 0 and X(1) = 0. Therefore, X0 (x) = ax − a for some nonzero
constant a.
(b) If λ = β 2 , then X(x) = a cos βx + b sin βx. We have

a + bβ = 0, a cos β + b sin β = 0.

The only possibility of the existence of the nontrivial a, b is that β = tan β.

(c) omit.
(d) If λ = −β 2 , then X(x) = aeβx + be−βx , and we have

a + b + aβ − bβ = 0, aeβ + be−β = 0.

Then we find the coefficient matrix is always nonsingular, therefore a = b = 0. We conclude

there isn’t any negative eigenvalue.
10. Solve the wave equation with Robin boundary conditions under the assumption that
(18) holds
Answer: Let u(x, t) = X(x)T (t), by the summary on Page 96. We can get
∞
X a0 a0
u(x, t) = (cn cos cβn t+dn sin cβn t)(cos βn x+ sin βn x)+(c0 cosh cγt+d0 sinh cγt)(cosh γx+ sinh γx),
n=1
β γ

where γ is determined by the intersection point of tanh γl = − (a 0 +al )γ

γ 2 +a0 al
. Then, using the initial
condition,
∞
X a0 a0
φ(x) = cn (cos βn x + sin βn x) + c0 (cosh γx + sinh γx),
n=1
β γ
∞
X a0 a0
ψ(x) = cβn dn (cos βn x + sin βn x) + cγd0 (cosh γx + sinh γx).
n=1
β γ

7
MATH 4220 (2012-2013) Partial differential equations CUHK

We can decide the value of ci , di , i = 0, 1, 2, · · · .

11. (a) Prove that the (total) energy is conserved for the wave equation with Dirichlet BCs,
where the energy is defined to be

1 l −2 2
Z
E= (c ut + u2x ) dx.
2 0
(Compare this definition with Section 2.2.)
(b) Do the same for the Neumann BCs.
(c) For the Robin BCs, show that

1 l −2 2
Z
1 1
ER = (c ut + u2x ) dx + al [u(l, t)]2 + a0 [u(0, t)]2
2 0 2 2
is conserved. Thus, while the total energy ER is still a constant, some of the internal energy is
“lost” to the boundary is a0 and al are positive and “gained” from the boundary if a0 and al
are negative.
Answer: (a) By the wave equation,
Z l
dE 1
= [ 2 ut utt + ux uxt ] dx
dt 0 c
Z l
= [ut uxx + ux uxt ] dx
0
= (ut ux )|l0 = ut (l, t)ux (l, t) − ut (0, t)ux (0, t).

By the Dirichlet boundary conditions u(l, t) = u(0, t) = 0, ut (l, t) = ut (0, t) = 0. Thus dE

dt
≡ 0.
(b) The proof is the same as above. Here we omit it.
(c) By the computation in (a) and the Robin boundary conditions, we can get that
dER
= ut ux |l0 + al ut (l, t)u(l, t) + a0 ut (0, t)ux (0, t) ≡ 0. 
dt
12. Consider the unusual eigenvalue problem

−vxx = λv for 0 < x < l

v(l) − v(0)
vx (0) = vx (l) = .
l
(a) Show that λ = 0 is a double eigenvalue.
(b) Get an equation√ for the positive eigenvalues λ > 0.
1
(c) Letting γ = 2 l λ, reduce the equation in part (b) to the equation

γ sin γ cos γ = sin2 γ.

(d) Use part (c) to find half of the eigenvalues explicitly and half of them graphically.
(e) Assuming that all the eigenvalues are nonnegative, make a list of all the eigenfunctions.
(f) Solve the problem ut = kuxx for 0 < x < l, with the BCs given above, and with
u(x, 0) = φ(x).

8
MATH 4220 (2012-2013) Partial differential equations CUHK

(g) Show that, as t → ∞, lim u(x, t) = A + Bx, assuming that you can take limits term by
term.
Answer: (a) Let λ = 0, we get −vxx = 0, which implies v(x) = Ax+B. Since v(x) = Ax+B
satisfy the boundary conditions for any numbers A and B, so λ = 0 is a double eigenvalue.
(b) Let λ = β 2 > 0 and suppose β > 0, −vxx = λv, then v(x) = C cos βx + D sin βx. Hence
by the boundary conditions, we have
C cos βl + D sin βl − C
Dβ = −Cβ sin βl + Dβ cos βl = .
l
Hence λ = β 2 , where β is a root of the following equation

sin(βl)(− sin βl + βl) = (1 − cos βl)2 .

√
(c) Let γ = 21 l λ, then γ is a root of the following equation

γ sin γ cos γ = sin2 γ.

2 2
(d) By (c), we have sin γ = 0 or γ = tan γ. Then λ1 = 0, λn = n l2π , for n = 2, 4, . . . , and
λn = 4γn2 /l2 , where γn = tan γn , for n = 3, 5, 7, . . . .
2 2
(e) By (a) and (d), for λ = 0, the eigenfunctions are 1 and x; for λn = n l2π , n = 2, 4, 6, ...,
the eigenfunctions are cos(nπx/l); for λn = 4γn2 /l2 , where γn = tan γn , n = 3, 5, 7, . . . , the
eigenfunctions are p p p
l λn cos( λn x) − 2 sin( λn x).
(f) From above, we have
X hp p p i
u(x, t) = A + Bx + An e−λn kt l λn cos( λn x) − 2 sin( λn x)
n>1 and odd
X
+ Bn e−λn kt cos(nπx/l).
n>1 and even

(g) By(f), we have limt→∞ u(x, t) = A + Bx since lim e−λn kt = 0. 

t→∞
15. Find the equation for the eigenvalues λ of the problem

(κ(x)X 0 )0 + λρ(x)X = 0 for 0 < x < l with X(0) = X(l) = 0,

where κ(x) = κ21 for x < a, κ(x) = κ22 for x > a, ρ(x) = ρ21 for x < a, and ρ(x) = ρ22 for x > a.
All these constants are positive and 0 < a < l.
Answer: Let λ = β 2 , then
βρ1 x βρ1 x
X(x) = A cos + B sin , 0 < x < a;
k1 k1
βρ2 x βρ2 x
X(x) = C cos + D sin , a < x < l.
k2 k2
Hence by the boundary conditions,
βρ2 l βρ2 l
A = 0, C cos + D sin = 0;
k2 k2

9
MATH 4220 (2012-2013) Partial differential equations CUHK

βρ1 a βρ2 a βρ2 a

B sin = C cos + D sin ;
k1 k2 k2
βρ1 βρ1 a βρ2 βρ2 a βρ2 βρ2 a
and B cos = −C sin +D cos .
k1 k1 k2 k2 k2 k2
Therefore λ = β 2 , where β is a root of the equation
ρ1 βρ1 a ρ2 βρ2 (l − a)
cot + cot = 0.
k1 k1 k2 k2
Let λ = 0, then we have

X(x) = Ax, 0 < a < l, X(x) = B(x − l), a < x < l.

Then by using the boundary condition and X(x) and with its first derivative should be contin-
uous at x = a. Such A and B can not exist except A = B = 0.
Let λ = −β 2 , then we have
βρ1 x βρ1 x
X(x) = A cosh + B sinh , 0 < x < a,
k1 k1
βρ2 x βρ2 x
X(x) = C cosh + D sinh , a < x < l.
k2 k2
By using the boundary condition and X(x) and with its first derivative should be continuous
at x = a. We have
βρ2 l βρ2 l
A = 0, C cosh + D sinh ,
k2 k2
βρ1 a βρ2 a βρ2 a
B sinh = C cosh + D sinh ,
k1 k2 k2
βρ1 βρ1 a βρ2 βρ2 a βρ2 βρ2 a
B cosh =C sinh +D cosh .
k1 k1 k2 k2 k2 k2
After computations, we find β is a root of the following equation
ρ1 βρ1 a ρ2 βρ2 (l − a)
coth + coth = 0.
k1 k1 k2 k2
However, since the left handside is always positive unless β = 0. Therefore, it has no negative
eigenvalues.
16. Find the positive eigenvalues and the corresponding eigenfunctions of the fourth-order
operator +d4 /dx4 with the four boundary conditions

X(0) = X(l) = X 00 (0) = X 00 (l) = 0.

Answer: Let λ = β 4 > 0, where β > 0 and X(x) = A cosh(βx) + B sinh(βx) + C cos(βx) +
D sin(βx).
Hence by the boundary conditions,
nπ nπ nπx
β= , λ = ( )4 , Xn (x) = sin( ), n = 1, 2, . . . . 
l l l
10
MATH 4220 (2012-2013) Partial differential equations CUHK

17. Solve the fourth-order eigenvalue problem X 0000 = λX in 0 < x < l, with the four
boundary conditions
X(0) = X 0 (0) = X(l) = X 0 (l) = 0,
where λ > 0. (Hint: First solve the fourth-order ODE.)
Answer: Let λ = β 4 > 0 where β > 0, and X(x) = A cosh βx + B sinh βx + C cos(βx) +
D sin(βx). Hence by the boundary conditions,
A + C = 0, B + D = 0,
A cosh βl + B sinh βl + C cos(βl) + D sin(βl) = 0,
and A sinh βl + B cosh βl − C sin(βl) + D cos(βl) = 0.
4
Hence λ = β , where β is a root of the equation
cosh βl cos βl = 1,
and the corresponding eigenfunction is
X(x) = (sinh βl − sin βl)(cosh βx − cos βx) − (cosh βl − cos βl)(sinh βx − sin βx). 
Suggested Answer to Extra Problems Problem 10. Solve the following
problem by using separation of variables.
utt + a2 uxxxx = 0, 0 < x < l, t > 0,
u(x, 0) = φ(x), ut (x, 0) = ψ(x),
u(0, t) = uxx (0, t) = u(l, t) = uxx (l, t) = 0.
00 0000
Answer: Let u = X(x)T (t), then −aT 2 T = XX = λ.
Let λ = 0, X(x) = ax3 + bx2 + cx + d, using the boundary condition, we obtain a = b =
c = d = 0. Hence zero can not be an eigenvalue.
Let λ = β 4 , X(x) = A cosh βx + B sinh βx + C cos βx + D sin βx.
Following a similar process as in (17). If we want to obtain a nontrivial solution, we must
have l = kπ
β
. The corresponding eigenfunction is
kπ
X(x) = D sin .
l
Function T (t) suffices the following equation
T (t) = C cos aβ 2 t + D sin aβ 2 t.
Using integration by parts, we can exactly prove this eigenvalue problem does not admit
negative eigenvalue. Therefore the solution to this problem is the following
∞
X nπx
u(x, t) = (Cn cos aβn2 t + Dn sin aβn2 t) sin .
n=1
l
Using the initial condition
∞ ∞
X nπx X nπx
φ(x) = Cn sin , ψ(x) = aβn2 Dn sin .
n=1
l n=1
l
We can decide the Cn , Dn , n = 1, 2, 3, · · · .

11