You are on page 1of 38

14

14
Boundary-Value Problems in
Other Coordinate Systems
EXERCISES 14.1
Problems in Polar Coordinates
1. We have
A
0
=
1
2
_

0
u
0
d =
u
0
2
A
n
=
1

_

0
u
0
cos n d = 0
B
n
=
1

_

0
u
0
sin n d =
u
0
n
[1 (1)
n
]
and so
u(r, ) =
u
0
2
+
u
0

n=1
1 (1)
n
n
r
n
sin n.
2. We have
A
0
=
1
2
_

0
d +
1
2
_
2

( ) d = 0
A
n
=
1

_

0
cos n d +
1

_
2

( ) cos n d =
2
n
2

[(1)
n
1]
B
n
=
1

_

0
sin n d +
1

_
2

( ) sin n d =
1
n
[1 (1)
n
]
and so
u(r, ) =

n=1
r
n
_
(1)
n
1
n
2

cos n +
1 (1)
n
n
sin n
_
.
3. We have
A
0
=
1
2
_
2
0
(2
2
) d =
2
2
3
A
n
=
1

_
2
0
(2
2
) cos n d =
4
n
2
B
n
=
1

_
2
0
(2
2
) sin n d = 0
and so
u(r, ) =
2
2
3
4

n=1
r
n
n
2
cos n.
755
14.1 Problems in Polar Coordinates
4. We have
A
0
=
1
2
_
2
0
d =
A
n
=
1

_
2
0
cos n d = 0
B
n
=
1

_
2
0
sin n d =
2
n
and so
u(r, ) = 2

n=1
r
n
n
sin n.
5. As in Example 1 in the text we have R(r) = c
3
r
n
+c
4
r
n
. In order that the solution be bounded as r we
must dene c
3
= 0. Hence
u(r, ) = A
0
+

n=1
r
n
(A
n
cos n +B
n
sin n)
A
0
=
1
2
_
2
0
f() d where
A
n
=
c
n

_
2
0
f() cos n d
B
n
=
c
n

_
2
0
f() sin n d.
6. Using the same reasoning as in Example 1 in the text we obtain
u(r, ) = A
0
+

n=1
r
n
(A
n
cos n +B
n
sin n).
The boundary condition at r = c implies
f() =

n=1
nc
n1
(A
n
cos n +B
n
sin n).
Since this condition does not determine A
0
, it is an arbitrary constant. However, to be a full Fourier series on
[0, 2] we must require that f() satisfy the condition A
0
= a
0
/2 = 0 or
_
2
0
f() d = 0. If this integral were
not 0, then the series for f() would contain a nonzero constant, which it obviously does not. With this as a
necessary compatibility condition we can then make the identications
nc
n1
A
n
= a
n
and nc
n1
B
n
= b
n
or
A
n
=
1
nc
n1

_
2
0
f() cos n d and B
n
=
1
nc
n1

_
2
0
f() sin n d.
7. Proceeding as in Example 1 in the text and again using the periodicity of u(r, ), we have
() = c
1
cos +c
2
sin
where = n for n = 0, 1, 2, . . . . Then
R(r) = c
3
r
n
+c
4
r
n
.
756
14.1 Problems in Polar Coordinates
[We do not have c
4
= 0 in this case since 0 < a r.] Since u(b, ) = 0 we have
u(r, ) = A
0
ln
r
b
+

n=1
__
b
r
_
n

_
r
b
_
n
_
[A
n
cos n +B
n
sin n] .
From
u(a, ) = f() = A
0
ln
a
b
+

n=1
__
b
a
_
n

_
a
b
_
n
_
[A
n
cos n +B
n
sin n]
we nd
A
0
ln
a
b
=
1
2
_
2
0
f() d,
__
b
a
_
n

_
a
b
_
n
_
A
n
=
1

_
2
0
f() cos n d,
and
__
b
a
_
n

_
a
b
_
n
_
B
n
=
1

_
2
0
f() sin n d.
8. Substituting u(r, ) = v(r, ) +(r) into the partial dierential equation we obtain

2
v
r
2
+

(r) +
1
r
_
v
r
+

(r)
_
+
1
r
2

2
v

2
= 0.
This equation will be homogeneous provided

(r) +
1
r

(r) = 0 or r
2

(r) +r

(r) = 0.
The general solution of this Cauchy-Euler dierential equation is
(r) = c
1
+c
2
ln r.
From
u
0
= u(a, ) = v(a, ) +(a) and u
1
= u(b, ) = v(b, ) +(b)
we see that in order for the boundary values v(a, ) and v(b, ) to be 0 we need (a) = u
0
and (b) = u
1
. From
this we have
(a) = c
1
+c
2
ln a = u
0
(b) = c
1
+c
2
ln b = u
1
.
Solving for c
1
and c
2
we obtain
c
1
=
u
1
ln a u
0
ln b
ln(a/b)
and c
2
=
u
0
u
1
ln(a/b)
.
Then
(r) =
u
1
ln a u
0
ln b
ln(a/b)
+
u
0
u
1
ln(a/b)
ln r =
u
0
ln(r/b) u
1
ln(r/a)
ln(a/b)
.
From Problem 7 with f() = 0 we see that the solution of

2
v
r
2
+
1
r
v
r
+
1
r
2

2
v

2
= 0, 0 < < 2, a < r < b,
v(a, ) = 0, v(b, ) = 0, 0 < < 2
is v(r, ) = 0. Thus the steady-state temperature of the ring is
u(r, ) = v(r, ) +(r) =
u
0
ln(r/b) u
1
ln(r/a)
ln(a/b)
.
757
14.1 Problems in Polar Coordinates
9. This is similar to the solution to Problem 7 above. When n = 0, () = c
5
+ c
6
and R(r) = c
7
+ c
8
ln r.
Periodicity implies c
5
= 0 and the insulation condition at r = a implies c
8
= 0. Thus, we take u
0
= A
0
= c
6
c
7
.
Then, for n = 1, 2, 3, . . . , () = c
1
cos n + c
2
sin n and R(r) = c
3
r
n
+ c
4
r
n
. From R

(a) = 0 we get
c
3
na
n1
c
4
na
n1
= 0, which implies c
4
= c
3
a
2n
. Then
R(r) = c
3
(r
n
+a
2n
r
n
) = c
3
r
2n
+a
2n
r
n
and
u(r, ) = A
0
+

n=1
r
2n
+a
2n
r
n
(A
n
cos n +B
n
sin n).
Taking r = b we have
f() = A
0
+

n=1
b
2n
+a
2n
b
n
(A
n
cos n +B
n
sin n),
which implies
A
0
=
a
0
2
=
1
2
_
2
0
f() d
and
b
2n
+a
2n
b
n
A
n
=
1

_
2
0
f() cos n d and
b
2n
+a
2n
b
n
B
n
=
1

_
2
0
f() sin 2n d.
Hence
A
n
=
b
n
(a
2n
+b
2n
)
_
2
0
f() cos n d and B
n
=
b
n
(a
2n
+b
2n
)
_
2
0
f() sin n d.
10. We solve

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0, 0 < <

2
, 0 < r < c,
u(c, ) = f(), 0 < <

2
,
u(r, 0) = 0, u(r, /2) = 0, 0 < r < c.
Proceeding as in Example 1 in the text we obtain the separated dierential equations
r
2
R

+rR

R = 0

+ = 0.
Taking =
2
the solutions are
() = c
1
cos +c
2
sin
R(r) = c
3
r

+c
4
r

.
Since we want R(r) to be bounded as r 0 we require c
4
= 0. Applying the boundary conditions (0) = 0
and (/2) = 0 we nd that c
1
= 0 and = 2n for n = 1, 2, 3, . . . . Therefore
u(r, ) =

n=1
A
n
r
2n
sin 2n.
From
u(c, ) = f() =

n=1
A
n
c
n
sin 2n
we nd
A
n
=
4
c
2n
_
/2
0
f() sin 2n d.
758
14.1 Problems in Polar Coordinates
11. Referring to the solution of Problem 10 above we have
() = c
1
cos +c
2
sin
R(r) = c
3
r

.
Applying the boundary conditions

(0) = 0 and

(/2) = 0 we nd that c
2
= 0 and = 2n for
n = 0, 1, 2, . . . . Therefore
u(r, ) = A
0
+

n=1
A
n
r
2n
cos 2n.
From
u(c, ) =
_
1, 0 < < /4
0, /4 < < /2
= A
0
+

n=1
A
n
c
2n
cos 2n
we nd
A
0
=
1
/2
_
/4
0
d =
1
2
and
c
2n
A
n
=
2
/2
_
/4
0
cos 2n d =
2
n
sin
n
2
.
Thus
u(r, ) =
1
2
+
2

n=1
1
n
sin
n
2
_
r
c
_
2n
cos 2n.
12. We solve

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0, 0 < < /4, r > 0
u(r, 0) = 0, r > 0
u(r, /4) = 30, r > 0.
Proceeding as in Example 1 in the text we nd the separated ordinary dierential equations to be
r
2
R

+rR

R = 0

+ = 0.
With =
2
> 0 the corresponding general solutions are
R(r) = c
1
r

+c
2
r

() = c
3
cos +c
4
sin .
The condition (0) = 0 implies c
3
= 0 so that = c
4
sin . Now, in order that the temperature be bounded
as r we dene c
1
= 0. Similarly, in order that the temperature be bounded as r 0 we are forced to
dene c
2
= 0. Thus R(r) = 0 and so no nontrivial solution exists for > 0. For = 0 the separated dierential
equations are
r
2
R

+rR

= 0 and

= 0.
Solutions of these latter equations are
R(r) = c
1
+c
2
ln r and () = c
3
+c
4
.
(0) = 0 still implies c
4
= 0, whereas boundedness as r 0 demands c
2
= 0. Thus, a product solution is
u = c
1
c
3
= A.
759
14.1 Problems in Polar Coordinates
From u(r, /4) = 0 we obtain A = 120/. Thus, a solution to the problem is
u(r, ) =
120

.
13. We solve

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0, 0 < < , a < r < b,
u(a, ) = ( ), u(b, ) = 0, 0 < < ,
u(r, 0) = 0, u(r, ) = 0, a < r < b.
Proceeding as in Example 1 in the text we obtain the separated dierential equations
r
2
R

+rR

R = 0

+ = 0.
Taking =
2
the solutions are
() = c
1
cos +c
2
sin
R(r) = c
3
r

+c
4
r

.
Applying the boundary conditions (0) = 0 and () = 0 we nd that c
1
= 0 and = n for n = 1, 2, 3, . . . .
The boundary condition R(b) = 0 gives
c
3
b
n
+c
4
b
n
= 0 and c
4
= c
3
b
2n
.
Then
R(r) = c
3
_
r
n

b
2n
r
n
_
= c
3
_
r
2n
b
2n
r
n
_
and
u(r, ) =

n=1
A
n
_
r
2n
b
2n
r
n
_
sin n.
From
u(a, ) = ( ) =

n=1
A
n
_
a
2n
b
2n
a
n
_
sin n
we nd
A
n
_
a
2n
b
2n
a
n
_
=
2

_

0
(
2
) sin n d =
4
n
3

[1 (1)
n
].
Thus
u(r, ) =
4

n=1
1 (1)
n
n
3
r
2n
b
2n
a
2n
b
2n
_
a
r
_
n
sin n.
14. Letting u(r, ) = v(r, ) + () we obtain

() = 0 and so () = c
1
+ c
2
. From (0) = 0 and () = u
0
we nd, in turn, c
2
= 0 and c
1
= u
0
/. Therefore () =
u
0

. Now u(1, ) = v(1, ) + () so that


v(1, ) = u
0

u
0

. From
v(r, ) =

n=1
A
n
r
n
sin n and v(1, ) =

n=1
A
n
sin n
we obtain
A
n
=
2

_

0
_
u
0

u
0


_
sin n d =
2u
0
n
.
760
14.1 Problems in Polar Coordinates
Thus
u(r, ) =
u
0

+
2u
0

n=1
r
n
n
sin n.
15. We solve

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0, 0 < < , 0 < r < 2,
u(2, ) =
_
u
0
, 0 < < /2
0, /2 < <
u

=0
= 0,
u

=
= 0, 0 < r < 2.
Proceeding as in Example 1 in the text we obtain the separated dierential equations
r
2
R

+rR

R = 0

+ = 0.
Taking =
2
the solutions are
() = c
1
cos +c
2
sin
R(r) = c
3
r

+c
4
r

.
Applying the boundary conditions

(0) = 0 and

() = 0 we nd that c
2
= 0 and = n for n = 0, 1, 2, . . . .
Since we want R(r) to be bounded as r 0 we require c
4
= 0. Thus
u(r, ) = A
0
+

n=1
A
n
r
n
cos n.
From
u(2, ) =
_
u
0
, 0 < < /2
0, /2 < <
= A
0
+

n=1
A
n
2
n
cos n
we nd
A
0
=
1
2
2

_
/2
0
u
0
d =
u
0
2
and
2
n
A
n
=
2u
0

_
/2
0
cos n d =
2u
0

sin n/2
n
.
Therefore
u(r, ) =
u
0
2
+
2u
0

n=1
1
n
_
sin
n
2
__
r
2
_
n
cos n.
16. (a) From Problem 1 in this section, with u
0
= 100,
u(r, ) = 50 +
100

n=1
1 (1)
n
n
r
n
sin n.
761
1 2 3 4 5 6
q
20
40
60
80
100
u
r=0.9
r=0.7
r=0.5
r=0.3
r=0.1
14.1 Problems in Polar Coordinates
(b)
(c) We could use S
5
from part (b) of this problem to compute the approximations, but in a CAS it is just as
easy to compute the sum with a much larger number of terms, thereby getting greater accuracy. In this
case we use partial sums including the term with r
99
to nd
u(0.9, 1.3) 96.5268 u(0.9, 2 1.3) 3.4731
u(0.7, 2) 87.871 u(0.7, 2 2) 12.129
u(0.5, 3.5) 36.0744 u(0.5, 2 3.5) 63.9256
u(0.3, 4) 35.2674 u(0.3, 2 4) 64.7326
u(0.1, 5.5) 45.4934 u(0.1, 2 5.5) 54.5066
(d) At the center of the plate u(0, 0) = 50. From the graphs in part (b) we observe that the solution curves
are symmetric about the point (, 50). In part (c) we observe that the horizontal pairs add up to 100, and
hence average 50. This is consistent with the observation about part (b), so it is appropriate to say the
average temperature in the plate is 50

.
17. Let u
1
be the solution of the boundary-value problem

2
u
1
r
2
+
1
r
u
1
r
+
1
r
2

2
u
1

2
= 0, 0 < < 2, a < r < b
u
1
(a, ) = f(), 0 < < 2
u
1
(b, ) = 0, 0 < < 2,
and let u
2
be the solution of the boundary-value problem

2
u
2
r
2
+
1
r
u
2
r
+
1
r
2

2
u
2

2
= 0, 0 < < 2, a < r < b
u
2
(a, ) = 0, 0 < < 2
u
2
(b, ) = g(), 0 < < 2.
762
14.2 Problems in Cylindrical Coordinates
Each of these problems can be solved using the methods shown in Problem 7 of this section. Now if u(r, ) =
u
1
(r, ) +u
2
(r, ), then
u(a, ) = u
1
(a, ) +u
2
(a, ) = f()
u(b, ) = u
1
(b, ) +u
2
(b, ) = g()
and u(r, ) will be the steady-state temperature of the circular ring with boundary conditions u(a, ) = f()
and u(b, ) = g().
EXERCISES 14.2
Problems in Cylindrical Coordinates
1. Referring to the solution of Example 1 in the text we have
R(r) = c
1
J
0
(
n
r) and T(t) = c
3
cos a
n
t +c
4
sin a
n
t
where the
n
are the positive roots of J
0
(c) = 0. Now, the initial condition u(r, 0) = R(r)T(0) = 0 implies
T(0) = 0 and so c
3
= 0. Thus
u(r, t) =

n=1
A
n
sin a
n
tJ
0
(
n
r) and
u
t
=

n=1
a
n
A
n
cos a
n
tJ
0
(
n
r).
From
u
t

t=0
= 1 =

n=1
a
n
A
n
J
0
(
n
r)
we nd
a
n
A
n
=
2
c
2
J
2
1
(
n
c)
_
c
0
rJ
0
(
n
r) dr x =
n
r, dx =
n
dr
=
2
c
2
J
2
1
(
n
c)
_
nc
0
1

2
n
xJ
0
(x) dx
=
2
c
2
J
2
1
(
n
c)
_
nc
0
1

2
n
d
dx
[xJ
1
(x)] dx see (4) of Section 12.6 in text
=
2
c
2

2
n
J
2
1
(
n
c)
xJ
1
(x)

nc
0
=
2
c
n
J
1
(
n
c)
.
Then
A
n
=
2
ac
2
n
J
1
(
n
c)
and
u(r, t) =
2
ac

n=1
J
0
(
n
r)

2
n
J
1
(
n
c)
sin a
n
t.
2. From Example 1 in the text we have B
n
= 0 and
A
n
=
2
J
2
1
(
n
)
_
1
0
r(1 r
2
)J
0
(
n
r) dr.
763
14.2 Problems in Cylindrical Coordinates
From Problem 10, Exercises 12.6 we obtained A
n
=
4J
2
(
n
)

2
n
J
2
1
(
n
)
. Thus
u(r, t) = 4

n=1
J
2
(
n
)
J
2
1
(
n
)
cos a
n
tJ
0
(
n
r).
3. Referring to Example 2 in the text we have
R(r) = c
1
J
0
(r) +c
2
Y
0
(r)
Z(z) = c
3
cosh z +c
4
sinh z
where c
2
= 0 and J
0
(2) = 0 denes the positive eigenvalues
n
=
2
n
. From Z(4) = 0 we obtain
c
3
cosh 4
n
+c
4
sinh 4
n
= 0 or c
4
= c
3
cosh 4
n
sinh 4
n
.
Then
Z(z) = c
3
_
cosh
n
z
cosh 4
n
sinh 4
n
sinh
n
z
_
= c
3
sinh 4
n
cosh
n
z cosh 4
n
sinh
n
z
sinh 4
n
= c
3
sinh
n
(4 z)
sinh 4
n
and
u(r, z) =

n=1
A
n
sinh
n
(4 z)
sinh 4
n
J
0
(
n
r).
From
u(r, 0) = u
0
=

n=1
A
n
J
0
(
n
r)
we obtain
A
n
=
2u
0
4J
2
1
(2
n
)
_
2
0
rJ
0
(
n
r) dr =
u
0

n
J
1
(2
n
)
.
Thus the temperature in the cylinder is
u(r, z) = u
0

n=1
sinh
n
(4 z)J
0
(
n
r)

n
sinh 4
n
J
1
(2
n
)
.
4. (a) The boundary condition u
r
(2, z) = 0 implies R

(2) = 0 or J

0
(2) = 0. Thus = 0 is also an eigenvalue
and the separated equations are in this case rR

+R

= 0 and z

= 0. The solutions of these equations are


then
R(r) = c
1
+c
2
ln r, Z(z) = c
3
z +c
4
.
Now Z(0) = 0 yields c
4
= 0 and the implicit condition that the temperature is bounded as r 0 demands
that we dene c
2
= 0. Thus we have
u(r, z) = A
1
z +

n=2
A
n
sinh
n
zJ
0
(
n
r). (1)
At z = 4 we obtain
f(r) = 4A
1
+

n=2
A
n
sinh 4
n
J
0
(
n
r).
764
14.2 Problems in Cylindrical Coordinates
Thus from (17) and (18) of Section 12.6 in the text we can write with b = 2,
A
1
=
1
8
_
2
0
rf(r) dr (2)
A
n
=
1
2 sinh 4
n
J
2
0
(2
n
)
_
2
0
rf(r)J
0
(
n
r) dr. (3)
A solution of the problem consists of the series (1) with coecients A
1
and A
n
dened in (2) and (3),
respectively.
(b) When f(r) = u
0
we get A
1
= u
0
/4 and
A
n
=
u
0
J
1
(2
n
)

n
sinh 4
n
J
2
0
(2
n
)
= 0
since J

0
(2) = 0 is equivalent to J
1
(2) = 0. A solution of the problem is then u(r, z) =
u
0
4
z.
5. Letting u(r, t) = R(r)T(t) and separating variables we obtain
R

+
1
r
R

R
=
T

kT
= and R

+
1
r
R

+R = 0, T

+kT = 0.
From the last equation we nd T(t) = e
kt
. If < 0, T(t) increases without bound as t . Thus we assume
=
2
> 0. Now
R

+
1
r
R

+
2
R = 0
is a parametric Bessel equation with solution
R(r) = c
1
J
0
(r) +c
2
Y
0
(r).
Since Y
0
is unbounded as r 0 we take c
2
= 0. Then R(r) = c
1
J
0
(r) and the boundary condition
u(c, t) = R(c)T(t) = 0 implies J
0
(c) = 0. This latter equation denes the positive eigenvalues
n
=
2
n
. Thus
u(r, t) =

n=1
A
n
J
0
(
n
r)e

2
n
kt
.
From
u(r, 0) = f(r) =

n=1
A
n
J
0
(
n
r)
we nd
A
n
=
2
c
2
J
2
1
(
n
c)
_
c
0
rJ
0
(
n
r)f(r) dr, n = 1, 2, 3, . . . .
6. If the edge r = c is insulated we have the boundary condition u
r
(c, t) = 0. Referring to the solution of
Problem 5 above we have
R

(c) = c
1
J

0
(c) = 0
which denes an eigenvalue =
2
= 0 and positive eigenvalues
n
=
2
n
. Thus
u(r, t) = A
0
+

n=1
A
n
J
0
(
n
r)e

2
n
kt
.
From
u(r, 0) = f(r) = A
0
+

n=1
A
n
J
0
(
n
r)
765
14.2 Problems in Cylindrical Coordinates
we nd
A
0
=
2
c
2
_
c
0
rf(r) dr
A
n
=
2
c
2
J
2
0
(
n
c)
_
c
0
rJ
0
(
n
r)f(r) dr.
7. Referring to Problem 5 above we have T(t) = e
kt
and R(r) = c
1
J
0
(r). The boundary condition
hu(1, t) +u
r
(1, t) = 0 implies hJ
0
() +J

0
() = 0 which denes positive eigenvalues
n
=
2
n
. Now
u(r, t) =

n=1
A
n
J
0
(
n
r)e

2
n
kt
where
A
n
=
2
2
n
(
2
n
+h
2
)J
2
0
(
n
)
_
1
0
rJ
0
(
n
r)f(r) dr.
8. We solve

2
u
r
2
+
1
r
u
r
+

2
u
z
2
= 0, 0 < r < 1, z > 0
u
r

r=1
= hu(1, z), z > 0
u(r, 0) = u
0
, 0 < r < 1.
assuming u = RZ we get
R

+
1
r
R

R
=
Z

Z
=
and so
rR

+R

+
2
rR = 0 and Z

Z = 0.
Letting =
2
we then have
R(r) = c
1
J
0
(r) +c
2
Y
0
(r) and Z(z) = c
3
e
z
+c
4
e
z
.
We use the exponential form of the solution of Z

2
Z = 0 since the domain of the variable z is a semi-innite
interval. As usual we dene c
2
= 0 since the temperature is surely bounded as r 0. Hence R(r) = c
1
J
0
(r).
Now the boundary-condition u
r
(1, z) +hu(1, z) = 0 is equivalent to
J

0
() +hJ
0
() = 0. (4)
The eigenvalues
n
are the positive roots of (4) above. Finally, we must now dene c
4
= 0 since the temperature
is also expected to be bounded as z . A product solution of the partial dierential equation that satises
the rst boundary condition is given by
u
n
(r, z) = A
n
e
nz
J
0
(
n
r).
Therefore
u(r, z) =

n=1
A
n
e
nz
J
0
(
n
r)
is another formal solution. At z = 0 we have u
0
= A
n
J
0
(
n
r). In view of (4) above we use equations (17) and
(18) of Section 12.6 in the text with the identication b = 1:
A
n
=
2
2
n
(
2
n
+h
2
) J
2
0
(
n
)
_
1
0
rJ
0
(
n
r)u
0
dr
766
14.2 Problems in Cylindrical Coordinates
=
2
2
n
u
0
(
2
n
+h
2
) J
2
0
(
n
)
2
n
tJ
1
(t)

n
0
=
2
n
u
0
J
1
(
n
)
(
2
n
+h
2
) J
2
0
(
n
)
. (5)
Since J

0
= J
1
[see equation (6) of Section 11.5 in the text] it follows from (4) above that
n
J
1
(
n
) = hJ
0
(
n
).
Thus (5) above simplies to
A
n
=
2u
0
h
(
2
n
+h
2
) J
0
(
n
)
.
A solution to the boundary-value problem is then
u(r, z) = 2u
0
h

n=1
e
nz
(
2
n
+h
2
) J
0
(
n
)
J
0
(
n
r).
9. Substituting u(r, t) = v(r, t) +(r) into the partial dierential equation gives

2
v
r
2
+
1
r
v
r
+

+
1
r

=
v
t
.
This equation will be homogeneous provided

+
1
r

= 0 or (r) = c
1
ln r + c
2
. Since ln r is unbounded as
r 0 we take c
1
= 0. Then (r) = c
2
and using u(2, t) = v(2, t) + (2) = 100 we set c
2
= (2) = 100.
Therefore (r) = 100. Referring to Problem 5 above, the solution of the boundary-value problem

2
v
r
2
+
1
r
v
r
=
v
t
, 0 < r < 2, t > 0,
v(2, t) = 0, t > 0,
v(r, 0) = u(r, 0) (r)
is
v(r, t) =

n=1
A
n
J
0
(
n
r)e

2
n
t
where
A
n
=
2
2
2
J
2
1
(2
n
)
_
2
0
rJ
0
(
n
r)[u(r, 0) (r)] dr
=
1
2J
2
1
(2
n
)
_
_
1
0
rJ
0
(
n
r)[200 100] dr +
_
2
1
rJ
0
(
n
r)[100 100] dr
_
=
50
J
2
1
(2
n
)
_
1
0
rJ
0
(
n
r) dr x =
n
r, dx =
n
dr
=
50
J
2
1
(2
n
)
_
n
0
1

2
n
xJ
0
(x) dx
=
50

2
n
J
2
1
(2
n
)
_
n
0
d
dx
[xJ
1
(x)] dx see (5) of Section 12.6 in text
=
50

2
n
J
2
1
(2
n
)
_
xJ
1
(x)
_

n
0
=
50J
1
(
n
)

n
J
2
1
(2
n
)
.
Thus
u(r, t) = v(r, t) +(r) = 100 + 50

n=1
J
1
(
n
)J
0
(
n
r)

n
J
2
1
(2
n
)
e

2
n
t
.
10. Letting u(r, t) = u(r, t)+(r) we obtain r

= r. The general solution of this nonhomogeneous Cauchy-


Euler equation is found with the aid of variation of parameters: = c
1
+ c
2
ln r r
2
/4. In order that this
767
14.2 Problems in Cylindrical Coordinates
solution be bounded as r 0 we dene c
2
= 0. Using (1) = 0 then gives c
1
= /4 and so (r) = (1 r
2
)/4.
Using v = RT we nd that a solution of

2
v
r
2
+
1
r
v
r
=
v
t
, 0 < r < 1, t > 0
v(1, t) = 0, t > 0
v(r, 0) = (r), 0 < r < 1
is
v(r, t) =

n=1
A
n
e

2
n
t
J
0
(
n
r)
where
A
n
=

4
2
J
2
1
(
n
)
_
1
0
r(1 r
2
)J
0
(
n
r) dr
and the
n
are dened by J
0
() = 0. From the result of Problem 10, Exercises 12.6 (see also Problem 2 of this
exercise set) we get
A
n
=
J
2
(
n
)

2
n
J
2
1
(
n
)
.
Thus from u = v +(r) it follows that
u(r, t) =

4
(1 r
2
)

n=1
J
2
(
n
)

2
n
J
2
1
(
n
)
e

2
n
t
J
0
(
n
r).
11. (a) Writing the partial dierential equation in the form
g
_
x

2
u
x
2
+
u
x
_
=

2
u
t
2
and separating variables we obtain
xX

+X

X
=
T

gT
= .
Letting =
2
we obtain
xX

+X

+
2
X = 0 and T

+g
2
T = 0.
Letting x =
2
/4 in the rst equation we obtain dx/d = /2 or d/dx = 2. Then
dX
dx
=
dX
d
d
dx
=
2

dX
d
and
d
2
X
dx
2
=
d
dx
_
2

dX
d
_
=
2

d
dx
_
dX
d
_
+
dX
d
d
dx
_
2

_
=
2

d
d
_
dX
d
_
d
dx
+
dX
d
d
d
_
2

_
d
dx
=
4

2
d
2
X
d
2

4

3
dX
d
.
Thus
xX

+X

+
2
X =

2
4
_
4

2
d
2
X
d
2

4

3
dX
d
_
+
2

dX
d
+
2
X =
d
2
X
d
2
+
1

dX
d
+
2
X = 0.
This is a parametric Bessel equation with solution
X() = c
1
J
0
() +c
2
Y
0
().
768
14.2 Problems in Cylindrical Coordinates
(b) To insure a nite solution at x = 0 (and thus = 0) we set c
2
= 0. The condition u(L, t) = X(L)T(t) = 0
implies X

x=L
= X

=2

L
= c
1
J
0
(2

L) = 0, which denes positive eigenvalues


n
=
2
n
. The solution
of T

+g
2
T = 0 is
T(t) = c
3
cos(
n

g t) +c
4
sin(
n

g t).
The boundary condition u
t
(x, 0) = X(x)T

(0) = 0 implies c
4
= 0. Thus
u(, t) =

n=1
A
n
cos(
n

g t)J
0
(
n
).
From
u(, 0) = f(
2
/4) =

n=1
A
n
J
0
(
n
)
we nd
A
n
=
2
(2

L)
2
J
2
1
(2
n

L)
_
2

L
0
J
0
(
n
)f(
2
/4) d v = /2, dv = d/2
=
1
2LJ
2
1
(2
n

L)
_

L
0
2vJ
0
(2
n
v)f(v
2
)2 dv
=
2
LJ
2
1
(2
n

L)
_

L
0
vJ
0
(2
n
v)f(v
2
) dv.
The solution of the boundary-value problem is
u(x, t) =

n=1
A
n
cos(
n

g t)J
0
(2
n

x).
12. (a) First we see that
R

+
1
r
R

+
1
r
2
R

R
=
T

a
2
T
= .
This gives T

+a
2
T = 0 and from
R

+
1
r
R

+R
R/r
2
=

=
we get

+ = 0 and r
2
R

+rR

+ (r
2
)R = 0.
(b) With =
2
and =
2
the general solutions of the dierential equations in part (a) are
T = c
1
cos at +c
2
sin at
= c
3
cos +c
4
cos
R = c
5
J

(r) +c
6
Y

(r).
(c) Implicitly we expect u(r, , t) = u(r, + 2, t) and so must be 2-periodic. Therefore = n, n = 0, 1,
2, . . . . The corresponding eigenfunctions are 1, cos , cos 2, . . . , sin , sin 2, . . . . Arguing that u(r, , t)
is bounded as r 0 we then dene c
6
= 0 and so R = c
3
J
n
(r). But R(c) = 0 gives J
n
(c) = 0; this
equation denes the eigenvalues
n
=
2
n
. For each n,
ni
= x
ni
/c, i = 1, 2, 3, . . . . The corresponding
eigenfunctions are J
n
(
ni
r) = 0.
769
2 4 6 8 10
r
-1
-0.5
0.5
1
u
t=0
t=4
t=10
t=12
t=20
14.2 Problems in Cylindrical Coordinates
(d) u(r, , t) =
n

i=1
(A
0i
cos a
0i
t +B
0i
sin a
0i
t)J
0
(
0i
r)
+

n=1

i=1
_
(A
ni
cos a
ni
t +B
ni
sin a
ni
t) cos n
+ (C
ni
cos a
ni
t +D
ni
sin a
ni
t) sin n
_
J
n
(
ni
r)
13. (a) With c = 10 in Example 1 in the text the eigenvalues are
n
=
2
n
= x
2
n
/100 where x
n
is a positive root
of J
0
(x) = 0. From a CAS we nd that x
1
= 2.4048, x
2
= 5.5201, and x
3
= 8.6537, so that the rst three
eigenvalues are
1
= 0.0578,
2
= 0.3047, and
3
= 0.7489. The corresponding coecients are
A
1
=
2
100J
2
1
(x
1
)
_
10
0
rJ
0
(x
1
r/10)(1 r/10) dr = 0.7845,
A
2
=
2
100J
2
1
(x
2
)
_
10
0
rJ
0
(x
2
r/10)(1 r/10) dr = 0.0687,
and
A
3
=
2
100J
2
1
(x
3
)
_
10
0
rJ
0
(x
3
r/10)(1 r/10) dr = 0.0531.
Since g(r) = 0, B
n
= 0, n = 1, 2, 3, . . . , and the third partial sum of the series solution is
S
3
(r, t) =

n=1
A
n
cos(x
n
t/10)J
0
(x
n
r/10)
= 0.7845 cos(0.2405t)J
0
(0.2405r) + 0.0687 cos(0.5520t)J
0
(0.5520r)
+ 0.0531 cos(0.8654t)J
0
(0.8654r).
(b)
14. Because of the nonhomogeneous boundary condition u(c, t) = 200 we use the substitution u(r, t) = v(r, t)+(r).
This gives
k
_

2
v
r
2
+
1
r
v
r
+

+
1
r

_
=
v
t
.
This equation will be homogeneous provided

+(1/r)

= 0 or (r) = c
1
ln r +c
2
. Since ln r is unbounded as
r 0 we take c
1
= 0. Then (r) = c
2
and using u(c, t) = v(c, t) +c
2
= 200 we set c
2
= 200, giving v(c, t) = 0.
Referring to Problem 5 in this section, the solution of the boundary-value problem
k
_

2
v
r
2
+
1
r
v
r
_
=
v
t
, 0 < r < c, t > 0
770
5000 10000
t
-100
100
200
u
r0
r5
14.2 Problems in Cylindrical Coordinates
v(c, t) = 0, t > 0
v(r, 0) = 200, 0 < r < c
is
v(r, t) =

n=1
A
n
J
0
(
n
r)e

2
n
kt
,
where the separation constant is =
2
. The eigenvalues are
n
=
2
= x
2
n
/c
2
where x
n
is a positive root
of J
0
(x) = 0 and the coecients A
n
are
A
n
=
2
c
2
J
2
1
(
n
c)
_
c
0
rJ
0
(
n
r)(200)dr =
400
c
2
J
2
1
(
n
c)
_
c
0
rJ
0
(
n
r)dr.
Taking c = 10 and k = 0.1 we have
u(r, t) = v(r, t) + 200 = 200 +

n=1
A
n
J
0
(x
n
r/10)e
0.01x
2
n
t/100
where
A
n
=
4
J
2
1
(x
n
)
_
10
0
rJ
0
(x
n
r/10) dr.
Using a CAS we nd that the rst ve values of x
n
are
x
1
= 2.4048, x
2
= 5.5201, x
3
= 8.6537, x
4
= 11.7915,
and x
5
= 14.9309, with corresponding eigenvalues

1
= 0.0578,
2
= 0.3047,
3
= 0.7489,
4
= 1.3904,
and
5
= 2.2293. The rst ve values of A
n
are
A
1
= 320.4, A
2
= 213.0, A
3
= 170.3, A
4
= 145.9,
and A
5
= 129.7. Using a root nding application in
a CAS we nd that u(5, t) = 100 when t 1331 and
u(0, t) = 100 when t 2005. Since u = 200 is an asymptote for the graphs of u(0, t) and u(5, t) we solve
u(5, t) = 199.9 and u(0, t) = 199.9. This gives t 13,265 and t 13,958, respectively.
15. (a) The boundary-value problem is
a
2
_

2
u
r
2
+
1
r
u
r
_
=

2
u
t
2
, 0 < r < 1, t > 0
u(1, t) = 0, t > 0
u(r, 0) = 0,
u
t

t=0
=
_
v
0
, 0 r < b
0, b r < 1
, 0 < r < 1,
and the solution is
u(r, t) =

n=1
(A
n
cos a
n
t +B
n
sin a
n
t)J
0
(
n
r),
where the eigenvalues
n
=
2
n
are dened by J
0
() = 0 and A
n
= 0 since f(r) = 0. The coecients B
n
771
-1 -0.5 0.5 1
r
-0.2
-0.1
0.1
0.2
u
-1 -0.5 0.5 1
r
-0.2
-0.1
0.1
0.2
u
-1 -0.5 0.5 1
r
-0.2
-0.1
0.1
0.2
u
-1 -0.5 0.5 1
r
-0.2
-0.1
0.1
0.2
u
-1 -0.5 0.5 1
r
-0.2
-0.1
0.1
0.2
u
-1 -0.5 0.5 1
r
-0.2
-0.1
0.1
0.2
u
14.2 Problems in Cylindrical Coordinates
are given by
B
n
=
2
a
n
J
2
1
(
n
)
_
b
0
rJ
0
(
n
r)g(r) dr =
2v
0
a
n
J
2
1
(
n
)
_
b
0
rJ
0
(
n
r)dr
let x =
n
r
=
2v
0
a
n
J
2
1
(
n
)
_
nb
0
x

n
J
0
(x)
1

n
dx =
2v
0
a
3
n
J
2
1
(
n
)
_
nb
0
xJ
0
(x) dx
=
2v
0
a
3
n
J
2
1
(
n
)
_
xJ
1
(x)
_

nb
0
=
2v
0
a
3
n
J
1
(
n
)
_

n
bJ
1
(
n
b)
_
=
2v
0
b
a
2
n
J
1
(
n
b)
J
2
1
(
n
)
.
Thus,
u(r, t) =
2v
0
b
a

n=1
1

2
n
J
1
(
n
b)
J
2
1
(
n
)
sin(a
n
t)J
0
(
n
r).
(b) The standing wave u
n
(r, t) is given by u
n
(r, t) = B
n
sin(a
n
t)J
0
(
n
r), which has frequency f
n
= a
n
/2,
where
n
is the nth positive zero of J
0
(x). The fundamental frequency is f
1
= a
1
/2. The next two
frequencies are
f
2
=
a
2
2
=

2

1
_
a
1
2
_
=
5.520
2.405
f
1
= 2.295f
1
and
f
3
=
a
3
2
=

3

1
_
a
1
2
_
=
8.654
2.405
f
1
= 3.598f
1
.
(c) With a = 1, b =
1
4
, and v
0
= 1, the solution becomes
u(r, t) =
1
2

n=1
1

2
n
J
1
(
n
/4)
J
2
1
(
n
)
sin(
n
t)J
0
(
n
r).
The graphs of S
5
(r, t) for t = 1, 2, 3, 4, 5, 6 are shown below.
772
14.3 Problems in Spherical Coordinates
(d) Three frames from the movie are shown.
EXERCISES 14.3
Problems in Spherical Coordinates
1. To compute
A
n
=
2n + 1
2c
n
_

0
f()P
n
(cos ) sin d
we substitute x = cos and dx = sin d. Then
A
n
=
2n + 1
2c
n
_
1
1
F(x)P
n
(x)(dx) =
2n + 1
2c
n
_
1
1
F(x)P
n
(x) dx
where
F(x) =
_
0, 1 < x < 0
50, 0 < x < 1
= 50
_
0, 1 < x < 0
1, 0 < x < 1
.
The coecients A
n
are computed in Example 3 of Section 12.6. Thus
u(r, ) =

n=0
A
n
r
n
P
n
(cos )
= 50
_
1
2
P
0
(cos ) +
3
4
_
r
c
_
P
1
(cos )
7
16
_
r
c
_
3
P
3
(cos ) +
11
32
_
r
c
_
5
P
5
(cos ) +
_
.
2. In the solution of the Cauchy-Euler equation,
R(r) = c
1
r
n
+c
2
r
(n+1)
,
we dene c
1
= 0 since we expect the potential u to be bounded as r . Hence
u
n
(r, ) = A
n
r
(n+1)
P
n
(cos )
u(r, ) =

n=0
A
n
r
(n+1)
P
n
(cos ).
When r = c we have
f() =

n=0
A
n
c
(n+1)
P
n
(cos )
773
14.3 Problems in Spherical Coordinates
so that
A
n
= c
n+1
(2n + 1)
2
_

0
f()P
n
(cos ) sin d.
The solution of the problem is then
u(r, ) =

n=0
_
2n + 1
2
_

0
f()P
n
(cos ) sin d
_
_
c
r
_
n+1
P
n
(cos ).
3. The coecients are given by
A
n
=
2n + 1
2c
n
_

0
cos P
n
(cos ) sin d =
2n + 1
2c
n
_

0
P
1
(cos )P
n
(cos ) sin d
x = cos , dx = sin d
=
2n + 1
2c
n
_
1
1
P
1
(x)P
n
(x) dx.
Since P
n
(x) and P
m
(x) are orthogonal for m = n, A
n
= 0 for n = 1 and
A
1
=
2(1) + 1
2c
1
_
1
1
P
1
(x)P
1
(x) dx =
3
2c
_
1
1
x
2
dx =
1
c
.
Thus
u(r, ) =
r
c
P
1
(cos ) =
r
c
cos .
4. The coecients are given by
A
n
=
2n + 1
2c
n
_

0
(1 cos 2)P
n
(cos ) sin d.
These were computed in Problem 18 of Section 12.6. Thus
u(r, ) =
4
3
P
0
(cos )
4
3
_
r
c
_
2
P
2
(cos ).
5. Referring to Example 1 in the text we have
= P
n
(cos ) and R = c
1
r
n
+c
2
r
(n+1)
.
Since u(b, ) = R(b)() = 0,
c
1
b
n
+c
2
b
(n+1)
= 0 or c
1
= c
2
b
2n1
,
and
R(r) = c
2
b
2n1
r
n
+c
2
r
(n+1)
= c
2
_
b
2n+1
r
2n+1
b
2n+1
r
n+1
_
.
Then
u(r, ) =

n=0
A
n
b
2n+1
r
2n+1
b
2n+1
r
n+1
P
n
(cos )
where
b
2n+1
a
2n+1
b
2n+1
a
n+1
A
n
=
2n + 1
2
_

0
f()P
n
(cos ) sin d.
6. Referring to Example 1 in the text we have
R(r) = c
1
r
n
and () = P
n
(cos ).
774
14.3 Problems in Spherical Coordinates
Now (/2) = 0 implies that n is odd, so
u(r, ) =

n=0
A
2n+1
r
2n+1
P
2n+1
(cos ).
From
u(c, ) = f() =

n=0
A
2n+1
c
2n+1
P
2n+1
(cos )
we see that
A
2n+1
c
2n+1
= (4n + 3)
_
/2
0
f() sin P
2n+1
(cos ) d.
Thus
u(r, ) =

n=0
A
2n+1
r
2n+1
P
2n+1
(cos )
where
A
2n+1
=
4n + 3
c
2n+1
_
/2
0
f() sin P
2n+1
(cos ) d.
7. Referring to Example 1 in the text we have
r
2
R

+ 2rR

R = 0
sin

+ cos

+sin = 0.
Substituting x = cos , 0 /2, the latter equation becomes
(1 x
2
)
d
2

dx
2
2x
d
dx
+ = 0, 0 x 1.
Taking the solutions of this equation to be the Legendre polynomials P
n
(x) corresponding to
= n(n + 1) for n = 1, 2, 3, . . . , we have = P
n
(cos ). Since
u

=/2
=

(/2)R(r) = 0
we have

(/2) = (sin /2)P

n
(cos /2) = P

n
(0) = 0.
As noted in the hint, P

n
(0) = 0 only if n is even. Thus = P
n
(cos ), n = 0, 2, 4, . . . . As in Example 1,
R(r) = c
1
r
n
. Hence
u(r, ) =

n=0
A
2n
r
2n
P
2n
(cos ).
At r = c,
f() =

n=0
A
2n
c
2n
P
2n
(cos ).
Using Problem 19 in Section 12.6, we obtain
c
2n
A
2n
= (4n + 1)
_
0
/2
f()P
2n
(cos )(sin ) d
and
A
2n
=
4n + 1
c
2n
_
/2
0
f() sin P
2n
(cos ) d.
775
14.3 Problems in Spherical Coordinates
8. Referring to Example 1 in the text we have
R(r) = c
1
r
n
+c
2
r
(n1)
and () = P
n
(cos ).
Since we expect u(r, ) to be bounded as r , we dene c
1
= 0. Also (/2) = 0 implies that n is odd, so
u(r, ) =

n=0
A
2n+1
r
2(n+1)
P
2n+1
(cos ).
From
u(c, ) = f() =

n=0
A
2n+1
c
2(n+1)
P
2n+1
(cos )
we see that
A
2n+1
c
2(n+1)
= (4n + 3)
_
/2
0
f() sin P
2n+1
(cos ) d.
Thus
u(r, ) =

n=0
A
2n+1
r
2(n+1)
P
2n+1
(cos )
where
A
2n+1
= (4n + 3)c
2(n+1)
_
/2
0
f() sin P
2n+1
(cos ) d.
9. Checking the hint, we nd
1
r

2
r
2
(ru) =
1
r

r
_
r
u
r
+u
_
=
1
r
_
r

2
u
r
2
+
u
r
+
u
r
_
=

2
u
r
2
+
2
r
u
r
.
The partial dierential equation then becomes

2
r
2
(ru) = r
u
t
.
Now, letting ru(r, t) = v(r, t) +(r), since the boundary condition is nonhomogeneous, we obtain

2
r
2
[v(r, t) +(r)] = r

t
_
1
r
v(r, t) +(r)
_
or

2
v
r
2
+

(r) =
v
t
.
This dierential equation will be homogeneous if

(r) = 0 or (r) = c
1
r +c
2
. Now
u(r, t) =
1
r
v(r, t) +
1
r
(r) and
1
r
(r) = c
1
+
c
2
r
.
Since we want u(r, t) to be bounded as r approaches 0, we require c
2
= 0. Then (r) = c
1
r. When r = 1
u(1, t) = v(1, t) +(1) = v(1, t) +c
1
= 100,
and we will have the homogeneous boundary condition v(1, t) = 0 when c
1
= 100. Consequently, (r) = 100r.
The initial condition
u(r, 0) =
1
r
v(r, 0) +
1
r
(r) =
1
r
v(r, 0) + 100 = 0
implies v(r, 0) = 100r. We are thus led to solve the new boundary-value problem

2
v
r
2
=
v
t
, 0 < r < 1, t > 0,
v(1, t) = 0, lim
r0
1
r
v(r, t) < ,
v(r, 0) = 100r.
776
14.3 Problems in Spherical Coordinates
Letting v(r, t) = R(r)T(t) and using the separation constant we obtain
R

+R = 0 and T

+T = 0.
Using =
2
> 0 we then have
R(r) = c
3
cos r +c
4
sin r and T(t) = c
5
e

2
t
.
The boundary conditions are equivalent to R(1) = 0 and lim
r0
R(r)/r < . Since
lim
r0
cos r
r
does not exist we must have c
3
= 0. Then R(r) = c
4
sin r, and R(1) = 0 implies = n for n = 1, 2, 3, . . . .
Thus
v
n
(r, t) = A
n
e
n
2

2
t
sin nr
for n = 1, 2, 3, . . . . Using the condition lim
r0
R(r)/r < it is easily shown that there are no eigenvalues
for = 0, nor does setting the common constant to =
2
when separating variables lead to any solutions.
Now, by the superposition principle,
v(r, t) =

n=1
A
n
e
n
2

2
t
sin nr.
The initial condition v(r, 0) = 100r implies
100r =

n=1
A
n
sin nr.
This is a Fourier sine series and so
A
n
= 2
_
1
0
(100r sin nr) dr = 200
_

r
n
cos nr

1
0
+
_
1
0
1
n
cos nr dr
_
= 200
_

cos n
n
+
1
n
2

2
sin nr

1
0
_
= 200
_

(1)
n
n
_
=
(1)
n
200
n
.
A solution of the problem is thus
u(r, t) =
1
r
v(r, t) +
1
r
(r) =
1
r

n=1
(1)
n
20
n
e
n
2

2
t
sin nr +
1
r
(100r)
=
200
r

n=1
(1)
n
n
e
n
2

2
t
sin nr + 100.
10. Referring to Problem 9 we have

2
v
r
2
+

(r) =
v
t
where (r) = c
1
r. Since
u(r, t) =
1
r
v(r, t) +
1
r
(r) =
1
r
v(r, t) +c
1
we have
u
r
=
1
r
v
r
(r, t)
1
r
2
v(r, t).
When r = 1,
u
r

r=1
= v
r
(1, t) v(1, t)
777
14.3 Problems in Spherical Coordinates
and
u
r

r=1
+hu(1, t) = v
r
(1, t) v(1, t) +h[v(1, t) +(1)] = v
r
(1, t) + (h 1)v(1, t) +hc
1
.
Thus the boundary condition
u
r

r=1
+hu(1, t) = hu
1
will be homogeneous when hc
1
= hu
1
or c
1
= u
1
. Consequently (r) = u
1
r. The initial condition
u(r, 0) =
1
r
v(r, 0) +
1
r
(r) =
1
r
v(r, 0) +u
1
= u
0
implies v(r, 0) = (u
0
u
1
)r. We are thus led to solve the new boundary-value problem

2
v
r
2
=
v
t
, 0 < r < 1, t > 0,
v
r
(1, t) + (h 1)v(1, t) = 0, t > 0,
lim
r0
1
r
v(r, t) < ,
v(r, 0) = (u
0
u
1
)r.
Separating variables as in Problem 9 leads to
R(r) = c
3
cos r +c
4
sin r and T(t) = c
5
e

2
t
.
The boundary conditions are equivalent to
R

(1) + (h 1)R(1) = 0 and lim


r0
1
r
R(r) < .
As in Problem 6 we use the second condition to determine that c
3
= 0 and R(r) = c
4
sin r. Then
R

(1) + (h 1)R(1) = c
4
cos +c
4
(h 1) sin = 0
and the
n
are the consecutive nonnegative roots of tan = /(1 h). Now
v(r, t) =

n=1
A
n
e

2
n
t
sin
n
r.
From
v(r, 0) = (u
0
u
1
)r =

n=1
A
n
sin
n
r
we obtain
A
n
=
_
1
0
(u
0
u
1
)r sin
n
r dr
_
1
0
sin
2

n
r dr
.
We compute the integrals
_
1
0
r sin
n
r dr =
_
1

2
n
sin
n
r
1

n
cos
n
r
_

1
0
=
1

2
n
sin
n

n
cos
n
and
_
1
0
sin
2

n
r dr =
_
1
2
r
1
4
n
sin 2
n
r
_

1
0
=
1
2

1
4
n
sin 2
n
.
778
14.3 Problems in Spherical Coordinates
Using
n
cos
n
= (h 1) sin
n
we then have
A
n
= (u
0
u
1
)
1

2
n
sin
n

1
n
cos
n
1
2

1
4n
sin 2
n
= (u
0
u
1
)
4 sin
n
4
n
cos
n
2
2
n

n
2 sin
n
cos
n
= 2(u
0
u
1
)
sin
n
+ (h 1) sin
n

2
n
+ (h 1) sin
n
sin
n
= 2(u
0
u
1
)h
sin
n

2
n
+ (h 1) sin
2

n
.
Therefore
u(r, t) =
1
r
v(r, t) +
1
r
(r) = u
1
+ 2(u
0
u
1
)h

n=1
sin
n
sin
n
r
r[
2
n
+ (h 1) sin
2

n
]
e

2
n
t
.
11. We write the dierential equation in the form
a
2
1
r

2
r
2
(ru) =

2
u
t
2
or a
2

2
r
2
(ru) = r

2
u
t
2
,
and then let v(r, t) = ru(r, t). The new boundary-value problem is
a
2

2
v
r
2
=

2
v
t
2
, 0 < r < c, t > 0
v(c, t) = 0, t > 0
v(r, 0) = rf(r),
v
t

t=0
= rg(r).
Letting v(r, t) = R(r)T(t) and using the separation constant =
2
we obtain
R

+
2
R = 0
T

+a
2

2
T = 0
and
R(r) = c
1
cos r +c
2
sin r
T(t) = c
3
cos at +c
4
sin at.
Since u(r, t) = v(r, t)/r, in order to insure boundedness at r = 0 we dene c
1
= 0. Then R(r) = c
2
sin r and
the condition R(c) = 0 implies = n/c. Thus
v(r, t) =

n=1
_
A
n
cos
na
c
t +B
n
sin
na
c
t
_
sin
n
c
r.
From
v(r, 0) = rf(r) =

n=1
A
n
sin
n
c
r
we see that
A
n
=
2
c
_
c
0
rf(r) sin
n
c
r dr.
From
v
t

t=0
= rg(r) =

n=1
_
B
n
na
c
_
sin
n
c
r
we see that
B
n
=
c
na

2
c
_
c
0
rg(r) sin
n
c
r dr =
2
na
_
c
0
rg(r) sin
n
c
r dr.
The solution is
u(r, t) =
1
r

n=1
_
A
n
cos
na
c
t +B
n
sin
na
c
t
_
sin
n
c
r,
779
14.3 Problems in Spherical Coordinates
where A
n
and B
n
are given above.
12. Proceeding as in Example 1 we obtain
() = P
n
(cos ) and R(r) = c
1
r
n
+c
2
r
(n+1)
so that
u(r, ) =

n=0
(A
n
r
n
+B
n
r
(n+1)
)P
n
(cos ).
To satisfy lim
r
u(r, ) = Er cos we must have A
n
= 0 for n = 2, 3, 4, . . . . Then
lim
r
u(r, ) = Er cos = A
0
1 +A
1
r cos ,
so A
0
= 0 and A
1
= E. Thus
u(r, ) = Er cos +

n=0
B
n
r
(n+1)
P
n
(cos ).
Now
u(c, ) = 0 = Ec cos +

n=0
B
n
c
(n+1)
P
n
(cos )
so

n=0
B
n
c
(n+1)
P
n
(cos ) = Ec cos
and
B
n
c
(n+1)
=
2n + 1
2
_

0
Ec cos P
n
(cos ) sin d.
Now cos = P
1
(cos ) so, for n = 1,
_

0
cos P
n
(cos ) sin d = 0
by orthogonality. Thus B
n
= 0 for n = 1 and
B
1
=
3
2
Ec
3
_

0
cos
2
sin d = Ec
3
.
Therefore,
u(r, ) = Er cos +Ec
3
r
2
cos .
780
CHAPTER 14 REVIEW EXERCISES
CHAPTER 14 REVIEW EXERCISES
1. We have
A
0
=
1
2
_

0
u
0
d +
1
2
_
2

(u
0
) d = 0
A
n
=
1
c
n

_

0
u
0
cos n d +
1
c
n

_
2

(u
0
) cos n d = 0
B
n
=
1
c
n

_

0
u
0
sin n d +
1
c
n

_
2

(u
0
) sin n d =
2u
0
c
n
n
[1 (1)
n
]
and so
u(r, ) =
2u
0

n=1
1 (1)
n
n
_
r
c
_
n
sin n.
2. We have
A
0
=
1
2
_
/2
0
d +
1
2
_
2
3/2
d =
1
2
A
n
=
1
c
n

_
/2
0
cos n d +
1
c
n

_
2
3/2
cos n d =
1
c
n
n
_
sin
n
2
sin
3n
2
_
B
n
=
1
c
n

_
/2
0
sin n d +
1
c
n

_
2
3/2
sin n d =
1
c
n
n
_
cos
3n
2
cos
n
2
_
and so
u(r, ) =
1
2
+
1

n=1
_
r
c
_
n
_
sin
n
2
sin
3n
2
n
cos n +
cos
3n
2
cos
n
2
n
sin n
_
.
3. The conditions (0) = 0 and () = 0 applied to = c
1
cos +c
2
sin give c
1
= 0 and = n,
n = 1, 2, 3, . . . , respectively. Thus we have the Fourier sine-series coecients
A
n
=
2

_

0
u
0
(
2
) sin n d =
4u
0
n
3

[1 (1)
n
].
Thus
u(r, ) =
4u
0

n=1
1 (1)
n
n
3
r
n
sin n.
4. In this case
A
n
=
2

_

0
sin sin n d =
1

_

0
[cos(1 n) cos(1 +n)] d = 0, n = 1.
For n = 1,
A
1
=
2

_

0
sin
2
d =
1

_

0
(1 cos 2) d = 1.
Thus
u(r, ) =

n=1
A
n
r
n
sin n
781
CHAPTER 14 REVIEW EXERCISES
reduces to
u(r, ) = r sin .
5. The insulation conditions are u

(r, 0) = 0 and u

(r, ) = 0. The eigenvalue = 0 corresponds to a constant


eigenfunction. The other eigenvalues are = n
2
, n = 1, 2, 3, . . . , with corresponding eigenfunctions () =
c
1
cos n. Also, assuming a bounded solution at r = 0, R(r) = c
3
r
n
and so
u(r, ) = A
0
+

n=1
A
n
r
n
cos n.
At r = c
f() = A
0
+

n=1
A
n
c
n
cos n,
from which we see that
A
0
=
a
0
2
=
1

_

0
f() d and A
n
=
2

_

0
f() cos n d.
6. Two of the boundary conditions are u(r, 0) = 0 and u

(r, ) = 0 which imply (0) = 0 and

() = 0. For
= [(2n 1)/2]
2
, n = 1, 2, 3, . . . , we have () = c
2
sin[(2n 1)/2]. Assuming a bounded solution at r = 0
we have R(r) = c
3
r
n
, so
u(r, ) =

n=1
A
n
r
n
sin
_
2n 1
2

_
and f() =

n=1
A
n
c
n
sin
_
2n 1
2

_
.
This is not a Fourier series but it is an orthogonal series expansion of f, so
A
n
= c
n
_
_
_
_
_

0
f() sin
_
2n 1
2

_
d
_

0
sin
2
_
2n 1
2

_
d
_
_
_
_
.
7. We solve

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0, 0 < <

4
,
1
2
< r < 1,
u(r, 0) = 0, u(r, /4) = 0,
1
2
< r < 1,
u(1/2, ) = u
0
, u
r
(1, ) = 0, 0 < <

4
.
Proceeding as in Example 1 in Section 14.1 using the separation constant =
2
we obtain
r
2
R

+rR

R = 0

+ = 0
with solutions
() = c
1
cos +c
2
sin
R(r) = c
3
r

+c
4
r

.
Applying the boundary conditions (0) = 0 and (/4) = 0 gives c
1
= 0 and = 4n for
n = 1, 2, 3, . . . . From R
r
(1) = 0 we obtain c
3
= c
4
. Therefore
u(r, ) =

n=1
A
n
_
r
4n
+r
4n
_
sin 4n.
782
CHAPTER 14 REVIEW EXERCISES
From
u(1/2, ) = u
0
=

n=1
A
n
_
1
2
4n
+
1
2
4n
_
sin 4n
we nd
A
n
_
1
2
4n
+
1
2
4n
_
=
2
/4
_
/4
0
u
0
sin 4n d =
2u
0
n
[1 (1)
n
]
or
A
n
=
2u
0
n(2
4n
+ 2
4n
)
[1 (1)
n
].
Thus the steady-state temperature in the plate is
u(r, ) =
2u
0

n=1
[r
4n
+r
4n
][1 (1)
n
]
n[2
4n
+ 2
4n
]
sin 4n.
8. The boundary-value problem in polar coordinates is

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0, 0 < < , a < r < b,
u(a, ) = 0, u(b, ) = f(), 0 < < ,
u(r, 0) = u
0
, u(r, ) = u
1
, a < r < b.
Since the boundary conditions are nonhomogeneous at = 0 and = we let u(r, ) = v(r, ) + ().
Substituting into the partial dierential equation we obtain
v
rr
+
1
r
v
r
+
1
r
2
(v

) = 0.
This equation will be homogeneous if we require

= 0. In this case () = c
1
+c
2
. The boundary conditions
(0) = u
0
and () = u
1
then give c
1
= u
0
and c
2
= (u
1
u
0
)/. Hence
() = u
0
+
u
1
u
0

.
The homogeneous boundary-value problem for v(r, ) is then

2
v
r
2
+
1
r
v
r
+
1
r
2

2
v

2
= 0, 0 < < , a < r < b,
v(a, ) = u
0

u
1
u
0

, 0 < < ,
v(b, ) = f() u
0

u
1
u
0

, 0 < < ,
v(r, 0) = 0, v(r, ) = 0, a < r < b.
Letting v(r, ) = R(r)() and separating variables gives
() = c
3
cos +c
4
sin and R(r) = c
5
r

+c
6
r

.
The boundary conditions (0) = 0 and () = 0 give c
3
= 0 and = n/. Thus, () = c
4
sin n/ and
v(r, ) =

n=1
(A
n
r
n/
+B
n
r
n/
) sin
n

.
At r = a we have
u
0

u
1
u
0

n=1
(A
n
a
n/
+B
n
a
n/
) sin
n

,
783
CHAPTER 14 REVIEW EXERCISES
so
A
n
a
n/
+B
n
a
n/
=
2

_

0
_
u
0
+
u
1
u
0


_
sin
n

d.
Similarly, at r = b we have
A
n
b
n/
+B
n
b
n/
=
2

_

0
_
f() u
0

u
1
u
0


_
sin
n

d.
Solving the above two simultaneous equations for A
n
and B
n
we get
A
n
=

2
b
n/
_

0
_
u
0
+
u
1
u
0


_
sin
n

d
2
a
n/
_

0
_
f() u
0

u
1
u
0


_
sin
n

d
_
a
b
_
n/

_
b
a
_
n/
and
B
n
=

2a
n/

_

0
_
f() u
0

u
1
u
0


_
sin
n

d +
2b
n/

_

0
_
u
0
+
u
1
u
0


_
sin
n

d
_
a
b
_
n/

_
b
a
_
n/
.
9. The boundary-value problem in polar coordinates is

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0, 0 < < 2, 1 < r < 2,
u(1, ) = sin
2
,
u
r

r=2
= 0, 0 < < 2.
Letting u(r, ) = R(r)() and separating variables we obtain
r
2
R

+rR

R = 0 and

+ = 0.
For = 0 we have

= 0 and r
2
R

+ rR

= 0. This gives = c
1
+ c
2
and R = c
3
+ c
4
ln r. The
periodicity assumption (as mentioned in Example 1 of Section 14.1 in the text) implies c
2
= 0, while the
boundary condition R

(2) = 0 implies c
4
= 0. Thus, for = 0, u = c
1
c
3
= A
0
. Now, for =
2
, the
dierential equations become

+
2
= 0 and r
2
R

+ rR


2
R = 0. The corresponding solutions are
= c
5
cos + c
6
sin and R = c
7
r

+ c
8
r

. In this case the periodicity assumption implies = n,


n = 1, 2, 3, . . . , while the boundary condition R

(2) = 0 implies R = c
7
(r
n
+ 4
n
r
n
). The product of the
solutions is u
n
= (r
n
+ 4
n
r
n
)(A
n
cos n +B
n
sin n) and the superposition principle implies
u(r, ) = A
0
+

n=1
(r
n
+ 4
n
r
n
)(A
n
cos n +B
n
sin n).
Using the boundary condition at r = 1 we have
sin
2
=
1
2
(1 cos 2) = A
0
+

n=1
(1 + 4
n
)(A
n
cos n +B
n
sin n).
From this we conclude that B
n
= 0 for all integers n, A
0
=
1
2
, A
1
= 0, A
2
=
1
34
, and A
m
= 0 for
m = 3, 4, 5, . . . . Therefore
u(r, ) = A
0
+A
2
cos 2 =
1
2

1
34
(r
2
+ 16r
2
) cos 2 =
1
2

_
1
34
r
2
+
8
17
r
2
_
cos 2.
784
CHAPTER 14 REVIEW EXERCISES
10. We solve

2
u
r
2
+
1
r
u
r
+
1
r
2

2
u

2
= 0, r > 1, 0 < < ,
u(r, 0) = 0, u(r, ) = 0, r > 1,
u(1, ) = f(), 0 < < .
Separating variables we obtain
() = c
1
cos +c
2
sin
R(r) = c
3
r

+c
4
r

.
Applying the boundary conditions (0) = 0, and () = 0 gives c
1
= 0 and = n for n = 1, 2, 3, . . . .
Assuming f() to be bounded, we expect the solution u(r, ) to also be bounded as r . This requires that
c
3
= 0. Therefore
u(r, ) =

n=1
A
n
r
n
sin n.
From
u(1, ) = f() =

n=1
A
n
sin n
we obtain
A
n
=
2

_

0
f() sin n d.
11. Letting u(r, t) = R(r)T(t) and separating variables we obtain
R

+
1
r
R

hR
R
=
T

T
=
so
R

+
1
r
R

( +h)R = 0 and T

T = 0.
From the second equation we nd T(t) = c
1
e
t
. If > 0, T(t) increases without bound as t . Thus we
assume =
2
< 0. Since h > 0 we can take =
2
h. Then
R

+
1
r
R

+
2
R = 0
is a parametric Bessel equation with solution
R(r) = c
1
J
0
(r) +c
2
Y
0
(r).
Since Y
0
is unbounded as r 0 we take c
2
= 0. Then R(r) = c
1
J
0
(r) and the boundary condition u(1, t) =
R(1)T(t) = 0 implies J
0
() = 0. This latter equation denes the positive eigenvalues
n
. Thus
u(r, t) =

n=1
A
n
J
0
(
n
r)e
(
2
n
h)t
.
From
u(r, 0) = 1 =

n=1
A
n
J
0
(
n
r)
we nd
A
n
=
2
J
2
1
(
n
)
_
1
0
rJ
0
(
n
r) dr x =
n
r, dx =
n
dr
=
2
J
2
1
(
n
)
_
n
0
1

2
n
xJ
0
(x) dx.
785
CHAPTER 14 REVIEW EXERCISES
From recurrence relation (5) in Section 12.6 of the text we have
xJ
0
(x) =
d
dx
[xJ
1
(x)].
Then
A
n
=
2

2
n
J
2
1
(
n
)
_
n
0
d
dx
[xJ
1
(x)] dx =
2

2
n
J
2
1
(
n
)
_
xJ
1
(x)
_

n
0
=
2
1
J
1
(
n
)

2
n
J
2
1
(
n
)
=
2

n
J
1
(
n
)
and
u(r, t) = 2e
ht

n=1
J
0
(
n
r)

n
J
1
(
n
)
e

2
n
t
12. Letting =
2
> 0 and proceeding in the usual manner we nd
u(r, t) =

n=1
A
n
cos a
n
tJ
0
(
n
r)
where the eigenvalues
n
=
2
n
are determined by J
0
() = 0. Then the initial condition gives
u
0
J
0
(x
k
r) =

n=1
A
n
J
0
(
n
r)
and so
A
n
=
2
J
2
1
(
n
)
_
1
0
r (u
0
J
0
(x
k
r)) J
0
(
n
r) dr.
But J
0
() = 0 implies that the eigenvalues are the positive zeros of J
0
, that is,
n
= x
n
for n = 1, 2, 3, . . . .
Therefore
A
n
=
2u
0
J
2
1
(
n
)
_
1
0
rJ
0
(
k
r)J
0
(
n
r) dr = 0, n = k
by orthogonality. For n = k,
A
k
=
2u
0
J
2
1
(
k
)
_
1
0
rJ
2
0
(
k
) dr = u
0
by (7) of Section 12.6. Thus the solution u(r, t) reduces to one term when n = k, and
u(r, t) = u
0
cos a
k
tJ
0
(
k
r) = u
0
cos ax
k
tJ
0
(x
k
r).
13. Letting the separation constant be =
2
and referring to Example 2 in Section 14.2 in the text we have
R(r) = c
1
J
0
(r) +c
2
Y
0
(r)
Z(z) = c
3
cosh z +c
4
sinh z
where c
2
= 0 and the positive eigenvalues
n
are determined by J
0
(2) = 0. From Z

(0) = 0 we obtain c
4
= 0.
Then
u(r, z) =

n=1
A
n
cosh
n
zJ
0
(
n
r).
From
u(r, 4) = 50 =

n=1
A
n
cosh 4
n
J
0
(
n
r)
we obtain (as in Example 1 of Section 14.1)
A
n
cosh 4
n
=
2(50)
4J
2
1
(2
n
)
_
2
0
rJ
0
(
n
r) dr =
50

n
J
1
(2
n
)
.
786
CHAPTER 14 REVIEW EXERCISES
Thus the temperature in the cylinder is
u(r, z) = 50

n=1
cosh
n
zJ
0
(
n
r)

n
cosh 4
n
J
1
(2
n
)
.
14. Using u = RZ and as a separation constant and then letting =
2
> 0 leads to
r
2
R

+rR

+
2
r
2
R = 0, R

(1) = 0, and Z

2
Z = 0.
Thus
R(r) = c
1
J
0
(r) +c
2
Y
0
(r)
Z(z) = c
3
cosh z +c
4
sinh z
for > 0. Arguing that u(r, z) is bounded as r 0 we dene c
2
= 0. Since the eigenvalues are dened by
J

0
() = 0 we know that = = 0 is an eigenvalue. The solutions are then
R(r) = c
1
+c
2
ln r and Z(z) = c
3
z +c
4
where boundedness again dictates that c
2
= 0. Thus,
u(r, z) = A
0
z +B
0
+

n=1
(A
n
sinh
n
z +B
n
cosh
n
z)J
0
(
n
r).
Finally, the specied conditions z = 0 and z = 1 give, in turn,
B
0
= 2
_
1
0
rf(r) dr
B
n
=
2
J
2
0
(
n
)
_
1
0
rf(r)J
0
(
n
r) dr
A
0
= B
0
+ 2
_
1
0
rg(r) dr
A
n
=
1
sinh
n
_
B
n
cosh
n
+
2
J
2
0
(
n
)
_
1
0
rg(r)J
0
(
n
r) dr
_
.
15. Referring to Example 1 in Section 14.3 of the text we have
u(r, ) =

n=0
A
n
r
n
P
n
(cos ).
For x = cos
u(1, ) =
_
100 0 < < /2
100 /2 < <
= 100
_
1, 1 < x < 0
1, 0 < x < 1
= g(x).
From Problem 22 in Exercise 12.6 we have
u(r, ) = 100
_
3
2
rP
1
(cos )
7
8
r
3
P
3
(cos ) +
11
16
r
5
P
5
(cos ) +
_
.
16. Since
1
r

2
r
2
(ru) =
1
r

r
_
r
u
r
+u
_
=
1
r
_
r

2
u
r
2
+
u
r
+
u
r
_
=

2
u
r
2
+
2
r
u
r
the dierential equation becomes
1
r

2
r
2
(ru) =

2
u
t
2
or

2
r
2
(ru) = r

2
u
t
2
.
787
CHAPTER 14 REVIEW EXERCISES
Letting v(r, t) = ru(r, t) we obtain the boundary-value problem

2
v
r
2
=

2
v
t
2
, 0 < r < 1, t > 0
v
r

r=1
v(1, t) = 0, t > 0
v(r, 0) = rf(r),
v
t

t=0
= rg(r), 0 < r < 1.
If we separate variables using v(r, t) = R(r)T(t) and separation constant then we obtain
R

R
=
T

T
=
so that
R

+R = 0
T

+T = 0.
Letting =
2
> 0 and solving the dierential equations we get
R(r) = c
1
cos r +c
2
sin r
T(t) = c
3
cos t +c
4
sin t.
Since u(r, t) = v(r, t)/r, in order to insure boundedness at r = 0 we dene c
1
= 0. Then R(r) = c
2
sin r. Now
the boundary condition R

(1) R(1) = 0 implies cos sin = 0. Thus, the eigenvalues


n
are determined
by the positive solutions of tan = . We now have
v
n
(r, t) = (A
n
cos
n
t +B
n
sin
n
t) sin
n
r.
For the eigenvalue = 0,
R(r) = c
1
r +c
2
and T(t) = c
3
t +c
4
,
and boundedness at r = 0 implies c
2
= 0. We then take
v
0
(r, t) = A
0
tr +B
0
r
so that
v(r, t) = A
0
tr +B
0
r +

n=1
(a
n
cos
n
t +B
n
sin
n
t) sin
n
r.
Now
v(r, 0) = rf(r) = B
0
r +

n=1
A
n
sin
n
r.
Since {r, sin
n
r} is an orthogonal set on [0, 1],
_
1
0
r sin
n
r dr = 0 and
_
1
0
sin
n
r sin
n
r dr = 0
for m = n. Therefore
_
1
0
r
2
f(r) dr = B
0
_
1
0
r
2
dr =
1
3
B
0
and
B
0
= 3
_
1
0
r
2
f(r) dr.
788
CHAPTER 14 REVIEW EXERCISES
Also
_
1
0
rf(r) sin
n
r dr = A
n
_
1
0
sin
2

n
r dr
and
A
n
=
_
1
0
rf(r) sin
n
r dr
_
1
0
sin
2

n
r dr
.
Now
_
1
0
sin
2

n
r dr =
1
2
_
1
0
(1 cos 2
n
r) dr =
1
2
_
1
sin 2
n
2
n
_
=
1
2
[1 cos
2

n
].
Since tan
n
=
n
,
1 +
2
n
= 1 + tan
2

n
= sec
2

n
=
1
cos
2

n
and
cos
2

n
=
1
1 +
2
n
.
Then
_
1
0
sin
2

n
r dr =
1
2
_
1
1
1 +
2
n
_
=

2
n
2(1 +
2
n
)
and
A
n
=
2(1 +
2
n
)

2
n
_
1
0
rf(r) sin
n
r dr.
Similarly, setting
v
t

t=0
= rg(r) = A
0
r +

n=1
B
n

n
sin
n
r
we obtain
A
0
= 3
_
1
0
r
2
g(r) dr
and
B
n
=
2(1 +
2
n
)

3
n
_
1
0
rg(r) sin
n
r dr.
Therefore, since v(r, t) = ru(r, t) we have
u(r, t) = A
0
t +B
0
+

n=1
(A
n
cos
n
t +B
n
sin
n
t)
sin
n
r
r
,
where the
n
are solutions of tan = and
A
0
= 3
_
1
0
r
2
g(r) dr
B
0
= 3
_
1
0
r
2
f(r) dr
A
n
=
2(1 +
2
n
)

2
n
_
1
0
rf(r) sin
n
r dr
B
n
=
2(1 +
2
n
)

3
n
_
1
0
rg(r) sin
n
r dr
for n = 1, 2, 3, . . . .
789
CHAPTER 14 REVIEW EXERCISES
17. We note that the dierential equation can be expressed in the form
d
dx
[xu

] =
2
xu.
Thus
u
n
d
dx
[xu

m
] =
2
m
xu
m
u
n
and
u
m
d
dx
[xu

n
] =
2
n
xu
n
u
m
.
Subtracting we obtain
u
n
d
dx
[xu

m
] u
m
d
dx
[xu

n
] = (
2
n

2
m
)xu
m
u
n
and
_
b
a
u
n
d
dx
[xu

m
] dx
_
b
a
u
m
d
dx
[xu

n
] = (
2
n

2
m
)
_
b
a
xu
m
u
n
dx.
Using integration by parts this becomes
u
n
xu

b
a

_
b
a
xu

m
u

n
dx u
m
xu

b
a
+
_
b
a
xu

n
u

m
dx
= b[u
n
(b)u

m
(b) u
m
(b)u

n
(b)] a[u
n
(a)u

m
(a) u
m
(a)u

n
(a)]
= (
2
n

2
m
)
_
b
a
xu
m
u
n
dx.
Since
u(x) = Y
0
(a)J
0
(x) J
0
(a)Y
0
(x)
we have
u
n
(b) = Y
0
(
n
a)J
0
(
n
b) J
0
(
n
a)Y
0
(
n
b) = 0
by the denition of the
n
. Similarly u
m
(b) = 0. Also
u
n
(a) = Y
0
(a)J
0
(
n
a) J
0
(
n
a)Y
0
(
n
a) = 0
and u
m
(a) = 0. Therefore
_
b
a
xu
m
u
n
dx =
1

2
n

2
m
(b[u
n
(b)u

m
(b) u
m
(b)u

n
(b)] a[u
n
(a)u

m
(a) u
m
(a)u

n
(a)]) = 0
and the u
n
(x) are orthogonal with respect to the weight function x.
18. Letting u(r, t) = R(r)T(t) and the separation constant be =
2
we obtain
rR

+R

+
2
rR = 0
T

+
2
T = 0,
with solutions
R(r) = c
1
J
0
(r) +c
2
Y
0
(r)
T(t) = c
3
e

2
t
.
Now the boundary conditions imply
R(a) = 0 = c
1
J
0
(a) +c
2
Y
0
(a)
R(b) = 0 = c
1
J
0
(b) +c
2
Y
0
(b)
790
CHAPTER 14 REVIEW EXERCISES
so that
c
2
=
c
1
J
0
(a)
Y
0
(a)
and
c
1
J
0
(b)
c
1
J
0
(a)
Y
0
(a)
Y
0
(b) = 0
or
Y
0
(a)J
0
(b) J
0
(a)Y
0
(b) = 0.
This equation denes
n
for n = 1, 2, 3, . . . . Now
R(r) = c
1
J
0
(r) c
1
J
0
(a)
Y
0
(a)
Y
0
(r) =
c
1
Y
0
(a)
_
Y
0
(a)J
0
(r) J
0
(a)Y
0
(r)
_
and
u
n
(r, t) = A
n
_
Y
0
(
n
a)J
0
(
n
r) J
0
(
n
a)Y
0
(
n
r)
_
e

2
n
t
= A
n
u
n
(r)e

2
n
t
.
Thus
u(r, t) =

n=1
A
n
u
n
(r)e

2
n
t
.
From the initial condition
u(r, 0) = f(r) =

n=1
A
n
u
n
(r)
we obtain
A
n
=
_
b
a
rf(r)u
n
(r) dr
_
b
a
ru
2
n
(r) dr
.
19. We use the superposition principle for Laplaces equation discussed in Section 13.5 and shown schematically in
Figure 13.15 in the text. That is,
Solution u = Solution u
1
of Problem 1 + Solution u
2
of Problem 2,
where in Problem 1 the boundary condition on the top and bottom of the cylinder is u = 0, while on the lateral
surface r = c it is u = h(z), and in Problem 2 the boundary condition on the top of the cylinder z = L is
u = f(r), on the bottom z = 0 it is u = g(r), and on the lateral surface r = c it is u = 0.
20.
Solution for u
1
(r, z)
Using as a separation constant we have
R

+
1
r
R

R
=
Z

Z
= ,
so
rR

+R

rR = 0 and Z

+Z = 0.
The dierential equation in Z, together with the boundary conditions Z(0) = 0 and Z(L) = 0 is a Sturm-
Liouville problem. Letting =
2
> 0 we note that the above dierential equation in R is a modied parametric
Bessel equation which is discussed in Section 5.3 in the text. Also, we have Z(z) = c
1
cos z + c
2
sin z. The
boundary conditions imply c
1
= 0 and sin L = 0. Thus,
n
= n/L, n = 1, 2, 3, . . . , so
n
= n
2

2
/L
2
and
R(r) = c
3
I
0
_
n
L
r
_
+c
4
K
0
_
n
L
r
_
.
791
CHAPTER 14 REVIEW EXERCISES
Now boundedness at r = 0 implies c
4
= 0, so R(r) = c
3
I
0
(nr/L) and
u
1
(r, z) =

n=1
A
n
I
0
_
n
L
r
_
sin
_
n
L
z
_
.
At r = c for 0 < z < L we have
h(z) = u
1
(c, z) =

n=1
A
n
I
0
_
n
L
c
_
sin
_
n
L
z
_
which gives
A
n
=
2
LI
0
(nc/L)
_
L
0
h(z) sin
_
n
L
z
_
dz.
Solution for u
2
(r, z)
In this case we use as a separation constant which leads to
R

+
1
r
R

R
=
Z

Z
= ,
so
rR

+R

+rR = 0 and Z

Z = 0.
The dierential equation in R is a parametric Bessel equation. Using =
2
we nd R(r) = c
1
J
0
(r)+c
2
Y
0
(r).
Boundedness at r = 0 implies c
2
= 0 so R(r) = c
3
J
0
(r). The boundary condition R(c) = 0 then gives the
dening equation for the eigenvalues: J
0
(c) = 0. Let
n
=
2
n
where
n
c = x
n
are the roots. The solution of
the dierential equation in Z is Z(z) = c
4
cosh
n
z +c
5
sinh
n
z, so
u
2
(r, z) =

n=1
(B
n
cosh
n
z +C
n
sinh
n
z)J
0
(
n
r).
At z = 0, for 0 < r < c, we have
f(r) = u
2
(r, 0) =

n=1
B
n
J
0
(
n
r),
so
B
n
=
2
c
2
J
2
1
(
n
c)
_
c
0
rf(r)J
0
(
n
r) dr.
At z = L, for 0 < r < c, we have
g(r) = u
2
(r, L) =

n=1
(B
n
cosh
n
L +C
n
sinh
n
L)J
0
(
n
r),
so
B
n
cosh
n
L +C
n
sinh
n
L =
2
c
2
J
2
1
(
n
c)
_
c
0
rg(r)J
0
(
n
r) dr
and
C
n
= B
n
cosh
n
L
sinh
n
L
+
2
c
2
(sinh
n
L)J
2
1
(
n
c)
_
c
0
rg(r)J
0
(
n
r) dr.
By the superposition principle the solution of the original problem is
u(r, z) = u
1
(r, z) +u
2
(r, z).
792

You might also like