1

Math 425 / AMCS 525 Dr. DeTurck

Assignment 5 Due Tuesday, February 23, 2016

Topics for this week Duhamels principle and other techniques for inhomogeneous problems
Solutions on finite domains - separated solutions Boundary conditions eigenvalue (Sturm-Liouville)
problems for ODEs

Fifth Homework Assignment - due Tuesday, February 23

Reading: Read sections 3.2, 3.3, 3.4, 4.1 and 4.2 of the text

Be prepared to discuss the following problems in class:

Page 66 (page 64 in the first ed), problems 1, 3, 9

Page 70 (page 68 in the first ed), problem 1
Page 79 (page 76 in the first ed), problems 1, 4, 10
Page 89 (page 87 in the first ed), problems 2, 3
Page 92 (page 90 in the first ed), problem 1

Page 66, problem 1. The Neumann problem for the wave equation on 0 < x < is ut t = c2 uxx
for all x > 0 and t > 0 subject to the initial conditions u(x, 0) = (x) and ut (x, 0) = (x) for x > 0
and the boundary condition ux (0, t) = 0 for t > 0. To solve this, we need the even extensions of
(x) and (x):
( (
(x) if x > 0 (x) if x > 0
(x) = and (x) =
(x) if x < 0 (x) if x < 0
We need to interpret dAlemberts solution
Z x+ct
1  1
u(x, t) = (x + ct) + (x ct) + (s) ds
2 2c xct
in terms of these extended functions. There are two cases: either x > ct or x < ct.

In the first case, x > ct, we have x + ct > 0 and x ct > 0. In this case,
Z x+ct
1  1
u(x, t) = (x + ct) + (x ct) + (s) ds
2 2c xct

In the second case, x < ct, we have x + ct > 0 and x ct < 0 (or ct x > 0). In this case,
Z 0 Z x+ct 
1  1
u(x, t) = (x + ct) + (ct x) + (s) ds + (s) ds
2 2c xct 0
Z ctx Z x+ct 
1   1
= (x + ct) + (ct x) + (q) dq + (s) ds
2 2c 0 0
 Z ctx Z x+ct 
1  1
= (x + ct) + (ct x) + 2 (q) dq + (s) ds
2 2c 0 ctx
2

where we made the substitution q = s in one of the integrals. We conclude that

Z x+ct
1  1
(x + ct) + (x ct) + (s) ds if x ct


2 2c xct


u(x, t) =  Z ctx Z x+ct 
1  1
(x + ct) + (ct x) + 2 (q) dq + (s) ds if x ct



2 2c 0 ctx

Page 66, problem 3. Since u(x, t) = f (x + ct) for x > 0 and t < 0, we have

u(x, 0) = f (x) and ut (x, 0) = cf 0 (x) (for x < 0)

We also have u(0, t) = 0 because the end is fixed. Because of this, we extend f and cf 0 as odd
functions:
( (
f (x) if x > 0 cf 0 (x) if x > 0
(x) = and (x) =
f (x) if x < 0 cf 0 (x) if x < 0

We need to interpret dAlemberts solution

Z x+ct
1  1
u(x, t) = (x + ct) + (x ct) + (s) ds
2 2c xct

in terms of these extended functions. There are two cases: either x > ct or x < ct.

In the first case, x > ct, we have x + ct > 0 and x ct > 0. In this case,
Z x+ct
1  1
u(x, t) = f (x + ct) + f (x ct) + cf 0 (s) ds
2 2c xct
1  1 
= f (x + ct) + f (x ct) + f (x + ct) f (x ct)
2 2
= f (x + ct)

In the second case, x < ct, we have x + ct > 0 and x ct < 0 (or ct x > 0). In this case,
Z 0 Z x+ct 
1  1 0 0
u(x, t) = f (x + ct) f (ct x) + cf (s) ds + cf (s) ds
2 2c xct 0

1  1 Z 0 
0
= f (x + ct) f (ct x) + f (q) dq + f (x + ct) f (0)
2 2 ctx

1  1 
= f (x + ct) f (ct x) + f (0) f (ct x) + f (x + ct) f (0)
2 2
= f (x + ct) f (ct x)

where we made the substitution q = s in one of the integrals. We conclude that

(
f (x + ct) if x ct
u(x, t) =
f (x + ct) f (ct x) if x ct
 3

Page 66, problem 9. We want to solve utt = uxx on the x-interval 0 < x < 1. Since c = 1,
the characteristics have slope 1. And because the boundary conditions u(0, t) = u(1, t) = 0 are
homogeneous Dirichlet conditions, we need to make the odd extensions ext and ext of the initial
data u(x, 0) = x2 (1 x) and ut (x, 0) = (1 x)2 at every integer (so these extensions will have period
2). In dAlemberts solution
Z x+ct
1  1
u(x, t) = ext (x + ct) + ext (x ct) + ext (s) ds
2 2c xct

we can use the oddness of the extensions to simplify the integral of ext , since whenever we integrate
over an x-interval of length 2 we will get zero.

(a) To find u( 23 , 2), we need initial data on the x-interval [ 23 2, 23 + 2] = [ 43 , 38 ]. Since this
interval has length 4, we know that the integral of the odd periodic extension of over it is zero.
So we know that

u( 32 , 2) = 12 (ext ( 43 ) + ext ( 38 )) = 21 (ext ( 23 ) + ext ( 43 )) = 12 (( 23 ) + ( 23 )) = u( 23 , 0) = 4

27

(b) To find u( 14 , 72 ), we need initial data on the x-interval [ 14 27 , 14 + 27 ] = [ 13 15

4 , 4 ]. Unfortunately,
this time the interval has length 7, so we cant conclude that the integral of ext is zero, but we can
simplify it a bit:
Z 15/4 Z 7/4 Z 1/4 Z 3/4
ext (s) ds = ext (s) ds = ext (s) ds = ext (s) ds
13/4 5/4 3/4 1/4

3/4 3/4
(1 x)3  1 27
Z
13
= ut (s, 0) ds = = =
1/4 3 
1/4 192 96

where between the last two integrals we used that ext is odd so that its integral from 41 to 1
4 is
zero. We conclude that:
   
u( 14 , 72 ) = 21 ext ( 13 4 ) + ext ( 15
4 ) + 1
2 ( 13
96 ) = 1
2 ext ( 5
4 ) + ext ( 7
4 ) 13
192
   
= 12 ext ( 34 ) + ext ( 41 ) 192 13
= 12 u( 34 , 0) u( 14 , 0) 192
13

 
= 12 64
9
64 3 13
192 = 64 3 13
192 = 1924
= 48 1
.

Page 70, problem 1. We want to solve ut kuxx = f (x, t) for x > 0 and t > 0 with u(0, t) = 0
and u(x, 0) = (x). Since were dealing with a Dirichlet condition at x = 0, we make odd extensions
of (x) and f (x, t) so we let
(
f (x, t) for x > 0
F (x, t) =
f (x, t) for x < 0

and (
(x) for x > 0
h(x)
(x) for x < 0
4

Then
2
Z t Z (xy) Z (xy)2
1 4k(ts) 1
u(x, t) = p e F (y, s) dy ds + e 4kt h(y) dy
0 4k(t s) 4kt

2
Then, writing S(x, t) = ex /4kt / 4kt as usual, we have
Z tZ Z
u(x, t) = S(x y, t s)F (y, s) dy ds + S(x y, t)h(y) dy
0
Z t Z Z 0 
= S(x y, t s)F (y, s) dy + S(x y, t s)F (y, s) dy ds
0 0
Z Z 0
+ S(x y, t)h(y) dy + S(x y, t)h(y) dy
0
Z t Z Z 0 
= S(x y, t s)f (y, s) dy + S(x y, t s)(f (y, s)) dy ds
0 0
Z Z 0
+ S(x y, t)(y) dy + S(x y, t)((y)) dy
0
Z t Z Z 
= S(x y, t s)f (y, s) dy + S(x + y, t s)(f (y, s)) dy ds
0 0 0
Z Z
+ S(x y, t)(y) dy + S(x + y, t)((y)) dy
0 0
2 2 2 2
Z tZ h (xy) h
(x+y) i (xy) (x+y)
Z
1 1 i
= p e 4k(ts) e 4k(ts) h(y, s) dy ds + e 4k(t) e 4k(t) (y) dy
0 0 4k(t s) 4kt 0

Page 79, problem 1. Since the initial position and velocity are both zero, the solution of utt =
c2 uxx + xt is given by
Z tZ x+ctcs Z t y=x+ctcs
1 1 1 2 
u(x, t) = ys dy ds = y s ds
2c 0 xct+cs 2c 0 2 y=xct+cs
Z t  s=t
1 1 4
= 4x(ct cs)s ds = 2xcts2 xcs3
4c 0 4c 3 s=0
 
1 4 3 1
= 2xct xct = xt3
3
4c 3 6

Page 79, problem 4. The formula for the solution of

utt = c2 utt + f (x, t) u(x, 0) = 0 and ut (x, 0) = (x) for x > 0, and u(0, t) = (x)

is given in Theorem 1 on page 71 of the text:

Z x+ct ZZ
1 1 1
u(x, t) = [(x + ct) + (x ct)] + (s) ds + f.
2 2c xct 2c
 5

And as you can see:

Z x+ct ZZ
1 1 1
u(x, t) = 2 [(x + ct) + (x ct)] + (s) ds + f
2c xct 2c

it is the sum of three terms, one involving , one involving and one involving f .

Page 79, problem 10. Just for practice, well use the Greens function method. So suppose that
u(x, t) (for x > 0 and t > 0) is the solution of

utt = c2 utt + f (x, t) u(x, 0) = 0 and ut (x, 0) = 0 for x > 0, and u(0, t) = 0

The domain of dependence D for u(x0 , t0 ) is either the triangle with vertices (x0 ct0 , 0), (x0 +xt0 , 0),
and (x0 , t0 ) if x0 > ct0 , or else it is the quadrilateral with vertices (ct0 x0 , 0), (x0 + ct0 , 0), (x0 , t0 )
and (0, t0 x0 /c). For x0 > ct0 we already know (see the notes) that
ZZ
1
u(x0 , t0 ) = f (y, s) dy ds.
2c D

For x0 < ct0 , we have to prove this. So here goes: Assume that u(x, t) is the solution of the problem,
and that ct > x. Then D is the quadrilateral described above and
ZZ ZZ I
f (y, s) dy ds = utt c2 uxx = ut dx c2 ux dt
D D bd(D)

The evaluation of the line integral has four parts:

The horizontal segment on the x-axis starting at (ct0 x0 , 0) and ending at (x0 + ct0 , 0). This
part evaluates to zero because ut (x, 0) = 0 by the initial conditions and dt = 0 on the segment
since t is identically zero there.
The diagonal segment from (x0 + ct0 , 0) to (x0 , t0 ). We have to do this one. Parametrize the
segment via
x = x0 + c(t0 s) and t = s for 0 s t0
and then dx = c ds and dt = ds so the integral becomes
Z (x0 ,t0 ) Z t0
2
ut dx c ux dt = cut (x0 + c(t0 s), s) c2 ux (x0 + c(t0 s), s) ds
(x0 +ct0 ,0) 0

Now we need an observation: with x and t defined as functions of s as in the parametrization,

we have that
du u dx u dt
= + = cux (x0 + c(t0 s), s) + ut (x0 + c(t0 s), s).
ds x ds t ds
This is precisely 1/c times the integrand, so the value of this part of the line integral is the
difference between the value of cu(x, t) at the end of the segment and at the beginning of it.
In other words:
Z (x0 ,t0 )
1 
ut dx c2 ux dt = u(x0 , t0 ) u(x0 + ct0 , 0) = c u(x0 , t0 )
(x0 +ct0 ,0) c

because of the initial condition u(x, 0) = 0.

6

The diagonal segment from (x0 , t0 ) to the point (0, t0 x0 /c) on the t-axis. We parametrize
this segment via
x0
x = x0 cs and t = t0 s for 0 s
c
and then dx = c ds and dt = ds so the integral becomes
Z (0,t0 x0 /c) Z x0 /c
2
ut dx c ux dt = cut (x0 cs, t0 s) + c2 ux (x0 cs, t0 s) ds
(x0 ,t0 ) 0

We make an observation as on the previous segment: with x and t defined as functions of s as

in the parametrization of this segment, we have that
du u dx u dt
= + = cux (x0 cs, t0 s) ut (x0 cs, t0 s).
ds x ds t ds
And this is precisely 1/c times the integrand, so the value of this part of the line integral is
the difference between the value of cu(x, t) at the end of the segment and at the beginning
of it. In other words:
Z (0,t0 x0 /c)   x0  
ut dx c2 ux dt = cc u 0, t0 u(x0 , t0 ) = c u(x0 , t0 )
(x0 ,t0 ) c

because of the boundary condition u(0, t) = 0.

Finally there is the segment from (0, t0 x0 /c) back to (ct0 x0 , 0). Follow the same drill,
and parametrize:
x0 x0
x = cs and t = t0 s for 0 s t0
c c
and then dx = c ds and dt = ds so the integral becomes
Z (ct0 x0 ,0) Z t0 x0 /c
2
 x0   x0 
ut dx c ux dt = cut cs, t0 s + c2 ux cs, t0 s ds
(0,t0 x0 /c) 0 c c

And now make the observation that with x and t defined as functions of s as in the parametriza-
tion of this segment, we have that
du u dx u dt  x0   x0 
= + = cux cs, t0 s ut cs, t0 s .
ds x ds t ds c c
And this is precisely 1/c times the integrand, so the value of this part of the line integral is
the difference between the value of cu(x, t) at the end of the segment and at the beginning of
it. But u is zero at both points, because of the boundary and initial conditions, so that the
value of the integral over this segment is zero.

Now we can insert the value of the line integral back into our original computation:
ZZ ZZ I
2
f (y, s) dy ds = utt c uxx = ut dxc2 ux dt = 0+cu(x0 , t0 )+cu(x0 , t0 )+0 = 2cu(x0 , t0 ).
D D bd(D)

Thus the solution of the initial-boundary value problem is

ZZ
1
u(x, t) = f (y, s) dy ds.
2c D
 7

Page 89, problem 2. The differential equation for the temperature of the rod starting at the
moment the ends are immersed is the heat equation ut = kuxx , together with the initial condition
u(x, 0) = 1 for 0 < x < ` and boundary conditions u(0, t) = u(`, t) = 0 for t > 0. This is the
initial/boundary-value problem given by equations (13), (14) and (15) on page 87 of the textbook
(with (x) = 1 for all x between 0 and `). The solution is given by equation (17):

X 2 nx
u(x, t) = An e(n/`) kt
sin ,
n=1
`

with the coefficients An chosen so that

X nx
1= An sin .
n=1
`

The problem tells us that

4

if n is odd
An = n
0 if n is even

So, writing the odd numbers as 2k + 1 for k = 0, 1, 2, . . . we can express the solution as

X 4 2 (2k + 1)x
u(x, t) = e[(2k+1)/`] kt sin
(2k + 1) `
k=0

Page 89,
problem 3. The Schrodinger equation looks just like the heat equation with k replaced
by i = 1. So the solution of the Dirichlet problem corresponding to it should look just like the
solution (equation (17) on page 87) of the heat equation, with k replaced by i:

X 2 nx
u(x, t) = An e(n/`) it
sin
n=1
`

Page 92, problem 1. If we seek separated solutions u(x, t) = X(x)T (t) of the diffusion equation
ut = kuxx , we get that X(x)T 0 (t) = kX 00 (x)T (t), or in other words

T 00 (t) X 00 (x)
= = .
kT (t) X(x)

The boundary conditions u(0, t) = 0 and ux (`, t) = 0 give us conditions X(0) = 0 and X 0 (`) = 0 on
X, so we solve the equation for X first.

Can we have < 0? In that case, the solution of X 00 + X = 0 would be

X(x) = c1 cosh( x) + c2 sinh( x).

The condition X(0) = 0 implies that c1 = 0, and then X 0 (`) = c2 cosh( `) = 0 implies that
R
c2 = 0 (since cosh x 6= 0 for all x ). So there are no negative eigenvalues.
8

Can we have = 0? In that case, the solution of X 00 = 0 is X(x) = c1 + c2 x, and the boundary
conditions again force c1 = c2 = 0, so zero is not an eigenvalue.

For > 0, the solution of X 00 + X = 0 is

X(x) = c1 cos( x) + c2 sin( x).

The condition X(0) = 0 implies that c1 = 0 and then X 0 (`) = c2 cos( `) = 0. So we can have
c2 6= 0 if is chosen so that
3 5
cos( `) = 0 , in other words ` = , , ,...
2 2 2
Therefore we have
(2n + 1)2 2
Eigenvalues: for n = 0, 1, 2, 3, . . .
4`2
(2n + 1)x
Eigenfunctions: Xn (x) = sin for n = 0, 1, 2, 3, . . .
2`

Now we deal with the T equation: T 0 + kT = 0. The solution of this is T (t) = cekt . So the
solution of the diffusion equation with the given boundary conditions is

X 2 (2n + 1)x
u(x, t) = An e[(2n+1)/2`] kt
sin .
n=0
2`

Write up solutions of the following to hand in:

Page 66 (page 64 in the first ed), problems 4, 10

Page 70 (page 68 in the first ed), problem 2
Page 79 (page 76 in the first ed), problems 2, 5, 13
Page 89 (page 87 in the first ed), problems 4, 6
Page 92 (page 90 in the first ed), problem 2

Page 66, problem 4. Since u(x, t) = f (x + ct) for x > 0 and t < 0, we have

u(x, 0) = f (x) and ut (x, 0) = cf 0 (x) (for x < 0)

We also have ux (0, t) = 0 because the end is free. Because of this, we extend f and cf 0 as even
functions:
( (
f (x) if x > 0 cf 0 (x) if x > 0
(x) = and (x) =
f (x) if x < 0 cf 0 (x) if x < 0

We need to interpret dAlemberts solution

Z x+ct
1  1
u(x, t) = (x + ct) + (x ct) + (s) ds
2 2c xct
 9

in terms of these extended functions. There are two cases: either x > ct or x < ct.

In the first case, x > ct, we have x + ct > 0 and x ct > 0. In this case,
Z x+ct
1  1
u(x, t) = f (x + ct) + f (x ct) + cf 0 (s) ds
2 2c xct
1  1 
= f (x + ct) + f (x ct) + f (x + ct) f (x ct)
2 2
= f (x + ct)

In the second case, x < ct, we have x + ct > 0 and x ct < 0 (or ct x > 0). In this case,
Z 0 Z x+ct 
1  1
u(x, t) = f (x + ct) + f (ct x) + cf 0 (s) ds + cf 0 (s) ds
2 2c xct 0
Z 0 
1   1
= f (x + ct) + f (ct x) + f 0 (q) dq + f (x + ct) f (0)
2 2 ctx

1   1  
= f (x + ct) + f (ct x) + f (ct x) f (0) + f (x + ct) f (0)
2 2
= f (x + ct) + f (ct x) f (0)
where we made the substitution q = s in one of the integrals. We conclude that
(
f (x + ct) if x ct
u(x, t) =
f (x + ct) + f (ct x) f (0) if x ct

Page 66, problem 10. We need to solve utt = 9uxx for 0 < x < /2 (so c = 3) with initial
conditions u(x, 0) = cos x and ut (x, 0) = 0 on that interval, and ux (0, t) = 0 (so the left end is
free) and u(/2 , t) = 0 (so the right end is fixed). Since cos(x) = cos x and cos(/2 + x) =
cos(/2 x), the appropriate periodic extension of u(x, 0) which should be even at multiples
of and odd at odd multiples of /2 is just cos x. So we can use dAlemberts solution with
(x) = cos x for all x and (x) = 0 and the solution is
1 
u(x, t) = cos(x + 3t) + cos(x 3t) = cos x cos 3t
2
after using the trig identities cos(A B) = cos A cos B sin A sin B.

Page 70, problem 2. We need to solve vt kvxx = f (x, t) for x > 0 and t > 0 together with the
initial condition v(x, 0) = (x) and the boundary condition v(0, t) = h(t). So to start, we suppress
the boundary condition by letting
w(x, t) = v(x, t) h(t).
Then w satisfies wt kwxx = f (x, t) h0 (t) for x > 0 and t > 0 together with the initial condition
w(x, 0) = (x) h(0) and the boundary condition w(0, t) = 0. We are going to need the odd
extension of w(x, 0), which were going to call Q(x):
(
(x) h(0) if x > 0
Q(x) =
(x) + h(0) if x < 0
10

Also, let F (x, t) be the odd extension of f (x, t) h0 (t), so

(
f (x, t) h0 (t) if x > 0
F (x, t) =
f (x, t) + h0 (t) if x < 0

Then
Z Z tZ
w(x, t) = S(x y, t)Q(y) dy + S(x y, t s)F (y, s) dy ds
0
Z   Z 0  
= S(x y, t) (y) h(0) dy + S(x y, t) (y) + h(0) dy
0
Z tZ   Z tZ 0  
0
+ S(x y, t s) f (y, s) h (s) dy ds + S(x y, t s) f (y, s) + h0 (s) dy ds
0 0 0
Z   Z  
= S(x y, t) (y) h(0) dy + S(x + y, t) (y) + h(0) dy
0 0
Z tZ   Z tZ  
+ S(x y, t s) f (y, s) h0 (s) dy ds + S(x + y, t s) f (y, s) + h0 (s) dy ds
0 0 0 0
Z   
= S(x y, t) S(x + y, t) (y) h(0) dy
0
Z tZ   
+ S(x y, t s) S(x + y, t s) f (y, s) h0 (s) dy ds
0 0

Theres only a bit of simplification (or Erf-ificaion) we can do here. The terms with h(0) and h0 (s)
can be brought out of at least one integral and then we need to compute an integral of the form:
Z Z  (x+y)2 (xy)2

1 4kt 4kt
S(x + y, t) S(x y, t) dy = e e dy.
0 0 4kt

Make the change of variables p = (x + y)/ 4kt (so dp = dy/ 4kt) in the first integral and q =
(x y)/ 4kt (so dq = dy/ 4kt) in the second to get
Z Z x/4kt        
1 p2 1 q2 1 1 x 1 1 x x
e dp e dq = Erf + Erf = Erf .
x/ 4kt 2 2 4kt 2 2 4kt 4kt
Using this, we can rewrite our expression for w(x, t) above as
Z  (xy)2 (x+y)2
  
1 x
w(x, t) = e 4kt e 4kt (y) dy h(0) Erf
4kt 0 4kt
2
(x+y)2
Z tZ ! Z t !
1 (xy) x
+ p e 4k(ts) e 4k(ts) f (y, s) dy ds h0 (s) Erf p ds
0 0 4k(t s) 0 4k(t s)

And finally (remember, we were solving for v(x, t)!)

Z  (xy)2 (x+y)2
  
1 4kt 4kt x
v(x, t) = h(t) + e e (y) dy h(0) Erf
4kt 0 4kt
2
(x+y)2
Z tZ ! Z t !
1 (xy) x
+ p e 4k(ts) e 4k(ts) f (y, s) dy ds h0 (s) Erf p ds
0 0 4k(t s) 0 4k(t s)
 11

Page 79, problem 2. We need to solve utt = c2 uxx +eax on the whole line with zero initial position
and velocity. By the Duhamel formula (only the term for the inhomogeneity of the differential
equation will be present):
Z tZ x+ctcs
1 1 t a(x+ctcs)
u(x, t) = eay dy ds = e e ea(xct+cs) ds
2c 0 xct+cs 2ac 0
1 h a(x+ctcs) it 1 h ax i
a(xct+cs) a(x+ct) a(xct)
= e e = 2e a e
2a2 c2 0 2a2 c2
1 h i
= 2 2 aa(x+ct) + ea(xct) 2eax
2a c

Page 79, problem 5. We need to verify that the solution of

utt = c2 uxx + f , u(x, 0) = 0 , ut (x, 0) = 0

on the whole line is given by

Z tZ x+ctcs
1
u(x, t) = f (y, s) dy ds.
2c 0 xct+cs

Its easy to see that

Z 0
1
u(x, 0) = = 0
2c 0

so thats one initial condition out of the way. Next, calculate

Z x Z t Z x+ctcs 
u 1 1
= f (y, t) dy + f (y, s) dy ds
t 2c x 2c 0 t xct+cs

using Leibnizs rule (or the fundamental theorem of calculus in its fanciest form). The first term is
clearly zero. Use Leibniz again on the second term (the only t0 s are in the limits of integration) to
conclude that:
1 t
Z
u
= f (x + ct cs, s) + f (x ct + cs, s) ds.
t 2 0
Observe that this shows that ut (x, 0) = 0 since well have another integral from 0 to 0.

Next, use Leibniz on the expression for ut above to get

t
2u
Z
c f f
2
= f (x, t) + (x + ct cs, s) (x ct + cs, s) ds
t 2 0 x x

because the only ts in the integrand occur in the x slot of f .

Now we move on to the x-derivatives of u:

Z t
u 1
= f (x + ct cs, s) f (x ct + cs, s) ds
x 2c 0
12

since the only place om the formula for u that x appears is in the limits of integration of the inner
integral. And then
t
2u
Z
1 f f
2
= (x + ct cs, s) (x ct + cs, s) ds
x 2c 0 x x

And comparing the expressions for utt ad uxx above, you can see that utt c2 uxx = f (x, t).

Page 79, problem 13. We need to solve the wave equation utt = c2 uxx for x > 0 with initial
conditions u(x, 0) = x and ut (x, 0) = 0 and boundary condition u(0, t) = t2 .

We will trade the inhomogeneous boundary condition for an inhomogeneous equation by letting
v(x, t) = u(x, t) t2 . Then v will satisfy

vtt = c2 vxx 2 for x > 0

together with initial conditions

v(x, 0) = x and vt (x, 0) = 0

and boundary condition

v(0, t) = 0.
Because the boundary condition is a Dirichlet condition, we need to extend the initial data and the
inhomogeneity in the equation to be odd functions in x. Fortunately, the initial data are already
odd functions. To work on the whole line we need to rewrite the differential equation is
(
2 2 for x > 0
vtt = c vxx + f (x) where f (x) =
2 for x < 0

Then Z tZ x+ct
1 1h i
v(x, t) = f (y) dy ds + (x + ct) + (x ct)
2c 0 xct 2
If x > ct then f (y) is identically 2 over the triangular domain D of integration, which has area
1 2 2
2 (2ct)t = ct , so for x > ct we have v(x, t) = t + x.

Forx < ct, the function f (x) is equal to +2 over the small triangular portion of D with vertices
x
(0, 0), 0, t , (x ct, 0), which has area
c
1  x 1
(ct x) t = (c2 t2 2ctx + x2 )
2 c 2c
So the integral
Z tZ x+ct
2x2
 
1 2 2
f (y) dy ds = 2ct2 + 4 (c t 2ctx + x2 ) = 4tx.
0 xct 2c c

Therefore
x t2

for x > ct
v(x, t) = 2
x + x 2tx for x < ct
c2 c
 13

Since u(x, t) = v(x, t) + t2 , we have

x for x > ct
u(x, t) = 2
x + x 2tx + t2 for x < ct
c2 c

Page 89, problem 4. We want to find separated solutions for the damped wave equation

utt = c2 uxx rut for 0 < x < ` and t > 0

2c
(where 0 < r < ) that satisfy the initial conditions
`
u(x, 0) = (x) and ut (x, 0) = (x)

and the boundary conditions

u(0, t) = 0 and u(`, t) = 0
Substitute u(x, t) = X(x)T (t) into the equation to obtain

XT 00 + rXT 0 = c2 X 00 T

or
T 00 + rT 0 X 00
= =
c2 T X
So the problem for X(x) is

X 00 + X = 0 and X(0) = X(`) = 0.

This tells us that

n2 2 nx
= and Xn (x) = sin , n = 1, 2, . . .
`2 `
Now T (t) must be a solution of
n2 c2 2
T 00 + rT 0 + c2 T = 0 or T 00 + rT + T = 0.
`2
The roots of the characteristic equation are
r
r 1 4n2 c2 2
r2
2 2 `2
2c
But the quantity under the radical is negative for n = 1, 2, 3, . . ., since 0 < r < , so we rewrite
`
it as r
r 1 4n2 c2 2
i r2
2 2 `2
and so r ! r !
4n 2 2 c2 t 4n 2 2 c2 t
Tn = An ert/2 cos r2 + Bn ert/2 sin r2
`2 2 `2 2
and so the solution of the initial/boundary-value problem is

" r ! r !#
X X
rt/2 4n2 2 c2 t rt/2 4n 2 2 c2 t nx
u(x, t) = Xn (x)Tn (t) = An e cos 2
r2 + Bn e sin 2
r2 sin
n=1 n=1
` 2 ` 2 `
14

where

X nx
An sin = (x)
n=1
`
and

r !
X 1 4n2 2 c2 r nx
r 2 B n An sin = (x).
n=1
2 `2 2 `

Page 89, problem 6. We want to find separated solutions to the equation

tut = uxx + 2u for 0 < x < and t > 0

that satisfy the initial condition u(x, 0) = 0 and the boundary conditions u(0, t) = u(, t) = 0.
Usually, wed expect there not to be any non-zero solution to this problem, but the t multiplying
the ut makes something odd happen for t = 0, so specifying the initial condition there is likely to
cause something unexpected. Substitute u(x, t) = X(x)T (t) into the equation to obtain

tXT 0 = X 00 T + 2XT

or
tT 0 X 00
2= =
T X
So the problem for X(x) is

X 00 + X = 0 and X(0) = X() = 0.

This tells us that

= n2 and Xn (x) = sin nx , n = 1, 2, . . .
2
With = n , the equation for T becomes
tT 0
2 = n2 or tT 0 = (2 n2 )T
T
which is a separable (in the ODE sense) first-order equation so
dT dt 1
= so ln T = ln t + ln c
(2 n2 )T t 2 n2
which give us that
2
Tn (t) = cn t2n .
For n = 1, T1 (t) = c1 t and for any value of c1 , the product

X1 (x)T1 (t) = c1 t sin x

satisfies all the conditions of the problem. Therefore, there are infinitely many solutions.

Page 92, problem 2. We want to solve the wave equation utt = c2 uxx for t > 0 and 0 < x < `
with boundary conditions ux (0, t) = 0 and u(`, t) = 0, so the string is fixed at its right end and free
at its left. So we do the usual separation of variables and obtain:
T 00 X 00
= = and X 0 (0) = X(`) = 0
c2 T X
 15

(well be nave here and pretend we dont know that is not going to be negative, just to practice
proving that it cant be).

We need to find the eigenvalues of X 00 X = 0 with X 0 (0) = X(`) = 0. Since the solution is
different in this case, we check the possibility of = 0 first: If = 0 the X 00 = 0, so X = ax + b.
Then X 0 (0) = 0 tells us that a = 0 and X(`) = 0 tells us that b = 0, so zero is not an eigenvalue for
this problem.

For 6= 0we have that X = c1 e
x
+ c2 e x . Take the derivative of this and set x = 0 to
0
get X (0) = c1 c2 = 0. Since were assuming
6= 0,this implies that c1 = c2 , so lets call
their

common value c so that at this point, X = c(e x
+ e x ). Next, set x = ` and we get that
` `
c(e + c2 e ) = 0, in other words (assuming c 6= 0 so that we really have an eigenfunction)
1
`
e = which implies e2 `
= 1
e `

which means that 2 ` must be an odd multiple of i. Therefore

 
(2n + 1)i 2n + 1
= which implies Xn (x) = cn cos x n = 0, 1, . . .
2` 2`

And with = (2n + 1)2 2 /(2`)2 , we can solve the T equation and get
   
2n + 1 2n + 1
Tn (t) = an cos ct + bn ct .
2` 2`

Now let An = cn an and Bn = cn bn and we can write the solution as

      
X 2n + 1 2n + 1 2n + 1
u(x, t) = An cos ct + Bn sin ct cos x
n=0
2` 2` 2`

which awaits determination of the An and Bn using initial conditions.