Anthony Varilly
Math 112, Spring 2002
1. End of Chapter 8, Exercise 24. Give an example to show that the following is not equivalent
to the integrability of f . For any > 0, there is a > 0 such that if P is any partition into
rectangles S1 , . . . , SN with sides < , there exist x1 S1 , . . . , xN SN such that
n
X
f (xi ) v (Si ) I < .
i=1
Solution. Assume that the condition is equivalent to the integrability of f . Consider the
function f : [0, 1] R, where f (x) = 0 if x is irrational and f (x) = 1 if x is rational. Then
for any partition P into rectangles S1 , . . . , SN with sides < , there exist irrational points
xi Si at which f (xi ) = 0. Thus
n
X
f (xi ) v (Si ) = 0
i=1
Then
n
X
f (xi ) v (Si ) I = |I| < ,
i=1
for any > 0, if and only if I = 0. This implies that f is integrable with integral zero. But as
we showed in class, f is not Riemann integrable since L(f, P ) = 0 whereas U (f, P ) = 1.
2. End of Chapter
8, Exercise 36. Prove that limn (n!)1/n /n = e1 by considering Riemann
R1
sums for 0 log xdx based on the partition 1/n < 2/n < < 1.
Solution. We first rewrite
(n!)1/n
n
n (n 1) (n 2) 1 1/n
=
=
n n n
1
2
1 1/n
=
1 1
1
n
n
n
!
n1
k
1X
= exp
log 1
n
n
n!
nn
1/n
k=0
log(1 x)dx =
log y dy = lim
0
log y dy
= lim 1 ( log )
0
0
R1
Therefore 0log(1
x)dx = 1. Since f (x) = log(1 x) is integrable on [0, 1], L (f, Pn ) =
P
1 n1
k
log 1
must approach 1 as n . Hence we conclude that
n k=0
n
(n!)1/n
= e1
n
n
lim
as desired.
3. End of Chapter 8, Exercise 38. Let f : [0, 1] R be defined by
0 if x is irrational
f (x) =
1/q if x = p/q
where p, q 0 with no common factor. Show that f is integrable, and compute
R1
0
f.
Solution. We first show that f is discontinuous at every rational number in [0, 1]. Let x
[0, 1] Q and choose < 1/q. There exists an irrational number y such that |x y| < ,
which implies that
|f (x) f (y)| = |1/q 0| = |1/q| <
Thus we have produced an for which there is no such that |f (x) f (y)| < when
|x y| < .
We now show that f is continuous at every irrational number in [0, 1]. Let c be an irrational
number in [0, 1] and > 0. Then there is a natural number n such that 1/n < . If we choose
small enough that the interval (c , c + ) contains no rational numbers with denominator
less than n, then it follows that for x in this interval we have |f (x) f (c)| = |f (x)| 1/n < .
Thus we have shown that f is discontinuous on D(f ) = [0, 1] Q Q. Since Q is countable,
we know that D(f ) is also countable. Since the interval [0, 1] is bounded and has volume,
and since f is bounded by 1/1 = 1, we have satisfied the conditions of Corollary 8.3.3, from
which it follows that f is integrable.
R1
Finally we show that 0 f = 0. Take any partition Pn = {x0 , x1 , . . . xn } and consider dj =
inf {f (x) : x [xj1 , xj ]}. Since every subinterval [xj1 , xj ] must contain irrational numbers,
it follows that dj = 0. Then
Z 1
f = sup {L (f, Pn )} = 0
0
and since f was shown to be integrable, the upper and lower integrals must agree.
4. Exercise 10.1.2. Show that
Z b
2
Z b
f (x)dx (b a)
|f (x)|2 dx,
a
and deduce that a square integrable function on [a, b], continuous on (a, b), is also integrable.
Is the converse true?
Solution. We apply the Cauchy-Schwarz inequality to functions f, g : [a, b] C, and we take
g(x) = 1 for all x [a, b]. Then
|hf, gi| kf k kgk
implies
Z
f (x)g(x)dx
2
Z
|f (x)| dx
Z
|g(x)| dx
which leads to
Z
2
Z b
Z b
2
f (x)dx
|f (x)| dx
dx
= (b a)
|f (x)|2 dx
Rb
a
Rb
a
f (x)dx converges
The converse, however, is not true. Consider the function f (x) = 1/ x. Then f is continuous
on (0, 1), and
Z 1
Z 1
1
dx = 2.
f (x)dx =
x
0
0
But
Z 1
Z 1
dx
|f (x)|2 dx =
= lim (log 1 log ) =
0
0
0 x
Therefore f is not square integrable even though f is integrable.
5. Exercise 10.2.2. Let g0 , g1 , g2 , . . . be linearly independent vectors in an inner product space.
Inductively define
n1
h0 = g0 , 0 =
X
hn
h0
, . . . , hn = gn
hgn , k ik , n =
....
kh0 k
khn k
k=0
Show that 0 , 1 , 2 , . . . are orthonormal. Why must we assume that the gs are linearly
independent?
3
since
hk , k i =
hk
hk
,
khk k khk k
hhk , hk i
=1
khk k2
j1
X
hgj , n in k =
6 0
n=0
r
sin nx
1
2 1
1
2
`
r
r
r 2
2 1
2n
2
2n
cos mx
2 1
2m
sin
cos
x=
sin
x. The functions
become
x=
`
`
`
`
`
`
r
2
2m
cos
x.
`
`
(c) Write the Fourier series of f for the families obtained in (b).
Solution.
f
* r
+r
X
1
1
2
2n
2
2n
+
=
f,
f,
sin
x
sin
x
`
`
`
`
`
` n=1
* r
+r
X
2
2m
2
2m
+
f,
cos
x
cos
x
`
`
`
`
m=1
1
1
2X
2n
2n
+
=
f,
f, sin
x sin
x
`
`
`
` ` n=1
2X
2m
2m
+
f, cos
x cos
`
`
`
m=1
f
n (u)du
` 2 0
2
r Z
r
`
`
`
=
g(u)n (u)du =
hg, n i
2 0
2
If the n in (a) are complete, then by Parsevals relation, we have
n=0
=
=
2
Z `
f `u du = 2
|f (y)|2 dy
2
`
0
0
2
2
2
hf, f i =
kf k
`
`
n=0
` X
` 2
|hf, n i| =
|hg, n i|2 =
kf k2 = kf k2
2
2 `
2
n=0
7. Exercise 10.2.4. Assume for the moment that the functions 1/ 2, (sin nx) / , (cos mx) /
are complete on the interval [0, 2].
(a) Apply this to the function x to show that x = 2
1
x,
2
X
1
sin nx sin nx X
cos mx cos mx
+
x,
+
x,
2 n=1
m=1
0
n
Z 2
cos mx
1
x cos(mx)dx = 0
x,
=
0
It follows that
x =
X 1 2 sin nx
1 4 2
1
2 2
2 n=1 n
= 2
X
(sin nx)
n=1
2
(b) Using
P the 2Fourier coefficients found in (a), apply Parsevals relation to show that /6 =
n=1 1/n .
n=1
Since
2
kxk =
Z
0
x2 dx =
8 3
3
we obtain
8 3
4 4 X 1 4 2
=
+
3
2
n2
n=1
or
8 3
2 3 X 4 2
=
+
3
3
n2
n=1
which leads to
2 X 1
=
6
n2
n=1
n=1 1/n
4.
2
n=1
m=1
0
2
1 2x
n2 x2 2
1 4 2
=
sin
nx
cos
nx
=
n3
n2
n
0
Z 2
cos mx
1
x2 cos(mx)dx
x2 ,
=
0
2
2x
m2 x2 2
1 4
1
=
cos mx +
sin mx
=
2
3
m
m2
m
0
Since
2 2
kx k =
x4 dx =
32 5
5
we obtain
32 5
5
1 8 3
2 3
2
2 X
X
1 4 2
1 4 2
+
+
n
m2
n=1
m=1
X
X
32 5
1
1
3
+ 16
+ 16
9
n2
m4
n=1
32 5
9
16 5
6
m=1
X
1
+ 16
m4
m=1
4 X 1
=
.
90
m4
n
X
cos k =
k=1
by using
ei
ei2
+ +
ein
ei 1 ein
=
.
1 ei
k
P
Solution. Since cos k = Re ei , we can rewrite nk=1 cos k as
" n
#
"
#
n
X
X k
ei 1 ein
i
cos k = Re
e
= Re
1 ei
k=1
k=1
2 (1 cos )
(1 ei ) (1 ei ) ei/2
"
#
ei(n+1/2)
1
= + Re i/2
2
e
ei/2
cos (n + 1/2) + i sin (n + 1/2)
1
= + Re
2
2i sin (/2)
1
i cos (n + 1/2) + sin (n + 1/2)
= + Re
2
2 sin (/2)
1 1 sin (n + 1/2)
= +
2 2
sin (/2)
Therefore
2
n
X
k=1
cos k =
as desired.