You are on page 1of 272

DEPARTMENT OF MATHEMATICS

NATIONAL UNIVERSITY OF SINGAPORE


2012-13 SEMESTER I
MA3110 MATHEMATICAL ANALYSIS II
1. The Real Number System
Notation:
N = set of all natural numbers (positive integers)
= 1, 2, 3, . . .
Z = set of all integers = . . . , 3, 2, 1, 0, 1, 2, 3, . . .
Q = set of all rational numbers =
m
n
: m, n Z, n ,= 0
R = set of all real numbers
An irrational number is a real number that is not a
rational number.
1.1 Integers
Axiom (Well Ordering Property of N):
Every nonempty subset S of N has a least element, i.e.,
there exists m S such that m n for every n S.
1
2 MA3110 MATHEMATICAL ANALYSIS II
Theorem 1.1. (Principle of Mathematical Induction)
Let S be a nonempty subset of N such that
(i) 1 S, and
(ii) for every k N, k S = k + 1 S.
Then S = N.
Proof. Suppose that S ,= N. Then NS ,= . By Well
Ordering Property of N, there exists a least element
m N S. Since 1 S, we have m ,= 1 and so
m 1 N. Since m is the smallest element not in
S, the number m 1 is in S. But by (ii), then m =
(m1) + 1 S, which is a contradiction.
Remark. Theorem 1.1 can be rephrased as:
For each n N, let P(n) be a statement about n.
Suppose that
(i) P(1) is true, and
(ii) for every k N, if P(k) is true, then P(k + 1) is
true.
Then P(n) is true for every n N.
MA3110 MATHEMATICAL ANALYSIS II 3
1.2 Real numbers
The set R of all real numbers has the following proper-
ties:
(1) Field properties: the four arithmetic operations
+, , , /.
(2) Order properties: We assume there exists a nonempty
subset P (called the set of positive real numbers)
of R satisfying:
(i) If a, b P, then a + b P.
(ii) If a, b P, then ab P.
(iii) (Trichotomy Property) For every a R,
exactly one of the following holds:
a P, a = 0, a P.
(Hence Ris the disjoint union of P, 0, P,
where P = a : a P).
Using the set P, dene the order relations <, >, ,
as follows:
If a P, then a is positive and write a > 0.
If a P 0, then a is nonnegative and write a 0.
If a P, then a is negative and write a < 0.
4 MA3110 MATHEMATICAL ANALYSIS II
If a P 0, then a is nonpositve and write a 0.
If a, b R and a b P, then write a > b or b < a.
If a, b R and a b P 0, then write a b or
b a.
Then one can derive those familiar rules of inequali-
ties for real numbers (Ex.). In particular:
(a) For any a, b R, exactly one of the following holds:
a < b, a = b, a > b.
(b) If a R and a ,= 0, then a
2
> 0.
(c) n > 0 for every n N.
Consequently, one may speak of the maximum and
minimum of a nite set of real numbers. In partic-
ular, we dene
[a[ = maxa, a for all a R.
Then we have the triangle inequality:
[a + b[ [a[ + [b[ for all a, b R.
The ordered eld R also has the property called com-
pleteness which we shall describe. First, some deni-
tions:
MA3110 MATHEMATICAL ANALYSIS II 5
Denition. Let A be a nonempty subset of R.
(1) A number M R is an upper bound of A if
a M for all a A.
If A has an upper bound, then we say that it is
bounded above.
(2) A number m R is a lower bound of A if m a
for all a A.
If A has a lower bound, then we say that it is
bounded below.
(3) If A is both bounded below and bounded above,
then we say that it is bounded.
Remark. Note that a set may have many dierent
upper bounds and lower bounds.
Example 1.1.
(1) Let A = x R : 0 < x < 1.
Then 1 is an upper bound of A, so is 5.
1 is a lower bound of A, so is 0.
(2) Let B =
n
n+1
: n N.
Then 1/2 is a lower bound of B and 1 is an upper
bound of B.
6 MA3110 MATHEMATICAL ANALYSIS II
Denition. Let A be a nonempty subset of R.
(1) A number M R is called the least upper bound
or supremum of A if
(a) M is an upper bound of A, and
(b) M u for any upper bound u of A.
The supremum of A is denoted by sup A.
(2) A number m R is called the greatest lower
bound or inmum of A if
(a) m is a lower bound of A, and
(b) m l for any lower bound l of A.
The inmum of A is denoted by inf A.
Remark. The supremum (respectively, inmum) of a
set, if it exists, is unique.
Lemma 1.2. Let A be a nonempty subset of R.
(i) An upper bound u of A is the supremum of
A if and only if for every > 0, there exists
a A such that u < a.
(ii) A lower bound v of A is the inmum of A if
and only if for every > 0, there exists b A
such that b < v + .
MA3110 MATHEMATICAL ANALYSIS II 7
Completeness Axiom of R:
Every nonempty subset of R that is bounded above has
a supremum. (Equivalently, every nonempty subset of
R that is bounded below has an inmum.)
The Completeness Axiom is a fundamental property
assumed of R. All the main results in elementary anal-
ysis, e.g., the intermediate value theorem, the extreme
value theorem, depend directly on it.
Proposition 1.3. (Archimedean Property) The set
N is not bounded above in R, i.e., for every x R,
there exists n
x
N such that x < n
x
.
Proof. Suppose the contrary that N is bounded above.
By the Completeness Axiom, u := sup N exists in R.
Now u 1 is not an upper bound of N, hence there
exists n N such that n > u 1.
Then n + 1 > u.
But n + 1 N. This contradicts that u is an upper
bound of N.
8 MA3110 MATHEMATICAL ANALYSIS II
Example 1.2. Recall in Example 1.1:
B =
n
n+1
: n N.
We know that 1/2 is a lower bound of B and 1 is an
upper bound of B.
We now show that inf B = 1/2 and sup B = 1.
(1a) 1/2 is a lower bound of B.
(1b) Suppose l is a lower bound of B.
Then l b for any b B.
Observe that 1/2 B.
Hence l 1/2.
(1a) and (1b) together show that inf B = 1/2.
(2a) 1 is an upper bound of B.
(2b) Let u be an upper bound of B. We have to show
that u 1.
Suppose otherwise that u < 1. Then 1 u > 0.
By Archimedean property, there exists n N such that
n >
1
1u
. Then n + 1 >
1
1u
.
Thus u < 1
1
n+1
=
n
n+1
.
This contradicts that u is an upper bound of B.
(Ex.: Use Lemma 1.2 to give another proof.)
MA3110 MATHEMATICAL ANALYSIS II 9
Proposition 1.4.

2 is an irrational number.
Proof. Suppose

2 is a rational number.
Then

2 = m/n, where m and n are positive integers.


We may assume that m and n have no common factors
except 1. Now
m
2
= 2n
2
.
Hence m
2
is even. Therefore, m is even (why?).
Write m = 2k for some positive integer k. Then
4k
2
= 2n
2
= 2k
2
= n
2
.
Thus n
2
is even and hence so is n.
Then m and n are both even and hence have 2 as a
common factor, contrary to the assumption.
(Have we completed the proof? Ex.)
10 MA3110 MATHEMATICAL ANALYSIS II
Theorem 1.5. (Density of rationals) Let x and y
be real numbers with x < y. Then there exists a
rational number r such that x < r < y.
Proof. Without loss of generality, we may assume that
x > 0 (why?). By Archimedean property, there exists
n N such that
1
n
< y x. Then
ny nx > 1, or nx < nx + 1 < ny.
There exists m N such that m1 nx < m (Ex.).
Hence nx < m nx + 1 < ny, i.e., x <
m
n
< y.
Theorem 1.6. (Density of irrationals) Let x and y
be real numbers with x < y. Then there exists an
irrational number q such that x < q < y.
(Ex.)
MA3110 MATHEMATICAL ANALYSIS II 11
1.3 Finite and innite sets
Denition. Let S be a set.
(1) S is nite if it is empty or if there is a bijec-
tion with domain S and range an initial segment
1, 2, . . . , n of N. S is innite if it is not nite.
(2) S is denumerable or enumerable if there is a bi-
jection from S onto N.
(3) S is countable if it is nite or denumerable. S is
uncountable if it is not countable.
Remark. A denumerable set is countably innite.
Proposition 1.7. A set S is countable if and only
if there is an injection from S into N.
Proposition 1.8.
(i) Any subset of a nite (respectively countable)
set is nite (respectively countable).
(ii) The union of a nite (respectively countable)
collection of nite (respectively countable) sets
is nite (respectively countable).
12 MA3110 MATHEMATICAL ANALYSIS II
Theorem 1.9.
(i) The set N N is countable.
(ii) The set Q is countable.
(iii) The set R (and hence the set of all irrational
numbers) is uncountable.
Proof. (ii) The set Q can be written as
Q =
m
n
: m Z, n N
=
nN
A
n
where A
n
=
m
n
: m Z is countable for each n N.
By Proposition 1.8, Qis countable. Alternatively, using
diagonal process, we can enumerate Q =
nN0
A
n
:
A
0
= 0
A
1
A
0
=
1
1
,
1
1
,
2
1
,
2
1
,
3
1
,
3
1
,
A
2
A
0
=
1
2
,
1
2
,
2
2
,
2
2
,
3
2
,
3
2
,
A
3
A
0
=
1
3
,
1
3
,
2
3
,
2
3
,
3
3
,
3
3
,
.
.
.
.
.
.
MA3110 MATHEMATICAL ANALYSIS II 13
(iii) (Cantors diagonal argument) Suppose the set I =
x R : 0 x 1 is countable (it must be in-
nite). Then we can enumerate the elements of the set I
as x
1
, x
2
, x
3
, . . .. Expressing these numbers in decimal
representations:
x
1
= 0. x
11
x
12
x
13
x
1n

x
2
= 0. x
21
x
22
x
23
x
2n

.
.
.
.
.
.
x
n
= 0. x
n1
x
n2
x
n3
x
nn

.
.
.
.
.
.
Dene a real number y = 0. y
1
y
2
y
3
y
n
in I
such that
y
n
=
_
7 if x
nn
4
2 if x
nn
5.
Therefore for all n, y
n
,= x
nn
. Hence for all n, y ,= x
n
,
since the n-th decimal place y
n
of y is dierent from
the n-th decimal place x
nn
of x
n
. But y is in I, leading
to a contradiction. Hence I is uncountable, and so is
the bigger set R by Proposition 1.8.
14 MA3110 MATHEMATICAL ANALYSIS II
1.4 Intervals
There are nine types of intervals in R: Let a, b R.
(a, b) = x R : a < x < b
[a, b] = x R : a x b
[a, b) = x R : a x < b
(a, b] = x R : a < x b
(a, ) = x R : x > a
(, b) = x R : x < b
[a, ) = x R : x a
(, b] = x R : x b
(, ) = R
In particular, [0, 1] = x R : 0 x 1 is the
closed unit interval.
MA3110 MATHEMATICAL ANALYSIS II 15
The next result gives a characterization of intervals.
Proposition 1.10. Let S be a subset of R with at
least two points and the property:
x, y S and x < y = [x, y] S.
Then S is an interval.
Proof. There are four cases:
(1) Suppose S is bounded.
Claim : S is of the form (a, b) or [a, b] or [a, b) or
(a, b].
Let a = inf S, b = sup S. By denition of inf
and sup, for every x S, a x b. Hence
S [a, b].
If a < z < b, then z is not a lower bound of S.
Therefore there exists x S such that x < z.
Also, z is not an upper bound of S. Therefore
there exists y S such that z < y.
Hence x < z < y. By assumption, z S.
Hence (a, b) S. So S must be one of the above
intervals.
16 MA3110 MATHEMATICAL ANALYSIS II
(2) Suppose S is bounded above but not bounded be-
low.
Then S is of the form (, b) or (, b]. (Ex.)
(3) Suppose S is bounded below but not bounded
above.
Then S is of the form (a, ) or [a, ). (Ex.)
(4) Suppose S is neither bounded above nor bounded
below.
Then S = (, ) = R. (Ex.)

MA3110 MATHEMATICAL ANALYSIS II 17


A sequence (I
n
)
nN
of intervals is nested if
I
1
I
2
I
n
I
n+1

Theorem 1.11. (Nested interval property)
If I
n
= [a
n
, b
n
], n N, is a nested sequence of closed
bounded intervals, then I
n
,= . Furthermore, if
infb
n
a
n
: n N = 0, then I
n
consists of only
one element.
Proof. Since I
n
I
1
for all n, the sets a
n
: n N
and b
n
: n N are bounded.
Claim : a
n
b
k
for all n, k N.
If n k, then I
n
I
k
; so a
n
a
k
b
k
.
If n > k, then I
k
I
n
; so a
n
b
n
b
k
.
Let a = supa
n
: n N and b = infb
n
: n N.
Then a b
k
for all k N
and so a b.
Claim : I
n
= [a, b]. This is because
x I
n
a
n
x b
n
for all n
a x and x b
x [a, b].
18 MA3110 MATHEMATICAL ANALYSIS II
Suppose infb
n
a
n
: n N = 0. Then for all > 0,
there exists n N such that
0 b a b
n
a
n
< .
Therefore b a = 0, i.e, a = b.
Hence I
n
= a.
MA3110 MATHEMATICAL ANALYSIS II 19
2. Sequences of real numbers
2.1 Sequences and limit theorems
Denition. A sequence of real numbers is a function
f : N R. Writing a
n
= f(n) for n N, we often
denote a sequence by (a
n
)

n=1
and call a
n
the n-th term
of the sequence.
Denition. Let (a
n
)

n=1
be a sequence of real num-
bers.
(1) We say that a real number L is a limit of (a
n
)

n=1
(or (a
n
)

n=1
converges to L) if for every > 0,
there exists N N such that [a
n
L[ < when-
ever n N. In this case, we write
lim
n
a
n
= L, or a
n
L as n .
(2) The sequence (a
n
)

n=1
is said to converge (or is
convergent) if it converges to some L R.
Otherwise, it is said to diverge (or is divergent).
(3) The sequence (a
n
)

n=1
diverges to +
(write lima
n
= +, or a
n
+)
if for every a R, there exists N N such that
a
n
a whenever n N.
20 MA3110 MATHEMATICAL ANALYSIS II
The sequence (a
n
)

n=1
diverges to
(write lima
n
= , or a
n
)
if for every b R, there exists N N such that
a
n
b whenever n N.
Example 2.1. The sequence (1/n)

n=1
converges to 0.
Solution. Given > 0, choose N N such that
N > 1/.
If n N, then 1/n 1/N < .
Thus [1/n 0[ = 1/n < . This proves the claim.
Example 2.2. Dene a sequence (a
n
)

n=1
by
a
n
=
_
_
_
2
2n+3
if n is odd
3
n+5
if n is even.
Determine if the sequence (a
n
)

n=1
is convergent.
MA3110 MATHEMATICAL ANALYSIS II 21
Solution. Given > 0, choose N N with N > 3/.
If n N, then
[a
n
0[ =
_
_
_
[
2
2n+3
[ if n is odd
[
3
n+5
[ if n is even

_
_
_
2
2n
if n is odd
3
n
if n is even

3
n

3
N
< .
Hence lim
n
a
n
= 0.
Example 2.3. Show that the sequence
((1)
n
n/(n + 1))

n=1
does not converge to 1.
Solution.We need to show that there exists > 0 so
that for any N N, there exists n N such that
[(1)
n
n
n + 1
1[ .
Take = 1. For any N N, there exists n N such
that n is odd. For this n, (1)
n
= 1. Hence
[(1)
n
n
n + 1
1[ = [
n
n + 1
1[ = 1 +
n
n + 1
1.
22 MA3110 MATHEMATICAL ANALYSIS II
Proposition 2.1. (Uniqueness of limits) If a real se-
quence (a
n
)

n=1
converges to both L
1
and L
2
, then
L
1
= L
2
.
Proof. Suppose on the contrary that L
1
,= L
2
.
Let = [L
1
L
2
[/2 > 0.
Since (a
n
)

n=1
converges to L
1
, there exists N
1
N such
that
[a
n
L
1
[ < whenever n N
1
.
Similarly, there exists N
2
N such that
[a
n
L
2
[ < whenever n N
2
.
If n maxN
1
, N
2
, then
[L
1
L
2
[ [L
1
a
n
[ + [a
n
L
2
[
< + = [L
1
L
2
[.
Hence we must have L
1
= L
2
.
MA3110 MATHEMATICAL ANALYSIS II 23
Proposition 2.2. A convergent real sequence (a
n
)

n=1
is necessarily bounded, i.e., there exists a positive
real number K such that [a
n
[ K for every n N.
Proof. Suppose (a
n
)

n=1
converges to L.
There exists N N such that [a
n
L[ < 1 for all
n N.
If n N, then by triangle inequality
[a
n
[ [a
n
L[ + [L[ < 1 + [L[.
Thus for all n N,
[a
n
[ max[a
1
[, [a
2
[, . . . , [a
N1
[, 1 + [L[.
Set K = max[a
1
[, [a
2
[, . . . , [a
N1
[, 1 + [L[.
Then for all n, [a
n
[ K = K a
n
K.
So (a
n
)

n=1
is bounded below (by K) and bounded
above (by K) and thus is bounded.
24 MA3110 MATHEMATICAL ANALYSIS II
Proposition 2.3. Let (a
n
)

n=1
and (b
n
)

n=1
be real se-
quences that converge to a and b respectively. Then
(i) (a
n
+ b
n
)

n=1
converges to a + b.
(ii) (a
n
b
n
)

n=1
converges to a b.
(iii) (c a
n
)

n=1
converges to c a for every c R.
(iv) (a
n
b
n
)

n=1
converges to ab.
(v) If b
n
,= 0 for all n and b ,= 0, then (a
n
/b
n
)

n=1
converges to a/b.
(vi) If a
n
b
n
for all n, then a b.
Remark. It is important when using Proposition 2.3
that the limits of (a
n
)

n=1
and (b
n
)

n=1
both exist.
For instance, it is WRONG to say that
lim
n
a
n
n
= lim
n
a
n
lim
n
1
n
= lim
n
a
n
0 = 0
without knowing what the sequence (a
n
)

n=1
is.
MA3110 MATHEMATICAL ANALYSIS II 25
Proposition 2.4. (Squeeze Theorem) Let (a
n
)

n=1
,
(b
n
)

n=1
and (c
n
)

n=1
be real sequences such that
a
n
b
n
c
n
for all n.
If (a
n
)

n=1
and (c
n
)

n=1
both converge to the same
number L, then (b
n
)

n=1
also converges to L.
Proof. Let > 0 be given.
Choose N
1
N such that [a
n
L[ < whenever
n N
1
.
Choose N
2
N such that [c
n
L[ < whenever
n N
2
.
Set N = maxN
1
, N
2
. If n N, then
b
n
L c
n
L [c
n
L[ <
and
< a
n
L b
n
L.
Thus
[b
n
L[ <
whenever n N. Hence (b
n
)

n=1
converges to L.
26 MA3110 MATHEMATICAL ANALYSIS II
2.2 Monotone sequences and subsequences
Denition. A real sequence (a
n
)

n=1
is said to be in-
creasing if a
n
a
n+1
for all n. It is strictly increasing
if is replaced by <. Decreasing and strictly de-
creasing are dened by replacing by and >
respectively. A sequence that belongs to one of these
four categories is said to be monotone.
Proposition 2.5. (Monotone Convergence Theorem)
A bounded monotone real sequence converges. More
precisely, a bounded increasing sequence converges
to its supremum and a bounded decreasing sequence
converges to its inmum.
Proof. Let (a
n
)

n=1
be a bounded increasing real se-
quence and let L = sup
n
a
n
.
Let > 0 be given.
Since L is not an upper bound of (a
n
)

n=1
, there
exists N N such that a
N
> L . If n N, then
a
n
a
n1
a
n2
a
N
> L .
MA3110 MATHEMATICAL ANALYSIS II 27
On the other hand, L is an upper bound of (a
n
)

n=1
implies that L a
n
for all n.
So if n N,
L < a
n
L
= < a
n
L 0 <
= [a
n
L[ < .

28 MA3110 MATHEMATICAL ANALYSIS II


Let (a
n
)

n=1
be a bounded real sequence. Dene
limsup
n
a
n
:= inf
mN
sup
nm
a
n
, liminf
n
a
n
:= sup
mN
inf
nm
a
n
.
Dene
b
m
:= sup
nm
a
n
= supa
m
, a
m+1
, . . . .
Since a
m
, a
m+1
, . . . a
m+1
, a
m+2
, . . . ,
b
m
= supa
m
, a
m+1
, . . . supa
m+1
, a
m+2
, . . . = b
m+1
.
So (b
m
)

m=1
is decreasing.
If L and M are a lower bound and an upper bound
of (a
n
)

n=1
respectively, then L and M are also a lower
bound and an upper bound of (b
m
)

m=1
. (Ex.)
So (b
m
)

m=1
is bounded.
By Proposition 2.5, (b
m
)

m=1
converges to inf
m
b
m
. Thus
limsup
n
a
n
:= inf
m
b
m
= lim
m
b
m
.
Similarly, if we dene
c
m
:= inf
nm
a
n
= infa
m
, a
m+1
, . . . ,
then (c
m
)

m=1
is increasing and bounded. Therefore
liminf
n
a
n
:= sup
m
c
m
= lim
m
c
m
.
MA3110 MATHEMATICAL ANALYSIS II 29
Example 2.4. Consider the sequence ((1)
n
(n+1)/n)

n=1
,
i.e.,
(a
m
)

m=1
= (2,
3
2
,
4
3
,
5
4
,
6
5
, . . . ).
(b
m
)

m=1
= (
3
2
,
3
2
,
5
4
,
5
4
, . . . ).
(c
m
)

m=1
= (2,
4
3
,
4
3
,
6
5
, . . . ).
Then
limsup
n
a
n
= inf
m
b
m
= lim
m
b
m
= 1,
liminf
n
a
n
= sup
m
c
m
= lim
m
c
m
= 1.
30 MA3110 MATHEMATICAL ANALYSIS II
Proposition 2.6. Let (a
n
)

n=1
be a bounded real se-
quence. Then
(i) liminf
n
a
n
limsup
n
a
n
.
(ii) (a
n
)

n=1
converges to L R if and only if
liminf
n
a
n
= L = limsup
n
a
n
.
Proof. Dene (b
m
)

m=1
and (c
m
)

m=1
as above.
(i) Claim : For any j and k, c
j
b
k
.
In fact, given j and k, let i = maxj, k. Then a
i
is in
both the sets a
j
, a
j+1
, . . . and a
k
, a
k+1
, . . . . So
c
j
= infa
j
, a
j+1
, . . . a
i
supa
k
, a
k+1
, . . . = b
k
.
This proves the claim.
The claim shows that for each j, c
j
is a lower bound of
the sequence (b
m
)

m=1
.
Thus c
j
inf
m
b
m
= limsup
n
a
n
for any j.
But this means that limsup
n
a
n
is an upper bound of
the sequence (c
m
)

m=1
.
Thus liminf
n
a
n
= sup
m
c
m
limsup
n
a
n
.
MA3110 MATHEMATICAL ANALYSIS II 31
(ii) Assume that liminf
n
a
n
= L = limsup
n
a
n
.
For any m,
c
m
= infa
m
, a
m+1
, . . . a
m
supa
m
, a
m+1
, . . . = b
m
.
Since (c
m
)

m=1
and (b
m
)

m=1
both converge to L (see dis-
cussion preceding the proposition), (a
m
)

m=1
also con-
verges to L by the Squeeze Theorem.
Conversely, assume that (a
n
)

n=1
converges to L.
Because of (i), it is enough to show that
limsup
n
a
n
liminf
n
a
n
.
Assume that this is not the case. Then
:= limsup
n
a
n
liminf
n
a
n
> 0.
Choose N so that [a
n
L[ <

3
whenever n N.
Then L

3
< a
n
< L +

3
if n N.
Thus, L +

3
is an upper bound of a
N
, a
N+1
, . . . .
Hence b
N
= supa
N
, a
N+1
, . . . L +

3
.
Similarly, c
N
L

3
. (Ex.)
32 MA3110 MATHEMATICAL ANALYSIS II
Then
limsup
n
a
n
liminf
n
a
n
= inf
m
b
m
sup
m
c
m
b
N
c
N
(L +

3
) (L

3
)
=
2
3
< .

MA3110 MATHEMATICAL ANALYSIS II 33


Denition. Let (a
n
)

n=1
be a real sequence. A subse-
quence of (a
n
)

n=1
is a sequence of the form (a
n
k
)

k=1
,
where n
1
< n
2
< n
3
< . . . .
Proposition 2.7. Let (a
n
)

n=1
be a real sequence.
(i) If (a
n
)

n=1
is bounded, then so is any subse-
quence.
(ii) If (a
n
)

n=1
converges to L, then any subsequence
also converges to L.
34 MA3110 MATHEMATICAL ANALYSIS II
Theorem 2.8. (Bolzano-Weierstrass Theorem) Ev-
ery bounded real sequence has a convergent subse-
quence.
Proof. Let (a
n
)

n=1
be a bounded real sequence.
As before, dene b
m
= sup
nm
a
n
for all m N.
It was observed that (b
m
) converges, to a real number
L say.
We choose n
1
< n
2
< one at a time as follows.
First, observe that b
1
1 is not an upper bound of
a
1
, a
2
, . . . .
So there exists n
1
N such that a
n
1
> b
1
1.
Surely a
n
1
b
1
as well. So b
1
1 < a
n
1
b
1
.
Assume that we have chosen n
1
< n
2
< < n
k1
,
for k 2, such that (with n
0
= 0)
b
n
j1
+1

1
j
< a
n
j
b
n
j1
+1
, 1 j k 1.
b
n
k1
+1

1
k
is not an upper bound of a
n
k1
+1
, a
n
k1
+2
, . . . .
So there exists n
k
> n
k1
such that
a
n
k
> b
n
k1
+1

1
k
.
Also, a
n
k
b
n
k1
+1
.
Thus b
n
k1
+1

1
k
< a
n
k
b
n
k1
+1
.
MA3110 MATHEMATICAL ANALYSIS II 35
In this way, we choose a subsequence (a
n
k
)

k=1
of (a
n
)

n=1
.
Note that (b
n
k1
+1
)

k=1
is a subsequence of (b
n
)

n=1
.
By Proposition 2.7, (b
n
k1
+1
)

k=1
also converges to L.
Also (
1
k
)

k=1
converges to 0.
By Proposition 2.3, (b
n
k1
+1

1
k
)

k=1
converges to L
0 = L.
By the Squeeze Theorem, (a
n
k
)

k=1
converges to L.
Remark. An unbounded sequence of real numbers
may or may not have a convergent subsequence.
Example 2.5.
(1) The sequence (n)

n=1
is unbounded and has no con-
vergent subsequence. In fact, any subsequence of (n)

n=1
is unbounded and hence not convergent.
(2) The sequence (a
n
)

n=1
dened by
a
n
=
_
_
_
n if n is even
1 if n is odd
is unbounded. However, it has a convergent subse-
quence (a
2n1
)

n=1
.
36 MA3110 MATHEMATICAL ANALYSIS II
2.3 Cauchy sequences and completeness
Denition. A real sequence (a
n
)

n=1
is said to be a
Cauchy sequence if for every > 0, there exists N N
such that [a
m
a
n
[ < whenever m, n N.
Example 2.6. Show that ((1)
n
n/(n +1))

n=1
is not
a Cauchy sequence.
Proof. Take = 1.
For any N N, let m = N and n = N + 1.
Then m, n N.
[a
m
a
n
[ = [(1)
N
N
N + 1
(1)
N+1
N + 1
N + 2
[
= [
N
N + 1
(1)
N + 1
N + 2
[
=
N
N + 1
+
N + 1
N + 2

1
2
+
1
2
= 1.

Proposition 2.9. Every Cauchy sequence is bounded.


MA3110 MATHEMATICAL ANALYSIS II 37
Theorem 2.10. (Cauchys criterion) A real sequence
is convergent if and only if it is a Cauchy sequence.
Proof. Let (a
n
)

n=1
be a convergent real sequence.
Denote its limit by L. Let > 0 be given. Choose
N N such that [a
n
L[ <

2
whenever n N.
If m, n N,
[a
m
a
n
[ = [(a
m
L) + (L a
n
)[
[a
m
L[ + [L a
n
[
<

2
+

2
= .
So (a
n
)

n=1
is a Cauchy sequence.
Conversely, suppose (a
n
)

n=1
is a Cauchy sequence.
Given > 0, choose N such that
[a
m
a
n
[ <

2
if m, n N.
By Proposition 2.9, (a
n
) is bounded.
By Bolzano-Weierstass Theorem, it has a subsequence
(a
n
k
)

k=1
converging to some real number L.
Claim : (a
n
)

n=1
converges to L also.
38 MA3110 MATHEMATICAL ANALYSIS II
By the convergence of (a
n
k
)

k=1
to L,
there exists k
0
N such that [a
n
k
0
L[ <

2
.
Since n
k
0
k
0
N, for all n N, [a
n
a
n
k
0
[ <

2
.
Then for all n N,
[a
n
L[ [a
n
a
n
k
0
[ + [a
n
k
0
L[
<

2
+

2
= .
Hence (a
n
)

n=1
converges to L.

Remark. By Cauchys criterion, the sequence


((1)
n
n/(n + 1))

n=1
(Example 2.6) diverges.
MA3110 MATHEMATICAL ANALYSIS II 39
3. Limits of Functions
3.1 Limits of functions and limit theorems
Denition. Let A be a nonempty subset of R. A real
number c is a cluster point of A if for every > 0,
(c , c + ) (A c) ,= , i.e., there exists x in
A c such that [x c[ < .
Remark. The point c may or may not be in the set
A. Cluster points are also called accumulation points
or limit points.
Proposition 3.1. Let A be a nonempty subset of R.
A real number c is a cluster point of A if and only if
there exists a sequence (a
n
)

n=1
in Ac converging
to c.
Proof. () Suppose c is a cluster point of A. Then
n N, a
n
A c such that [a
n
c[ <
1
n
,
i.e., c
1
n
< a
n
< c +
1
n
.
By Squeeze Theorem, (a
n
)

n=1
converges to c.
() Let > 0. By the convergence of (a
n
)

n=1
(in
Ac) to c, N N such that 0 < [a
N
c[ < .
40 MA3110 MATHEMATICAL ANALYSIS II
Example 3.1.
Let A = (0, 1) = x R : 0 < x < 1.
(1) 0 is a cluster point of A even though 0 / A.
In fact, the sequence (1/(n+1))

n=1
is a sequence in
A that converges to 0 and 1/(n + 1) ,= 0 for all n.
(2) Morever, [0, 1] is the set of all cluster points of A.
MA3110 MATHEMATICAL ANALYSIS II 41
Example 3.2. The set N has no cluster points.
Proof 1: Use denition. (Ex.)
Proof 2: Use Proposition 3.1
Suppose on the contrary that N has a cluster point c.
Then there is a sequence (b
n
)

n=1
in N such that
(b
n
)

n=1
converges to c and b
n
,= c for all n.
Hence there exists N N such that [b
n
c[ < 1/2
whenever n N.
So, if n N, then b
n
is an integer and lies in the
interval (c 1/2, c + 1/2).
The interval (c 1/2, c + 1/2) is an open interval of
length 1 and so can contain at most one integer.
Thus all b
n
s are equal when n N.
Take this common value to be m.
Then (b
n
)

n=1
converges to c and also to m. (Ex.)
By uniqueness of limits, m = c.
Then b
n
= m = c whenever n N.
42 MA3110 MATHEMATICAL ANALYSIS II
Denition. Let Abe a nonempty subset of Rand let c
be a cluster point of A. Suppose that f is a real-valued
function dened on A and L R. We say that L is a
limit of f(x) as x approaches c (or simply, a limit of
f at c) if for every > 0, there exists > 0 such that
[f(x) L[ < for every x A and 0 < [x c[ < .
In this case, we also say that f converges to L at c,
and write lim
xc
f(x) = L, or f(x) L as x c.
If the limit of f at c does not exist (in R), we say that
f diverges at c.
Proposition 3.2. (Uniqueness of limits) If lim
xc
f(x) =
L
1
and also lim
xc
f(x) = L
2
, then L
1
= L
2
.
MA3110 MATHEMATICAL ANALYSIS II 43
Example 3.3. Show that lim
x3
x
2
= 9.
Proof. Given > 0, need to nd > 0 such that
if 0 < [x 3[ < , then [x
2
9[ < .
Since [x
2
9[ = [x 3[[x + 3[, we want an inequality
involving [x + 3[.
First place a restriction on , say 1.
Then 0 < [x 3[ < and 1
[x[ [x 3[ + 3 < + 3 4
[x + 3[ [x[ + 3 < 7
[x
2
9[ = [x 3[[x + 3[ < 7.
Hence put another restriction on : 7 , or

7
.
Therefore, take = min1,

7
.
If 0 < [x 3[ < , then by above arguments
[x
2
9[ < 7 .
44 MA3110 MATHEMATICAL ANALYSIS II
Example 3.4. Dene f : R R by
f(x) =
_

_
5x 2 if x < 1
4 if x = 1
2x + 1 if x > 1
Show that lim
x1
f(x) = 3.
Proof. Given > 0, choose = /5 > 0.
Suppose x R and 0 < [x 1[ < , then either
x (1 , 1) or x (1, 1 + ). We analyze the two
cases separately.
Case 1. x (1 , 1).
In this case
[f(x) 3[ = [(5x 2) 3[ = 5[x 1[ < 5 = .
Case 2. x (1, 1 + ).
In this case
[f(x) 3[ = [(2x + 1) 3[ = 2[x 1[ < 2 < .
Done.
MA3110 MATHEMATICAL ANALYSIS II 45
The following result gives the key connection between
limits of functions and limits of sequences.
Theorem 3.3. Let A be a nonempty subset of R
and let c be a cluster point of A. Suppose that f
is a real-valued function dened on A and L R.
Then the following conditions are equivalent.
(i) lim
xc
f(x) = L.
(ii) For every sequence (x
n
)

n=1
in Ac that con-
verges to c, the sequence (f(x
n
))

n=1
converges
to L.
Proof. (i) (ii): Suppose lim
xc
f(x) = L.
Let (x
n
)

n=1
be a sequence in A c converging to c.
Then x
n
,= c for all n.
Let > 0 be given. Choose > 0 such that
[f(x) L[ < whenever x A and 0 < [x c[ < .
Choose N Nsuch that [x
n
c[ < whenever n N.
Therefore, if n N, then x
n
Aand 0 < [x
n
c[ < .
Hence [f(x
n
) L[ < for every n N.
Thus (f(x
n
))

n=1
converges to L.
46 MA3110 MATHEMATICAL ANALYSIS II
(ii) (i): Suppose that lim
xc
f(x) ,= L. Then there
exists > 0 so that for any > 0, there exists x A,
0 < [x c[ < , and [f(x) L[ .
In particular, for any n N, there exists
x
n
A, 0 < [x
n
c[ < 1/n such that [f(x
n
) L[ .
Now (x
n
)

n=1
is a sequence in A c,
as 0 < [x
n
c[ implies x
n
,= c for all n.
Also, c 1/n < x
n
< c + 1/n for all n.
By the Squeeze Theorem, (x
n
)

n=1
converges to c.
On the other hand, (f(x
n
))

n=1
does not converge to L
since [f(x
n
) L[ for all n.
This contradicts (ii).
MA3110 MATHEMATICAL ANALYSIS II 47
Example 3.5. Dene f : R R by
f(x) =
_

_
x
2
if x < 0
3 if x = 0
1 + x if x > 0.
Show that lim
x0
f(x) does not exist.
Proof. Suppose on the contrary that the limit exists
and is equal to L.
By Theorem 3.3, if (x
n
)

n=1
is a sequence in R0 that
converges to 0, then (f(x
n
))

n=1
converges to L.
First consider the sequence (1/n)

n=1
.
It is dierent from 0 for all n and converges to 0.
So (f(1/n))

n=1
converges to L.
Since f(1/n) = 1 + 1/n, we conclude that L = 1.
Next, consider the sequence (1/n)

n=1
.
It is dierent from 0 for all n and converges to 0.
So (f(1/n))

n=1
converges to L.
Since f(1/n) = (1/n)
2
= 1/n
2
, taking limits
L = 0.
48 MA3110 MATHEMATICAL ANALYSIS II
Proposition 3.4. Let A be a nonempty subset of R
and let c be a cluster point of A. If f : A R has a
limit at c, then f is bounded on some interval about
c, i,e., there exist positive constants K and such
that [f(x)[ K whenever x A and 0 < [xc[ < .
Proposition 3.5. Let A be a nonempty subset of
R, let c be a cluster point of A and f, g : A R.
Suppose that lim
xc
f(x) = L and lim
xc
g(x) = M.
Then
(i) lim
xc
(f + g)(x) = L + M.
(ii) lim
xc
(f g)(x) = L M.
(iii) lim
xc
(kf)(x) = kL, for every k R.
(iv) lim
xc
(fg)(x) = LM.
(v) If g(x) ,= 0 for all x A c and M ,= 0,
then
lim
xc
f(x)
g(x)
=
L
M
.
(vi) If f(x) g(x) for all x A c,
then L M.
MA3110 MATHEMATICAL ANALYSIS II 49
Proposition 3.6. (Squeeze Theorem) Let A be a
nonempty subset of R, let c be a cluster point of
A and f, g, h : A R. If f(x) g(x) h(x)
for every x in A c and if lim
xc
f(x) = L =
lim
xc
h(x), then lim
xc
g(x) = L.
Proposition 3.7. Let A be a nonempty subset of
R and let c be a cluster point of A. If f : A R
has a positive limit at c, then f is positive on some
interval about c, i,e., there exists > 0 such that
f(x) > 0 for every x A and 0 < [x c[ < .
50 MA3110 MATHEMATICAL ANALYSIS II
3.2 One-sided limits
Denition. Let A be a nonempty subset of R and let
f : A R.
(1) Let c be a cluster point of A(c, ). We say that
a real number L is a limit of f(x) as x approaches
c from the right (or simply, a right-hand limit of f
at c) if for every > 0, there exists > 0 such that
[f(x) L[ < for every x A and 0 < x c < .
In this case, we write lim
xc+
f(x) = L, or
f(x) L as x c+.
(2) Let c be a cluster point of A(, c). We say that
a real number L is a limit of f(x) as x approaches
c from the left (or simply, a left-hand limit of f at
c) if for every > 0, there exists > 0 such that
[f(x) L[ < for every x A and 0 < c x < .
In this case, we write lim
xc
f(x) = L, or
f(x) L as x c.
(3) The limits lim
xc+
f(x) and lim
xc
f(x) (which
may or may not exist) are called one-sided limits of
f at c.
MA3110 MATHEMATICAL ANALYSIS II 51
Remark. Sometimes lim
xc
f(x) is called a two-sided
limit. Previous theorems on limits in Section 3.1 (with
obvious modications) remain valid if x c is re-
placed by x c+ (repsectively, by x c).
Theorem 3.8. Let c be a cluster point of both
A(c, ) and A(, c). Then lim
xc
f(x) exists
and equals a real number L if and only if
lim
xc+
f(x) = L = lim
xc
f(x).
52 MA3110 MATHEMATICAL ANALYSIS II
3.3 Innite limits
Denition. Let A be a nonempty subset of R, c a
cluster point of A and f : A R.
(1) We say that the limit of f(x) as x approaches c is
+ (or f(x) tends to + as x c) if for every
K > 0, there exists > 0 such that f(x) > K for
every x A and 0 < [x c[ < . In this case, we
write lim
xc
f(x) = +.
(2) We say that the limit of f(x) as x approaches c is
(or f(x) tends to as x c) if for every
K < 0, there exists > 0 such that f(x) < K for
every x A and 0 < [x c[ < . In this case, we
write lim
xc
f(x) = .
Remark. Similar denitions can be formulated for
innite one-sided limits:
lim
xc+
f(x) = +, lim
xc+
f(x) = ,
lim
xc
f(x) = +, lim
xc
f(x) = .
In all these cases, the limits do not exist in the previous
sense (of Section 3.1 and Section 3.2), as +and
are just symbols and they are not real numbers.
MA3110 MATHEMATICAL ANALYSIS II 53
3.4 Limits at innity
Denition. Let A be a nonempty subset of R and
f : A R.
(1) Suppose A is not bounded above (i.e., for every
M > 0, there exists x A such that x > M).
We say that a real number L is a limit of f(x) as
x approaches + if for every > 0, there exists
M > 0 such that [f(x)L[ < for every x Aand
x > M. In this case, we write lim
x+
f(x) = L,
or f(x) L as x +.
(2) Suppose A is not bounded below (i.e., for every
M < 0, there exists x A such that x < M).
We say that a real number L is a limit of f(x) as
x approaches if for every > 0, there exists
M < 0 such that [f(x)L[ < for every x Aand
x < M. In this case, we write lim
x
f(x) = L,
or f(x) L as x .
Remarks.
(1) The concept of the limit of a sequence is a special
case of the above denition (with A = N).
54 MA3110 MATHEMATICAL ANALYSIS II
(2) Previous theorems on limits in Section 3.1 (with ob-
vious modications) remain valid if x c is re-
placed by x + (respectively, by
x ).
3.5 Innite limits at innity
Denition. Let A be a nonempty subset of R,
f : A R and A is not bounded above.
(1) We say that f(x) tends to + as x +, writ-
ten lim
x+
f(x) = +, if K > 0, M > 0
such that f(x) > K for every x A and x > M.
(2) We say that f(x) tends to as x +, writ-
ten lim
x+
f(x) = , if K < 0, M > 0
such that f(x) < K for every x A and x > M.
Remark. Similar denitions can be formulated for
lim
x
f(x) = + and lim
x
f(x) = for A
not bounded below.
Theorem 3.9. Let f : (0, ) R and let g be
dened by g(x) = f(
1
x
) for x > 0. Then
lim
x+
f(x) = lim
x0+
g(x).
MA3110 MATHEMATICAL ANALYSIS II 55
4. Continuous Functions
4.1 Continuous functions and general prop-
erties
Denition. Let A be a nonempty subset of R and let
a A. A function f : A R is said to be continuous
at a if for every > 0, there exists > 0 ( depending
on and a) such that [f(x) f(a)[ < whenever
x A and [x a[ < . If f is not continuous at a, we
say that f is discontinuous at a (and a is a point of
discontinuity of f).
If f is continuous at every point in A, we say that f is
continuous on A.
56 MA3110 MATHEMATICAL ANALYSIS II
Remarks.
(1) If a A is not a cluster point of A, then f is always
continuous at a (since there exists > 0 such that
A (a , a + ) = a).
(2) If a A is a cluster point of A (in particular if A is
an interval and a Ais not an end point of A), then
f is continuous at a if and only if f(a) = lim
xa
f(x).
MA3110 MATHEMATICAL ANALYSIS II 57
Example 4.1. Let A = [0, 1] 2 and f : A R
be dened by f(x) = x. Then 2 A but 2 is not a
cluster point of A, and f is continuous at 2.
Example 4.2. Show that the function f : (0, )
R dened by f(x) = 1/x is continuous on (0, ).
Proof. Let a (0, ). Let > 0 be given.
Choose = min
a
2
2
,
a
2
.
(Note: depends on both and a.)
If x (a , a + ) (0, ), then
x > a a
a
2
=
a
2
.
Hence
[f(x) f(a)[ =

1
x

1
a

=
[a x[
[ax[
<

ax
<
2
a
2
.
Therefore f is continuous at a.
58 MA3110 MATHEMATICAL ANALYSIS II
The following sequential criterion for continuity is anal-
ogous to Theorem 3.3 for existence of limits.
Theorem 4.1. Let A R, a A and f : A R.
The following conditions are equivalent.
(i) f is continuous at a.
(ii) For every sequence (x
n
)

n=1
in A that converges
to a, the sequence (f(x
n
))

n=1
converges to f(a).
MA3110 MATHEMATICAL ANALYSIS II 59
Example 4.3. (Dirichlet) Let f : R R be dened
by
f(x) =
_
1 if x is rational
0 if x is irrational.
Then f is not continuous at any real number.
Proof. Let c R. Take a sequence (r
n
)

n=1
of ratio-
nal numbers converging to c, and a sequence (q
n
)

n=1
of irrational numbers converging to c. (Such sequences
exist by density of rational numbers, respectively irra-
tional numbers, in R.)
Then the sequence (f(r
n
))

n=1
converges to 1, whereas
the sequence (f(q
n
))

n=1
converges to 0.
Hence f is not continuous at c.
60 MA3110 MATHEMATICAL ANALYSIS II
Example 4.4. (Thomae) Let g : [0, 1] R be de-
ned by g(x) =
_

_
1 if x = 0
1
n
if x =
m
n
, m, n N, (m, n) = 1
0 if x is irrational.
Then g is continuous at every irrational number in [0, 1]
and discontinuous at every rational number in [0, 1].
Proof. If r is a rational number in [0, 1], then g(r) > 0.
Choose a sequence (q
n
)

n=1
of irrational numbers in [0, 1]
that converges to r. Then (g(q
n
))

n=1
is the constant 0
sequence which does not converge to g(r) > 0. Hence
g is not continuous at r.
Let q [0, 1] be irrational. Given > 0, want to nd
> 0 such that if [
m
n
q[ < , m, n N, (m, n) = 1,
then
1
n
= [
1
n
0[ = [g(
m
n
) g(q)[ < . (Why?)
There are only nitely many n N with n
1

. Hence
B =
m
n
[0, 1] : m, n N, (m, n) = 1, n
1

is
nite. Pick 0 < < min [
m
n
q[ :
m
n
B.
If [
m
n
q[ < , m, n N, (m, n) = 1, then
m
n
/ B, n >
1

and so [g(
m
n
) g(q)[ =
1
n
< .
Hence g is continuous at q.
MA3110 MATHEMATICAL ANALYSIS II 61
Proposition 4.2. Let A R and let f, g : A R
be functions that are continuous at a A. Then the
functions f +g, f g, kf and fg are all continuous
at a, for every k R. If g(x) ,= 0 for all x A,
then f/g is also continuous at a.
Example 4.5. Any polynomial is continuous on R.
62 MA3110 MATHEMATICAL ANALYSIS II
Proposition 4.3. Let A, B be subsets of R and let
g : A R and f : B R be functions such that
g(x) B for all x A. If g is continuous at a
A and f is continuous at g(a), then the composite
function f g is continuous at a.
Proof. Suppose (x
n
)

n=1
is a sequence in A that con-
verges to a.
By Theorem 4.1, the sequence (g(x
n
))

n=1
converges to
g(a).
By the assumption, (g(x
n
))

n=1
is a sequence in B and
g(a) B.
By Theorem 4.1, the sequence (f(g(x
n
)))

n=1
converges
to f(g(a)).
Thus the sequence ((f g)(x
n
))

n=1
converges to (f
g)(a).
Hence f g is continuous at a by Theorem 4.1 again.

MA3110 MATHEMATICAL ANALYSIS II 63


Example 4.6. The function h : R R dened by
h(x) = 1/(x
2
+ 1) is continuous on R.
Proof. The function g : R R, g(x) = x
2
+ 1 is con-
tinuous on R since it is a polynomial function.
The function f : (0, ) R, f(x) = 1/x is continu-
ous on (0, ) (Example 4.2).
Note that g(x) (0, ) for all x R.
By Proposition 4.3, h = f g is continuous on R.
64 MA3110 MATHEMATICAL ANALYSIS II
Remarks. Let a A be a cluster point of A and
f : A R.
(1) If f is discontinuous at a, then there are two possi-
bilities:
(i) lim
xa
f(x) does not exist in R.
In this case, we say that f has an essential dis-
continuity at a. In particular, if lim
xa
f(x)
and lim
xa+
f(x) both exist but
lim
xa
f(x) ,= lim
xa+
f(x),
we say that f has a jump discontinuity at a.
(ii) lim
xa
f(x) exists in R but lim
xa
f(x) ,= f(a).
In this case, we say that f has a removable dis-
continuity at a.
f can be redened so that the new function F is
continuous at a, by dening
F(x) =
_
f(x) if x ,= a
lim
ta
f(t) if x = a.
(Then F is dened at a and
lim
xa
F(x) = lim
xa
f(x) = F(a).)
MA3110 MATHEMATICAL ANALYSIS II 65
(2) Suppose lim
xa
f(x) and lim
xa+
f(x) both exist
in R. Dene the jump of f at a by
j
f
(a) = lim
xa+
f(x) lim
xa
f(x).
Then f has a removable discontinuity (or no discon-
tinuity) at a if j
f
(a) = 0,
and f has a jump discontinuity at a if j
f
(a) ,= 0.
66 MA3110 MATHEMATICAL ANALYSIS II
4.2 Continuous functions on intervals
Proposition 4.4. Let a, b R with a < b. Suppose
that f : [a, b] R is continuous on [a, b]. Then f
is a bounded function, i.e., the range of f (the set
f([a, b]) := f(x) : x [a, b]) is bounded.
Proof. Suppose on the contrary that f is not bounded.
Then [f(x)[ : x [a, b] is not bounded above. (cf.
Tutorial 1)
For every n N, x
n
[a, b] such that [f(x
n
)[ > n.
The sequence (x
n
)

n=1
is a bounded sequence.
By the Bolzano-Weierstrass Theorem, (x
n
)

n=1
has a
convergent subsequence (x
n
k
)

k=1
converging to, say c.
Since a x
n
k
b for all k, a c b by Proposi-
tion 2.3.
By Theorem 4.1, (f(x
n
k
))

k=1
converges to f(c).
By Proposition 2.2, (f(x
n
k
))

k=1
is a bounded sequence.
Hence M such that [f(x
n
k
)[ M for all k N.
Choose k such that k M. Then n
k
k M and
so [f(x
n
k
)[ > n
k
M.
MA3110 MATHEMATICAL ANALYSIS II 67
Denition. Let A R and let f : A R.
(1) f has an absolute maximum on A if there exists
x

A such that f(x

) f(x) for every x A. In


this case, x

is called an absolute maximum point


for f on A.
(2) f has an absolute minimum on A if there exists
x

A such that f(x

) f(x) for every x A. In


this case, x

is called an absolute minimum point


for f on A.
Theorem 4.5. (Extreme Value Theorem) Let a, b
R with a < b. Suppose that f : [a, b] R is con-
tinuous on [a, b]. Then f has an absolute maximum
and an absolute minimum on [a, b], i.e., there exist
c
1
, c
2
[a, b] such that f(c
1
) f(x) f(c
2
) for all
x [a, b].
68 MA3110 MATHEMATICAL ANALYSIS II
Proof. We will show how to obtain c
2
.
By Proposition 4.4, S := f(x) : x [a, b] is bounded.
It is clearly nonempty and so L := sup S exists in R.
For each n N, L 1/n is not an upper bound of S.
So x
n
[a, b] such that L 1/n < f(x
n
) L.
The sequence (x
n
)

n=1
is a bounded sequence.
By the Bolzano-Weierstrass Theorem, it has a conver-
gent subsequence (x
n
k
)

k=1
converging to, say c
2
.
Since a x
n
k
b for all k,
a c
2
b by Proposition 2.3.
By Theorem 4.1, (f(x
n
k
))

k=1
converges to f(c
2
).
Since L1/n < f(x
n
) L for all n, by Squeeze The-
orem (f(x
n
))

n=1
converges to L.
By Proposition 2.7, the subsequence (f(x
n
k
))

k=1
also
converges to L.
By uniqueness of limits, f(c
2
) = L = sup S.
Hence f(x) f(c
2
) for all x [a, b].
MA3110 MATHEMATICAL ANALYSIS II 69
Theorem 4.6. (Intermediate Value Theorem) Let a, b
R with a < b. Suppose that f : [a, b] R is
continuous on [a, b]. For any number L between
f(a) and f(b) (i.e., either f(a) < L < f(b) or
f(b) < L < f(a)), there exists c (a, b) such that
f(c) = L.
Proof. Without loss of generality, we may assume
f(a) < f(b).
(i) First consider the case f(a) < 0 < f(b).
Let S = x [a, b] : f(x) < 0. Since a S and S is
bounded above (by b), c := sup S exists in R.
We will show: (1) c (a, b) and (2) f(c) = 0.
(1) Since f is continuous on [a, b], f(a) < 0, f(b) > 0,
by Proposition 3.7, there exists a small > 0 such that
f(x) < 0 on [a, a+) and f(x) > 0 on (b , b]. (Ex.)
If x S, then x b .
Hence a < a +

2
c b < b.
(2) If f(c) < 0, then there exists a small > 0 such
that f(x) < 0 on (c , c + ). So c +

2
S.
But this contradicts that c is an upper bound of S.
70 MA3110 MATHEMATICAL ANALYSIS II
If f(c) > 0, then there exists a small > 0 such that
f(x) > 0 on (c, c+). Since c = sup S, there exists
x S such that c < x c. By the denition of
S, f(x) < 0.
Hence f(c) = 0.
(ii) Let f(a) < L < f(b).
Dene g on [a, b] by g(x) = f(x) L.
Then g is continuous on [a, b] and g(a) < 0 < g(b).
By Case (i), there exists c (a, b) such that g(c) = 0.
Therefore f(c) = L.
MA3110 MATHEMATICAL ANALYSIS II 71
Example 4.7. Any polynomial of odd degree must
have at least one real root.
Proof. First consider a polynomial P of odd degree
with leading coecient 1:
P(x) = x
2n1
+ a
2n2
x
2n2
+ + a
1
x + a
0
.
Let M := [a
2n2
[ + +[a
1
[ +[a
0
[ +1. Since M 1,
[a
2n2
M
2n2
[ + + [a
1
M[ + [a
0
[
([a
2n2
[ + + [a
1
[ + [a
0
[)M
2n2
< M
2n1
.
Thus
P(M) = M
2n1
+ a
2n2
M
2n2
+ + a
1
M + a
0
M
2n1
([a
2n2
M
2n2
[ + + [a
1
M[ + [a
0
[)
> M
2n1
M
2n1
= 0.
Similarly,
P(M) = (M)
2n1
+ a
2n2
(M)
2n2
+ + a
1
(M) + a
0
M
2n1
+ [a
2n2
M
2n2
[ + + [a
1
M[ + [a
0
[
< M
2n1
+ M
2n1
= 0.
72 MA3110 MATHEMATICAL ANALYSIS II
Now P is continuous on [M, M], and
P(M) < 0 < P(M).
By the Intermediate Value Theorem, there exists
c (M, M) such that P(c) = 0.
If Q is a polynomial of odd degree, let its leading coef-
cient be a ,= 0. Then P =
1
a
Q is a polynomial of odd
degree with leading coecient 1.
By the above, there exists c R such that P(c) = 0.
Then Q(c) = 0 as well. So c is a real root of Q.
MA3110 MATHEMATICAL ANALYSIS II 73
Theorem 4.7. Let a, b R with a < b. Sup-
pose that f : [a, b] R is continuous on [a, b].
Then the range f([a, b]) is the closed and bounded
interval [m, M], where m = min f([a, b]) and M =
max f([a, b]).
Proof. By Extreme Value Theorem, there exist c
1
, c
2
in
[a, b] such that f(c
1
) = m and f(c
2
) = M.
If m < L < M, then by Intermediate Value Theorem,
there exists c between c
1
and c
2
such that
L = f(c) f([a, b]). Hence [m, M] f([a, b]).
Conversely, by the denition of m and M,
we have f([a, b]) [m, M].
Hence f([a, b]) = [m, M]
Remark. Supposing f(a) < f(b), the range f([a, b])
is not necessarily the interval [f(a), f(b)] .
74 MA3110 MATHEMATICAL ANALYSIS II
The next result extends Theorem 4.7 to other types of
intervals.
Theorem 4.8. (Preservation of intervals) Let I be an
interval in R and let f : I R be continuous on I.
Then f(I) is an interval.
Proof. We may assume that f is not a constant func-
tion. Let x, y f(I) and x < y.
Then x = f(a) and y = f(b) for some a, b I.
By Intermediate Value Theorem, [f(a), f(b)] f(I),
i.e., [x, y] f(I).
By Proposition 1.10, f(I) is an interval.
Remark. The image interval f(I) may or may not
have the same form as the domain interval I.
Example 4.8. The function f(x) = 1/x is continuous
on (0, ) (Example 4.2).
(1) The interval (0, 1) is bounded
but f((0, 1)) = (1, ) is unbounded.
(2) The interval [1, ) is unbounded
but f([1, )) = (0, 1] is bounded.
MA3110 MATHEMATICAL ANALYSIS II 75
4.3 Uniform continuity
Denition. Let A be a nonempty subset of R. We say
that a function f : A R is uniformly continuous on
A if for every > 0, there exists > 0 ( depending on
only) such that [f(x) f(y)[ < whenever x, y A
and [x y[ < .
Remarks.
(1) If f is uniformly continuous on A, then f is con-
tinuous on A (i.e., continuous at every point in A).
The converse is not true in general.
(2) If f is uniformly continuous on A, then f is uni-
formly continuous on B for every nonempty subset
B of A.
76 MA3110 MATHEMATICAL ANALYSIS II
Example 4.9. Let f(x) = 1/x be dened on (0, ).
(1) f is continuous on (0, ). (Example 4.2)
(2) f is not uniformly continuous on (0, ).
(3) For any c > 0, f is uniformly continuous on [c, ).
Proof. (2): Take = 1. Given > 0, choose n N
such that n > 1/

. Let x
n
=
1
n
and y
n
=
1
n+1
.
Then x
n
, y
n
(0, ),
[x
n
y
n
[ = [
1
n

1
n+1
[ =
1
n(n+1)
<
1
n
2
< ,
but [f(x
n
) f(y
n
)[ = [n (n + 1)[ = 1 = .
Hence f is not uniformly continuous on (0, ).
(3): Let c > 0. Given > 0, let = c
2
.
If x, y [c, ) and [xy[ < , then x c, y c, and
[f(x) f(y)[ =

1
x

1
y

=
[x y[
[xy[
<

c
2
= .
Hence f is uniformly continuous on [c, ).
MA3110 MATHEMATICAL ANALYSIS II 77
Example 4.10. Let c > 0. The function f(x) = x
2
dened on A = [0, c] is uniformly continuous on A.
Proof. Let > 0. Dene =

2c
.
Then for every x, y A and [x y[ < ,
0 x c, 0 y c, and
[f(x) f(y)[ = [x
2
y
2
[ = [x +y[[x y[ < 2c = .
Hence f is uniformly continuous on A.
78 MA3110 MATHEMATICAL ANALYSIS II
Theorem 4.9. Let f : [a, b] R be continuous
on [a, b]. Then f is uniformly continuous on [a, b]:
for every > 0, there exists > 0 ( depending
on only) such that [f(x) f(y)[ < whenever
x, y [a, b] and [x y[ < .
Proof. Suppose otherwise, then there exists > 0 such
that for any n N, there exist x
n
and y
n
in [a, b],
[x
n
y
n
[ <
1
n
and [f(x
n
) f(y
n
)[ .
By the Bolzano-Weierstrass Theorem, (x
n
)

n=1
has a
convergent subsequence (x
n
k
)

k=1
converging to c say.
Since (
1
n
k
)

k=1
is a subsequence of (
1
n
)

n=1
, lim
k
1
n
k
= 0.
Since x
n
k

1
n
k
< y
n
k
< x
n
k
+
1
n
k
,
by the Squeeze Theorem, lim
k
y
n
k
= c too.
Since a x
n
k
b for all k, a c b by Proposi-
tion 2.3.
Since f is continuous at c, (f(x
n
k
))

k=1
and (f(y
n
k
))

k=1
both converge to f(c) by Theorem 4.1.
MA3110 MATHEMATICAL ANALYSIS II 79
In particular, there are N
1
and N
2
such that
[f(x
n
k
) f(c)[ <

2
whenever k N
1
and
[f(y
n
k
) f(c)[ <

2
whenever k N
2
.
Letting N = maxN
1
, N
2
, then
[f(x
n
N
)f(y
n
N
)[ [f(x
n
N
)f(c)[+[f(c)f(y
n
N
)[ < .

80 MA3110 MATHEMATICAL ANALYSIS II
4.4 Monotone and inverse functions on inter-
vals
Denition. Let A R and let f : A R.
(1) f is increasing on A if f(x
1
) f(x
2
) whenever
x
1
, x
2
A and x
1
< x
2
. f is decreasing on A if
f(x
1
) f(x
2
) whenever x
1
, x
2
A and x
1
< x
2
.
A function is monotone on Aif it is either increasing
or decreasing on A.
(3) f is strictly increasing on Aif f(x
1
) < f(x
2
) when-
ever x
1
, x
2
A and x
1
< x
2
.
f is strictly decreasing on Aif f(x
1
) > f(x
2
) when-
ever x
1
, x
2
A and x
1
< x
2
.
A function is strictly monotone on A if it is either
strictly increasing or strictly decreasing on A.
Notation. For the rest of this section, let I be an
interval in R containing more than one point. The
interval may be open or closed at either end, and it
may also be unbounded.
MA3110 MATHEMATICAL ANALYSIS II 81
The next result should be compared with the Monotone
Convergence Theorem for sequences (Proposition 2.5).
Proposition 4.10. Let f : I R be increasing on
I. If c I is not an end point of I, then
(i) lim
xc
f(x) = sup f(x) : x I, x < c.
(ii) lim
xc+
f(x) = inf f(x) : x I, x > c.
Proof. (i): Let S = f(x) : x I, x < c. Since c
is not an end point of I, there exists > 0 such that
(c , c + ) I. Hence S is nonempty. Also, since
f is increasing on I, f(c) is an upper bound of S.
Let L := sup S and let > 0 be given.
Then there exists u I and u < c such that
L < f(u) L.
Let = c u.
If x I and 0 < c x < , then u < x < c and so
L < f(u) f(x) L.
Hence [f(x)L[ < for every x I and 0 < cx < .
Therefore lim
xc
f(x) = L.
(ii): The proof is similar. (Ex.)
82 MA3110 MATHEMATICAL ANALYSIS II
Denition. Let f : I R be increasing on I.
(1) If c I is not an end point of I, dene the jump of
f at c by
j
f
(c) = lim
xc+
f(x) lim
xc
f(x).
(2) If the left-hand end point a of I is in I, dene the
jump of f at a by
j
f
(a) = lim
xa+
f(x) f(a).
(3) If the right-hand end point b of I is in I, dene the
jump of f at b by
j
f
(b) = f(b) lim
xb
f(x).
Proposition 4.11. Let f : I R be increasing on
I and c I. Then f is continuous at c if and only
if j
f
(c) = 0.
Proof. Suppose c is an interior point of I. By Proposi-
tion 4.10, lim
xc
f(x) f(c) lim
xc+
f(x).
Then the desired result follows from Theorem 3.8.
If c is the left-hand or the right-hand end point of I,
the proof is similar.
MA3110 MATHEMATICAL ANALYSIS II 83
Theorem 4.12. Let f : I R be monotone on
I. Then the set of points of discontinuity of f is a
countable set.
Proof. Assume f is increasing on I.
By Proposition 4.10, j
f
(x) 0 for all x I.
By Proposition 4.11, the set D of points of discontinu-
ity of f is given by D = x I : j
f
(x) > 0.
For simplicity of argument, consider the case I is a
closed and bounded interval [a, b] and f(a) < f(b).
Let a < x
1
< . . . < x
n
< b.
For each i = 1, . . . , n 1, as x
i
< x
i+1
, by Proposi-
tion 4.10, lim
xx
i
+
f(x) lim
xx
i+1

f(x).
84 MA3110 MATHEMATICAL ANALYSIS II
Since each j
f
(x
i
) := lim
xx
i
+
lim
xx
i

0, we have
f(a) f(a) + j
f
(x
1
) + j
f
(x
2
) + . . . + j
f
(x
n
)
= f(a) + ( lim
xx
1
+
lim
xx
1

) + ( lim
xx
2
+
lim
xx
2

)
+. . . + ( lim
xx
n
+
lim
xx
n

)
f(a) lim
xx
1

+ lim
xx
n
+
f(a) f(a) + f(b)
= f(b).
Hence j
f
(x
1
) + j
f
(x
2
) + . . . + j
f
(x
n
) f(b) f(a).
Therefore for each k N, there can be at most k dis-
tinct points x in (a, b) with j
f
(x)
f(b)f(a)
k
.
Since Da, b

k=1
x (a, b) : j
f
(x)
f(b)f(a)
k
,
which is the union of a countable collection of count-
able (indeed nite) sets, by Proposition 1.8 the set D
is countable.
MA3110 MATHEMATICAL ANALYSIS II 85
Theorem 4.13. (Continuous Inverse Theorem) Let
f : I R be strictly monotone and continuous
on I. Then the inverse function g of f is strictly
monotone and continuous on the interval J := f(I).
Proof. Assume f is strictly increasing and continuous
on I. By Theorem 4.8, J = f(I) is an interval.
Since f is strictly increasing on I, it is also injective on
I. Hence its inverse g : J I exists.
(i) We will rst show that g is also strictly increasing
on J. Let y
1
, y
2
J and y
1
< y
2
.
Then y
1
= f(x
1
) and y
2
= f(x
2
) for some x
1
, x
2
I.
If x
1
x
2
, since f is (strictly) increasing, then
y
1
= f(x
1
) f(x
2
) = y
2
.
Hence g(y
1
) = x
1
< x
2
= g(y
2
).
(ii) Next we will show that g is continuous on J. Sup-
pose on the contrary that g is discontinuous at some
c J. Consider only the case when c is not an end
point of I (the proof is similar for the case when c is an
end point). Then by Proposition 4.11, j
f
(c) > 0, i.e.,
lim
yc
g(y) < lim
yc+
g(y).
86 MA3110 MATHEMATICAL ANALYSIS II
Note that lim
yc
g(y) g(c) lim
yc+
g(y).
Pick a number x ,= g(c) and
lim
yc
g(y) < x < lim
yc+
g(y).
By Proposition 4.10,
sup g(y) : y J, y < c = lim
yc
g(y) < x,
x < lim
yc+
g(y) = inf g(y) : y J, y > c.
Hence g(y) < x for every y J and y < c, and
x < g(z) for every z J and z > c.
Thus x / g(J) = I.
But there exist y, z J, y < c, z > c such that
g(y) < x < g(z) and g(y), g(z) are in I.
Since I is an interval, x I.
MA3110 MATHEMATICAL ANALYSIS II 87
Example 4.11. (n-th root function)
(1) n even positve integer:
In this case, consider f(x) = x
n
, x I = [0, ).
f is strictly increasing (Ex.) and continuous on I
(Example 4.5).
Claim: J := f(I) = [0, ).
Obviously, J [0, ). Conversely, let y > 0. By
Archimedean property, there exists k N such that
y < k. Since f(0) = 0 < y < k k
n
= f(k), by
Intermediate Value Theorem, y f(I) = J.
By Theorem 4.13, the function g inverse to f is
strictly increasing and continuous on J, and is called
the n-th root function. We write
g(x) = x
1
n
, x 0,
and call x
1
n
the n-th root of x.
As inverses, g(f(x)) = x = f(g(x)), x 0,
i.e., (x
n
)
1
n
= x = (x
1
n
)
n
, x 0.
88 MA3110 MATHEMATICAL ANALYSIS II
(2) n odd positve integer:
In this case, consider F(x) = x
n
, x R.
F is strictly increasing and continuous on R, and
F(R) = R. (Ex.)
By Theorem 4.13, the function G inverse to F is
strictly increasing and continuous on R, and is called
the n-th root function. As before, we write
G(x) = x
1
n
, x R,
and call x
1
n
the n-th root of x. We also have
(x
n
)
1
n
= x = (x
1
n
)
n
, x R.
MA3110 MATHEMATICAL ANALYSIS II 89
5. Differentiation
5.1 The derivative and general properties
Denition. Let I be an interval in R and let c I.
A function f : I R is said to be dierentiable at c
if the limit
lim
xc
f(x) f(c)
x c
= lim
t0
f(c + t) f(c)
t
= L
exists in R, i.e.,
for every > 0, there exists > 0 such that
[
f(x)f(c)
xc
L[ < for every x I and 0 < [xc[ < .
In this case, we denote the limit L by f
/
(c) (or
df
dx
(c),
or
df
dx
[
x=c
) and call this value the derivative of f at c.
If f is dierentiable at every point in a set S, then
we say that f is dierentiable on S, and the derivative
function f
/
(or
df
dx
) is dened on S.
90 MA3110 MATHEMATICAL ANALYSIS II
Example 5.1. For n N, let f : R R be dened
by f(x) = x
n
, x R.
Then f is dierentiable on R
and f
/
(c) = nc
n1
for all c R.
(1) n = 1: f(x) = x; f
/
(c) = 1 for all c R. (Ex.)
(2) n = 2:
f(x) f(c)
x c
=
x
2
c
2
x c
= x + c if x ,= c.
Hence
f
/
(c) = lim
xc
f(x) f(c)
x c
= lim
xc
(x + c) = 2c.
(3) n 2:
f
/
(c) = lim
t0
f(c + t) f(c)
t
= lim
t0
(c + t)
n
c
n
t
= lim
t0
(c
n
+ nc
n1
t +
1
2
n(n 1)c
n2
t
2
+ . . . + nct
n1
+ t
n
) c
n
t
= lim
t0
(nc
n1
+
1
2
n(n 1)c
n2
t + . . . + nct
n2
+ t
n1
)
= nc
n1
.
MA3110 MATHEMATICAL ANALYSIS II 91
Example 5.2. Let
f(x) =
_
_
_
x
2
if x 0
x
3
if x > 0.
Using the denition, f
/
(0) = lim
x0
f(x) f(0)
x 0
= lim
x0
f(x)
x
.
Now lim
x0
f(x)
x
= lim
x0
x
2
x
= lim
x0
(x) = 0,
and lim
x0+
f(x)
x
= lim
x0+
x
3
x
= lim
x0+
(x
2
) = 0.
Therefore f
/
(0) = lim
x0
f(x)
x
= 0.
By Example 5.1,
f
/
(x) =
_
_
_
2x if x < 0
3x
2
if x > 0.
92 MA3110 MATHEMATICAL ANALYSIS II
Proposition 5.1. If f : I R is dierentiable at
c I, then it is continuous at c.
Proof. For every x Ic,
f(x) =
f(x) f(c)
x c
(x c) + f(c).
Since f
/
(c) exists, by Proposition 3.5,
lim
xc
f(x) = lim
xc
_
f(x) f(c)
x c
_
lim
xc
(x c) + f(c)
= f
/
(c) 0 + f(c)
= f(c).
Hence f is continuous at c.
Remark. The converse of Proposition 5.1 is false.
Example 5.3. Let f(x) = [x[, x R.
Then f is continuous on R. (Ex.)
f is dierentiable at every x ,= 0,
f
/
(x) =
_
_
_
1 if x < 0
1 if x > 0,
but f is not dierentiable at 0. (Ex.)
MA3110 MATHEMATICAL ANALYSIS II 93
Proposition 5.2. Suppose that f, g : I R are
dierentiable at c I. Then
(i) f + g is dierentiable at c and
(f + g)
/
(c) = f
/
(c) + g
/
(c).
(ii) f g is dierentiable at c and
(f g)
/
(c) = f
/
(c) g
/
(c).
(iii) kf is dierentiable at c and (kf)
/
(c) = k f
/
(c)
for every k R.
(iv) (Product Rule) fg is dierentiable at c and
(fg)
/
(c) = f
/
(c)g(c) + f(c)g
/
(c).
(v) (Quotient Rule) If g(x) ,= 0 for all x I, then
f/g is dierentiable at c and
_
f
g
_
/
(c) =
f
/
(c)g(c) f(c)g
/
(c)
(g(c))
2
.
94 MA3110 MATHEMATICAL ANALYSIS II
Proof. (iv): For every x Ic,
f(x)g(x) f(c)g(c)
x c
=
(f(x) f(c))g(x) + f(c)(g(x) g(c))
x c
=
f(x) f(c)
x c
g(x) + f(c)
g(x) g(c)
x c
.
By Proposition 5.1, lim
xc
g(x) = g(c). Taking limits
as x c and using Proposition 3.5,
lim
xc
f(x)g(x) f(c)g(c)
x c
= f
/
(c)g(c) + f(c)g
/
(c).

MA3110 MATHEMATICAL ANALYSIS II 95


Example 5.4. For n N, dene h : R0 R by
h(x) = 1/x
n
= x
n
. Then h is dierentiable at any
c ,= 0 and h
/
(c) = n/c
n+1
= nc
n1
.
Proof. Let g : R0 R, g(x) = x
n
.
By Example 5.1, g
/
(c) = nc
n1
for any c ,= 0.
Also dene f : R0 R, f(x) = 1. Then f
/
(c) = 0.
Note that g(x) ,= 0 if x ,= 0.
By Quotient Rule, h = f/g is dierentiable at any
c ,= 0 and
h
/
(c) = (f/g)
/
(c) =
f
/
(c)g(c) f(c)g
/
(c)
(g(c))
2
=
nc
n1
c
2n
=
n
c
n+1
.

Remark. We will also assume the following results.


(1) If f : R R, f(x) = sin x, then f is dierentiable
on R and f
/
(x) = cos x.
(2) If f : R R, f(x) = cos x, then f is dierentiable
on R and f
/
(x) = sin x.
The derivatives of the other trigonometric functions
can be obtained by using the above facts and Proposi-
tion 5.2.
96 MA3110 MATHEMATICAL ANALYSIS II
Lemma 5.3. (Caratheodory) Let f : I R and let
c I. The following conditions are equivalent:
(i) f is dierentiable at c.
(ii) There exists a function dened on I such
that is continuous at c and
f(x) f(c) = (x)(x c) for every x I.
In this case, (c) = f
/
(c).
Proof. (i) (ii): Suppose f
/
(c) exists. Dene a func-
tion on I by
(x) =
_
f(x)f(c)
xc
if x Ic
f
/
(c) if x = c.
Since lim
xc
(x) = lim
xc
f(x) f(c)
x c
= f
/
(c) = (c),
is continuous at c. It is obvious that
f(x) f(c) = (x)(x c) for all x I.
(ii) (i): Suppose (ii) holds. By continuity of at c,
lim
xc
f(x) f(c)
x c
= lim
xc
(x) = (c).
Hence f is dierentiable at c and f
/
(c) = (c).
MA3110 MATHEMATICAL ANALYSIS II 97
Theorem 5.4. (Chain Rule) Let I, J be intervals in
R and let g : I R and f : J R be functions
such that g(x) J for all x I. If g is dieren-
tiable at c I and f is dierentiable at g(c) J,
then f g is dierentiable at c and
(f g)
/
(c) = f
/
(g(c)) g
/
(c).
Remarks.
(1) Another form of the Chain Rule:
Let y = f(g(x)) = f(u), where u = g(x). Then
dy
dx
= (f g)
/
(x) = f
/
(g(x))g
/
(x) = f
/
(u)u
/
=
dy
du
du
dx
.
Formally :
dy
dx
=
dy
du
du
dx
.
(2) Rewrite a complicated function as a composite func-
tion (making a substitution) before using the Chain
Rule to nd its derivative.
98 MA3110 MATHEMATICAL ANALYSIS II
Proof. Let > 0 be given. Set

1
= min
_
1,

[g
/
(c)[ + 1 + [f
/
(g(c))[
_
.
Since g is dierentiable at c, there exists
1
> 0 such
that if x I and 0 < [x c[ <
1
, then
(5.1)

g(x) g(c)
x c
g
/
(c)

<
1
.
Since f is dierentiable at g(c), there exists
2
> 0 such
that if u J and 0 < [u g(c)[ <
2
, then

f(u) f(g(c))
u g(c)
f
/
(g(c))

<
1
.
Hence if u J and [u g(c)[ <
2
, then
(5.2)
[f(u) f(g(c)) f
/
(g(c))(u g(c))[
1
[u g(c)[.
Since g is dierentiable at c, it is continuous at c by
Proposition 5.1. Thus there exists
3
> 0 such that if
x I and [x c[ <
3
, then
(5.3) [g(x) g(c)[ <
2
.
Set = min
1
,
3
.
If x I and 0 < [x c[ < , then
(A) x I and 0 < [x c[ <
1
, hence (5.1) holds.
MA3110 MATHEMATICAL ANALYSIS II 99
(B) x I and [x c[ <
3
, hence (5.3) holds. Thus
g(x) J and [g(x) g(c)[ <
2
, and hence (5.2)
holds with u = g(x).
Therefore, for x I with 0 < [x c[ < ,

f(g(x)) f(g(c))
x c
f
/
(g(c))g
/
(c)

=
[f(g(x)) f(g(c)) (x c)f
/
(g(c))g
/
(c)[
[x c[

[f(g(x)) f(g(c)) f
/
(g(c))(g(x) g(c))[
[x c[
+
[f
/
(g(c))(g(x) g(c)) (x c)f
/
(g(c))g
/
(c)[
[x c[

1

g(x) g(c)
x c

+ [f
/
(g(c))[

g(x) g(c)
x c
g
/
(c)

<
1
([g
/
(c)[ +
1
) + [f
/
(g(c))[
1

1
([g
/
(c)[ + 1) + [f
/
(g(c))[
1
=
1
([g
/
(c)[ + 1 + [f
/
(g(c))[)


[g
/
(c)[ + 1 + [f
/
(g(c))[
([g
/
(c)[ + 1 + [f
/
(g(c))[)
= .
We are done!
100 MA3110 MATHEMATICAL ANALYSIS II
Example 5.5. Dene f : R R by
f(x) =
_
_
_
x
2
sin(1/x) if x ,= 0
0 if x = 0.
Then f is dierentiable at every point in R.
Proof. Dene h : R0 R, h(x) = 1/x
and g : R R, g(x) = sin x.
If c ,= 0, then h is dierentiable at c (Example 5.4) and
g is dierentiable at h(c) (previous remark).
Hence s = gh is dierentiable at c by the Chain Rule.
So s(x) = sin(1/x) is dierentiable at c ,= 0, and
s
/
(c) = (g h)
/
(c) = g
/
(h(c))h
/
(c)
= cos(h(c))(1/c
2
) =
cos(1/c)
c
2
.
Hence f(x) = x
2
s(x) is dierentiable at c (,= 0) by the
Product Rule, and
f
/
(c) = 2cs(c) + c
2
s
/
(c) = 2c sin(1/c) cos(1/c).
For c = 0, f
/
(0) = lim
x0
f(x) f(0)
x 0
= lim
x0
x
2
sin(1/x) 0
x 0
= lim
x0
x sin(1/x) = 0, by Squeeze Theorem.
MA3110 MATHEMATICAL ANALYSIS II 101
Theorem 5.5. Let I be an interval in R, f : I R
strictly monotone and continuous on I, J := f(I)
and g : J I the strictly monotone and continuous
function inverse to f. If f is dierentiable at c I
and f
/
(c) ,= 0, then g is dierentiable at d := f(c)
J and g
/
(d) =
1
f
/
(c)
, i.e.,
g
/
(d) =
1
f
/
(g(d))
, or g
/
(f(c)) =
1
f
/
(c)
.
102 MA3110 MATHEMATICAL ANALYSIS II
Proof. Suppose f
/
(c) exists and is nonzero. By Lemma 5.3,
there exists a function dened on I such that is con-
tinuous at c, (c) = f
/
(c) ,= 0 and
f(x) f(c) = (x)(x c) for every x I.
By continuity of at c with (c) ,= 0, there exists > 0
such that (x) ,= 0 for all x V := I (c , c + ).
Note that V is a subinterval of I containing c. By
Theorem 4.8, the set U := f(V ) J is an interval
containing d := f(c) and g(U) = V .
For all y U, we have g(y) V and y = f(g(y)),
so (g(y)) ,= 0 and
y d = f(g(y)) f(c) = (g(y))(g(y) g(d)).
Hence g(y) g(d) =
1
(g(y))
(y d), y U.
The function
1
g
is dened on U and it is continuous
at d by Proposition 4.3 and Proposition 4.2. Then by
Lemma 5.3, g is dierentible at d and
g
/
(d) = 1/(g(d)) = 1/(c) = 1/f
/
(c).
MA3110 MATHEMATICAL ANALYSIS II 103
Remarks.
(1) In Theorem 5.5, the hypothesis that
f
/
(c) ,= 0 is crucial.
Example 5.6. Let f(x) = x
3
and c = 0.
Then f is strictly increasing and dierentiable on
R, and f
/
(0) = 0.
However, the inverse function g(x) = x
1/3
is not
dierentiable at 0. (Ex.)
(2) Indeed, if f
/
(c) = 0, then the inverse function g of
f is not dierentiable at d = f(c). (Ex.)
104 MA3110 MATHEMATICAL ANALYSIS II
5.2 Mean Value Theorem and applications
The circle of related theorems: Rolles Theorem, Mean
Value Theorem, Cauchys Mean Value Theorem and
Taylors Theorem, are fundamental results in dieren-
tial calculus. In this and the next section, we will dis-
cuss these results and a few standard applications.
Denition. Let I be an interval in R, c I and
f : I R.
(1) f has a relative maximum at c if there exists > 0
such that f(c) f(x) for every x I(c, c+).
In this case, c is called a relative maximum point
for f on I.
(2) f has a relative minimum at c if there exists > 0
such that f(c) f(x) for every x I(c, c+).
In this case, c is called a relative minimum point
for f on I.
(3) f has a relative extremum at c (and c is a relative
extremum point for f on I) if f has either a relative
maximum or a relative minimum at c.
MA3110 MATHEMATICAL ANALYSIS II 105
Lemma 5.6. Let I be an interval in R, f : I R
and f
/
(c) exists for some c I.
(i) If f
/
(c) > 0, then there exists > 0 such that
f(x) < f(c) for every x I (c , c), and
f(x) > f(c) for every x I (c, c + ).
(ii) If f
/
(c) < 0, then there exists > 0 such that
f(x) > f(c) for every x I (c , c), and
f(x) < f(c) for every x I (c, c + ).
Proof. (i) Suppose f
/
(c) = lim
xc
f(x) f(c)
x c
> 0.
By Proposition 3.7, there exists > 0 such that
f(x) f(c)
x c
> 0 on I (c , c + ) c.
For x I (c , c),
x c < 0, and so f(x) f(c) < 0.
For x I (c, c + ),
x c > 0, and so f(x) f(c) > 0.
The proof of (ii) is similar.
106 MA3110 MATHEMATICAL ANALYSIS II
Proposition 5.7. (Interior Extremum Theorem) Sup-
pose that c is an interior point of an interval I and
f : I R is dierentiable at c. If f has a relative
extremum at c, then f
/
(c) = 0.
Proof. Suppose f has a relative maximum at c and
f
/
(c) exists.
(1) If f
/
(c) > 0, by Lemma 5.6, there exists a small
positive such that (c , c + ) I (since c is an
interior point of I), and f(x) < f(c) for every x
(c , c), and f(x) > f(c) for every x (c, c + ).
This contradicts that f has a relative maximum at c.
(2) If f
/
(c) < 0, likewise we get a contradiction using
Lemma 5.6.
Hence f
/
(c) = 0.
The case of a relative minimum at c is similar.
MA3110 MATHEMATICAL ANALYSIS II 107
Remarks.
(1) f may have a relative extremum at c but f
/
(c) does
not exist.
Example : Let f(x) = [x[, x R (Example 5.3).
Then f has a relative (indeed absolute) minimum
at 0, but f
/
(0) does not exist.
(2) The converse of Proposition 5.7 is false: f
/
(c) = 0
does not imply that f has a relative extremum at c.
Example : Let f(x) = x
3
, x R (Example 5.6).
Then f
/
(0) = 0 but 0 is not a relative extremum
point of f.
108 MA3110 MATHEMATICAL ANALYSIS II
Theorem 5.8. (Rolles Theorem) Let f : [a, b] R
be continuous on [a, b], dierentiable on (a, b) and
f(a) = f(b) = 0. Then there exists c (a, b) such
that f
/
(c) = 0.
Proof. If f is the constant zero function on [a, b], the
result is obvious. Hence without loss of generality, as-
sume that there exists x [a, b] such that f(x) > 0.
By the Extreme Value Theorem, f attains absolute
maximum on [a, b] at some c [a, b].
In particular, f(c) f(x) > 0.
Therefore c (a, b). Hence c is an interior point of
[a, b] and it is a relative maximum point for f on [a, b].
By Proposition 5.7, f
/
(c) = 0.
Remark. The conditions of continuity on [a, b] and
dierentiabilty on (a, b) in Rolles Theorem cannot be
omitted. (Ex.)
MA3110 MATHEMATICAL ANALYSIS II 109
Theorem 5.9. (Mean Value Theorem) Let f : [a, b]
R be continuous on [a, b] and dierentiable on (a, b).
Then there exists c (a, b) such that
f(b) f(a) = f
/
(c)(b a).
Proof. Dene g : [a, b] R by
g(x) = f(x) f(a)
f(b) f(a)
b a
(x a).
Then g is continuous on [a, b] by Proposition 4.2 and
dierentiable on (a, b) by Proposition 5.2.
Also g(a) = 0 = g(b).
By Rolles Theorem, there exists c (a, b) such that
g
/
(c) = 0.
Since g
/
(c) = f
/
(c)
f(b) f(a)
b a
, we have
f
/
(c) =
f(b) f(a)
b a
= f(b) f(a) = f
/
(c)(b a).

110 MA3110 MATHEMATICAL ANALYSIS II


Theorem 5.10. (Cauchys Mean Value Theorem) Let
f and g be continuous on [a, b] and dierentiable on
(a, b), and assume that g
/
(x) ,= 0 for all x (a, b).
Then there exists c (a, b) such that
f(b) f(a)
g(b) g(a)
=
f
/
(c)
g
/
(c)
.
MA3110 MATHEMATICAL ANALYSIS II 111
Proposition 5.11. Let f be dierentiable on an
interval I. Then f is increasing (respectively de-
creasing) on I if and only if f
/
(x) 0 (respectively
f
/
(x) 0) for all x I.
Proof. if: Suppose that f
/
(x) 0 for all x I.
Let x
1
, x
2
I, with x
1
< x
2
. Applying Mean Value
Theorem to f on [x
1
, x
2
], there exists c (x
1
, x
2
) such
that f(x
2
) f(x
1
) = f
/
(c)(x
2
x
1
) 0.
Therefore, f(x
1
) f(x
2
).
only if: Suppose f is dierentiable and increasing on
I. Hence for any x, c in I with x ,= c, f(x) f(c) and
x c have the same sign and so
f(x) f(c)
x c
0.
Taking limit, f
/
(c) = lim
xc
f(x) f(c)
x c
0.
112 MA3110 MATHEMATICAL ANALYSIS II
Remarks.
(1) Let f be dierentiable on an interval I. If f
/
(x) > 0
(respectively f
/
(x) < 0) for all x I, then f is
strictly increasing (respectively strictly decreasing)
on I. (Same proof as above.)
(2) The converse of (1) is false.
Example : Let f(x) = x
3
and c = 0 (Example 5.6).
Then f is dierentiable and strictly increasing on
R, but f
/
(0) = 0.
Proposition 5.12. Let f, g : [a, b] R be continu-
ous on [a, b] and dierentiable on (a, b). If f
/
(x) =
g
/
(x) for all x (a, b), then there exists a constant
C such that f(x) = g(x) + C for all x [a, b].
MA3110 MATHEMATICAL ANALYSIS II 113
Proposition 5.13. (First Derivative Test) Let f be
continuous on [a, b], c (a, b) and let f
/
exist on
(a, b) except possibly at c.
(i) If f
/
(x) 0 for every x (a, c) and
f
/
(x) 0 for every x (c, b),
then f has a relative maximum at c.
(ii) If f
/
(x) 0 for every x (a, c) and
f
/
(x) 0 for every x (c, b),
then f has a relative minimum at c.
114 MA3110 MATHEMATICAL ANALYSIS II
If f is dierentiable on an interval, we can consider
the dierentiability of the function f
/
. The derivative
of f
/
is denoted by f
//
and called the second derivative
of f. Higher derivatives are dened similarly. The n-th
derivative (where it exists) is denoted by f
(n)
.
Proposition 5.14. (Second Derivative Test) Let f
be dened on an interval I and let its derivative f
/
exist on I. Suppose that c is an interior point of I
such that f
/
(c) = 0 and f
//
(c) exists.
(i) If f
//
(c) > 0, then f has a relative minimum
at c.
(ii) If f
//
(c) < 0, then f has a relative maximum
at c.
Proof. (i): Suppose (f
/
)
/
(c) = f
//
(c) > 0. By Lemma 5.6
(applied to f
/
), there exists > 0 such that
f
/
(x)
_
_
_
< f
/
(c) = 0 if x (c , c)
> f
/
(c) = 0 if x (c, c + ).
By First Derivative Test, f has a relative minimum at
c. The proof of (ii) is similar.
MA3110 MATHEMATICAL ANALYSIS II 115
Remarks.
(1) The condition f
/
(c) = 0 in Second Derivative Test
cannot be omitted.
Example : Let f(x) = x
2
, x R.
Then f
/
(x) = 2x, f
//
(x) = 2, x R.
Now f
/
(0) = 0 and f
//
(0) > 0. By Second Deriva-
tive Test, f has a relative minimum at 0.
However, f
//
(1) > 0 but 1 is not a relative extremum
point for f (here f
/
(1) ,= 0).
(2) If f
//
(c) = 0, Second Derivative Test is inconclusive.
Examples :
(a) Let f(x) = x
3
, x R. Then f
/
(0) = f
//
(0) = 0,
but 0 is not a relative extremum point for f.
(b) Let g(x) = x
4
, x R. Then g
/
(0) = g
//
(0) = 0,
and 0 is a relative minimum point for g by applying
First Derivative Test: g
/
(x) = 4x
3
_
_
_
< 0 if x < 0
> 0 if x > 0.
(3) Second Derivative Test is easier to be applied, but
it is less powerful than First Derivative Test (see
Example (b) in (2) above).
116 MA3110 MATHEMATICAL ANALYSIS II
Theorem 5.15. (LHospitals Rule: 0/0 case) Let f
and g be dierentiable on (a, b), and assume that
g(x) ,= 0 and g
/
(x) ,= 0 for all x (a, b). Sup-
pose that lim
xa+
f(x) = 0 = lim
xa+
g(x) and that
lim
xa+
f
/
(x)
g
/
(x)
equals L R , .
Then lim
xa+
f(x)
g(x)
equals L.
Remark. LHospitals Rule (with obvious modica-
tions) remains valid if the right-hand limits as x
a+ are replaced throughout by the left-hand limits as
x b, or replaced throughout by the limits as
x c, or as x , or as x .
MA3110 MATHEMATICAL ANALYSIS II 117
Proof. Let a < y < x < b. Then by Rolles Theo-
rem, g(x) ,= g(y) (why?). By Cauchys Mean value
Theorem, there exists u (y, x) such that
(5.4)
f(x) f(y)
g(x) g(y)
=
f
/
(u)
g
/
(u)
.
(i): Suppose lim
xa+
f
/
(x)/g
/
(x) = L R. Given
> 0, there exists > 0 such that if u (a, a + ),
then L <
f
/
(u)
g
/
(u)
< L + .
Hence if a < y < x < a + , then by (5.4) and the
above inequality, we have L <
f(x)f(y)
g(x)g(y)
< L + .
Taking limit as y a+, we have
L
f(x)
g(x)
L + for all a < x < a + .
(ii): Suppose lim
xa+
f
/
(x)/g
/
(x) = .
Given M > 0, there exists > 0 such that if u
(a, a + ), then
f
/
(u)
g
/
(u)
> M.
Hence if a < y < x < a + , then by (5.4) and the
above inequality, we have
f(x)f(y)
g(x)g(y)
> M.
Taking limit as y a+, we have
f(x)
g(x)
M for all a < x < a + .
(iii) The proof of the case L = is similar.
118 MA3110 MATHEMATICAL ANALYSIS II
5.3 Taylors Theorem and applications
Theorem 5.16. (Taylors Theorem) Let f be a func-
tion such that f, f
/
, . . . , f
(n)
are continuous on [a, x]
and f
(n+1)
exists on (a, x). Then there exists c
(a, x) such that
f(x) =
n

k=0
f
(k)
(a)
k!
(x a)
k
+
f
(n+1)
(c)
(n + 1)!
(x a)
n+1
.
Proof. There exists a number R
n
(depending on a, x
and n) such that
f(x) = f(a) + f
/
(a)(x a) +
f
//
(a)
2!
(x a)
2
+. . . +
f
(n)
(a)
n!
(x a)
n
+ R
n
.
Dene a new function g on [a, x] by
g(t) = f(x) f(t) f
/
(t)(x t)
f
//
(t)
2!
(x t)
2
. . .
f
(n)
(t)
n!
(x t)
n
R
n
(x t)
n+1
(x a)
n+1
.
MA3110 MATHEMATICAL ANALYSIS II 119
Then g is continuous on [a, x]. Dierentiating with
respect to the variable t, we have
g
/
(t) = f
/
(t) + f
/
(t) f
//
(t)(x t)
+ 2
f
//
(t)
2!
(x t)
f
(3)
(t)
2!
(x t)
2
+. . .
+ n
f
(n)
(t)
n!
(x t)
n1

f
(n+1)
(t)
n!
(x t)
n
+ (n + 1) R
n
(x t)
n
(x a)
n+1
=
f
(n+1)
(t)
n!
(x t)
n
+ (n + 1)R
n
(x t)
n
(x a)
n+1
.
Since g(x) = g(a) = 0, by Rolles Theorem, there exists
c (a, x) such that g
/
(c) = 0.
Therefore, R
n
=
f
(n+1)
(c)
(n+1)!
(x a)
n+1
.
120 MA3110 MATHEMATICAL ANALYSIS II
Example 5.7. Suppose that a function f : R R
is dierentiable on R, f
/
(x) = f(x) for all x R and
f(0) = 1. What must f be?
Since f is dierentiable on R and f
/
= f, f
/
is dier-
entiable on R and f
//
= (f
/
)
/
= f
/
= f.
Continuing in this way, we conclude that for any n N,
f is n-times dierentiable on R and f
(n)
= f.
Fix a number x R, x ,= 0.
By Taylors Theorem (taking a = 0), there exists c
(depending on x and n) between 0 and x such that
f(x) =
n

k=0
f
(k)
(0)
k!
x
k
+
f
(n+1)
(c)
(n + 1)!
x
n+1
=
n

k=0
f(0)
k!
x
k
+
f(c)
(n + 1)!
x
n+1
=
n

k=0
1
k!
x
k
+
f(c)
(n + 1)!
x
n+1
.
Let P
n
(x) be the sum

n
k=1
x
k
k!
= 1 +x+
x
2
2
+. . . +
x
n
n!
.
Then f(x) P
n
(x) = f(c)
x
n+1
(n+1)!
for some c between 0
and x. (This c may be dierent for dierent ns.)
MA3110 MATHEMATICAL ANALYSIS II 121
Since f is dierentiable on the interval I := [0, x] or
[x, 0], depending on whether x > 0 or x < 0, it is con-
tinuous on the same interval (Proposition 5.1).
Therefore f is bounded on I by Proposition 4.4.
Choose M (depending on x but not on n) such that
[f(u)[ M for all u I. In particular, [f(c)[ M.
Thus [f(x)P
n
(x)[ M[x[
n+1
/(n+1)! for all n N.
Claim : lim
n
[x[
n
n!
= 0.
We will use the fact that 2
n
n for all n N. (Ex.)
Choose N
1
N such that 2[x[ N
1
. If n N
1
,
[x[
n
n!
[x[
N
1
[x[
nN
1
(N
1
+ 1)(N
1
+ 2) n
= [x[
N
1
1
2
nN
1
2[x[
N
1
+ 1

2[x[
N
1
+ 2

2[x[
n
[x[
N
1
1
2
nN
1
= (2[x[)
N
1
1
2
n
(2[x[)
N
1

1
n
.
122 MA3110 MATHEMATICAL ANALYSIS II
Given > 0, choose N
2
Nsuch that N
2
> (2[x[)
N
1

.
If n maxN
1
, N
2
, then
[x[
n
n!
(2[x[)
N
1

1
n
(2[x[)
N
1

1
N
2
< .
This proves the claim.
Since

M[x[
n+1
(n + 1)!
f(x) P
n
(x)
M[x[
n+1
(n + 1)!
,
by the Squeeze Theorem,
lim
n
(f(x) P
n
(x)) = 0.
By Proposition 3.5, lim
n
P
n
(x) = f(x).
Therefore we have
f(x) = lim
n
P
n
(x) =

k=0
x
k
k!
, x R.
(Note that this equation is trivially true for x = 0, since
f(0) = 1 = P
n
(0) for all n N.)
MA3110 MATHEMATICAL ANALYSIS II 123
6. The Riemann Integral
6.1 The Riemann integral and general prop-
erties
Denition. Let [a, b] be a closed and bounded inter-
val, a < b. A partition P of [a, b] is a nite (ordered)
set of points a = x
0
< x
1
< < x
n
= b.
Denition. Let f : [a, b] R be a bounded function
and let P = a = x
0
< x
1
< < x
n
= b be
a partition of [a, b]. The lower and upper sums of f
with respect to the partition P are dened to be
L(f, P) =
n

k=1
m
k
(x
k
x
k1
) =
n

k=1
m
k
x
k
U(f, P) =
n

k=1
M
k
(x
k
x
k1
) =
n

k=1
M
k
x
k
respectively, where for each k = 1, . . . , n,
m
k
= m
k
(f, P) = inf f(x) : x [x
k1
, x
k
],
M
k
= M
k
(f, P) = sup f(x) : x [x
k1
, x
k
],
x
k
= x
k
x
k1
.
124 MA3110 MATHEMATICAL ANALYSIS II
Example 6.1. Let f : [0, ] R, f(x) = sin x and
let P = 0,

6
,

4
,
5
12
,
3
4
, .
Then
m
1
= 0, m
2
= sin

6
, m
3
= sin

4
, m
4
= sin
3
4
, m
5
= 0
M
1
= sin

6
, M
2
= sin

4
, M
3
= sin
5
12
, M
4
= sin

2
,
M
5
= sin
3
4
.
Hence
L(f, P) = 0(

6
) + (sin

6
)(

4


6
) + (sin

4
)(
5
12


4
)
+ (sin
3
4
)(
3
4

5
12
) + 0(
3
4
)
U(f, P) = (sin

6
)(

6
) + (sin

4
)(

4


6
) + (sin
5
12
)(
5
12


4
)
+ (sin

2
)(
3
4

5
12
) + (sin
3
4
)(
3
4
).
MA3110 MATHEMATICAL ANALYSIS II 125
Example 6.2. (Dirichlet, Example 4.3)
Let f : [0, 1] R be dened by
f(x) =
_
_
_
1 if x is rational
0 if x is irrational.
If P = 0 = x
0
< x
1
< < x
n
= 1 is any partition
of [0, 1],
then m
k
= 0 for any k and M
k
= 1 for any k.
Hence L(f, P) = 0 and U(f, P) = 1 for any partition
P of [0, 1].
126 MA3110 MATHEMATICAL ANALYSIS II
Proposition 6.1. Let f : [a, b] R be a bounded
function and let P be a partition of [a, b].
Let m = inff(x) : x [a, b]
and M = supf(x) : x [a, b].
Then m(b a) L(f, P) U(f, P) M(b a).
MA3110 MATHEMATICAL ANALYSIS II 127
Proof. Let P = a = x
0
< x
1
< < x
n
= b.
For any k = 1, . . . , n and any x [x
k1
, x
k
] [a, b],
we have f(x) supf(u) : u [a, b] = M.
Thus M is an upper bound of f(x) : x [x
k1
, x
k
].
So M
k
= supf(x) : x [x
k1
, x
k
] M.
Similarly, m m
k
for any k.
Therefore,
U(f, P) =
n

k=1
M
k
(x
k
x
k1
)
n

k=1
M(x
k
x
k1
)
= M(x
n
x
0
) = M(b a);
L(f, P) =
n

k=1
m
k
(x
k
x
k1
)
n

k=1
m(x
k
x
k1
)
= m(x
n
x
0
) = m(b a).
Clearly m
k
M
k
for all k. Therefore,
L(f, P) =
n

k=1
m
k
(x
k
x
k1
)
n

k=1
M
k
(x
k
x
k1
) = U(f, P).

128 MA3110 MATHEMATICAL ANALYSIS II


Denition. Let f : [a, b] R be a bounded function.
The lower and upper integrals of f on [a, b] are
L(f) = sup L(f, P) : P is a partition of [a,b]
U(f) = inf U(f, P) : P is a partition of [a,b].
Other notations for the lower and upper integrals:
L(f) =
_
b
a
f =
_
f U(f) =
_
b
a
f =
_
f
Example 6.3. (see Example 6.2) Let f : [0, 1] R,
f(x) =
_
_
_
1 if x is rational
0 if x is irrational.
It was previously shown that L(f, P) = 0 and
U(f, P) = 1 for any partition P of [0, 1].
Hence
U(f) = infU(f, P) : P is a partition of [0, 1] = 1
and
L(f) = supL(f, P) : P is a partition of [0, 1] = 0.
MA3110 MATHEMATICAL ANALYSIS II 129
Example 6.4. Let f : [0, ] R, f(x) = sin x and
let P = 0,

6
,

4
,
5
12
,
3
4
, . From Example 6.1,
L(f, P) = 0(

6
) + (sin

6
)(

4


6
) + (sin

4
)(
5
12


4
)
+ (sin
3
4
)(
3
4

5
12
) + 0(
3
4
)
U(f, P) = (sin

6
)(

6
) + (sin

4
)(

4


6
) + (sin
5
12
)(
5
12


4
)
+ (sin

2
)(
3
4

5
12
) + (sin
3
4
)(
3
4
).
Suppose we add a point

3
to the partition P, and
let Q = 0,

6
,

4
,

3
,
5
12
,
3
4
, . Then L(f, P)
L(f, Q) = 0(

6
) + (sin

6
)(

4


6
) + (sin

4
)(

3


4
)
+ (sin

3
)(
5
12


3
) + (sin
3
4
)(
3
4

5
12
) + 0(
3
4
)
U(f, Q) = (sin

6
)(

6
) + (sin

4
)(

4


6
) + (sin

3
)(

3


4
)
+ (sin
5
12
)(
5
12


3
) + (sin

2
)(
3
4

5
12
)
+ (sin
3
4
)(
3
4
) U(f, P).
Adding a point to the partition increases the lower sum
and decreases the upper sum.
130 MA3110 MATHEMATICAL ANALYSIS II
Proposition 6.2. Let f : [a, b] R be a bounded
function and let P and Q be partitions of [a, b].
Then
L(f, P) L(f, P Q) U(f, P Q) U(f, Q).
Consequently, L(f) U(f).
Proof. Suppose R = a = x
0
< < x
n
= b is a
partition of [a, b] and S = R c for some c [a, b].
Let M
k
= supf(x) : x [x
k1
, x
k
], 1 k n.
Suppose that x
j1
< c < x
j
.
Set y = supf(x) : x [x
j1
, c] and
z = supf(x) : x [c, x
j
].
Observe that y, z M
j
. Then
U(f, R) U(f, S) = M
j
(x
j
x
j1
)
(y(c x
j1
) + z(x
j
c))
= (M
j
y)(c x
j1
) + (M
j
z)(x
j
c)
0.
This shows that adding one point to a partition de-
creases the value of the upper sum.
Since it is possible to go from the partition Q to the
MA3110 MATHEMATICAL ANALYSIS II 131
partition P Q by the addition of one point at a time
for a nite number of times, U(f, P Q) U(f, Q).
Similarly, L(f, P) L(f, P Q).
Also, L(f, P Q) U(f, P Q) by Proposition 6.1.
Thus
L(f, P) L(f, P Q) U(f, P Q) U(f, Q).
So L(f, P) U(f, Q) for any partitions P and Q.
In particular, for a xed partition P, L(f, P) is a lower
bound of the set U(f, Q) : Q is a partition of [a, b].
Hence
L(f, P) infU(f, Q) : Q is a partition of [a, b] = U(f).
But this shows that U(f) is an upper bound of the set
L(f, P) : P is a partition of [a, b].
Hence
U(f) supL(f, Q) : Q is a partition of [a, b] = L(f).

132 MA3110 MATHEMATICAL ANALYSIS II


Denition. A bounded function f : [a, b] R is said
to be (Riemann) integrable on [a, b] if
_
b
a
f = L(f) = U(f) =
_
b
a
f.
In this case, the common value is called the (Riemann)
integral of f on [a, b] and is denoted by
_
b
a
f or
_
b
a
f(x) dx.
We also dene
_
a
b
f =
_
b
a
f.
(Example 6.3) The function f : [0, 1] R,
f(x) =
_
_
_
1 if x is rational
0 if x is irrational.
is not Riemann integrable because it was shown in Ex-
ample 6.3 that U(f) = 1 and L(f) = 0.
MA3110 MATHEMATICAL ANALYSIS II 133
Example 6.5. Consider any constant function
f : [a, b] R, f(x) = c.
For any partition P = a = x
0
< < x
n
= b,
m
k
= inff(x) : x [x
k1
, x
k
] = c and
M
k
= supf(x) : x [x
k1
, x
k
] = c for all k.
Hence
L(f, P) =

n
k=1
c(x
k
x
k1
) = c(b a)
and U(f, P) =

n
k=1
c(x
k
x
k1
) = c(b a).
Thus L(f) = c(b a) and U(f) = c(b a).
So f is integrable on [a, b] and
_
b
a
f = c(b a).
134 MA3110 MATHEMATICAL ANALYSIS II
Example 6.6. Consider the function h : [0, 1] R,
h(x) = x. Well compute
_
1
0
h using the denition of
the Riemann integral.
For any n N, let P
n
be the partition
0, 1/n, 2/n, . . . , (n 1)/n, 1 of [0, 1].
For k = 0, 1, . . . , n, let x
k
= k/n.
For k = 1, . . . , n, let m
k
= infh(x) : x [x
k1
, x
k
]
and M
k
= suph(x) : x [x
k1
, x
k
].
Then m
k
= x
k1
and M
k
= x
k
.
Hence
L(h, P
n
) =
n

k=1
m
k
(x
k
x
k1
) =
n

k=1
k 1
n

1
n
,
U(h, P
n
) =
n

k=1
M
k
(x
k
x
k1
) =
n

k=1
k
n

1
n
.
Hence
L(h, P
n
) =
1
n
2
n

k=1
(k 1) =
1
n
2

n(n 1)
2
=
n 1
2n
,
U(h, P
n
) =
1
n
2
n

k=1
k =
1
n
2

n(n + 1)
2
=
n + 1
2n
.
MA3110 MATHEMATICAL ANALYSIS II 135
Thus
L(h) = supL(h, P) : P is a partition of [0, 1]
sup
n
L(h, P
n
) = sup
n
n 1
2n
=
1
2
,
U(h) = infU(h, P) : P is a partition of [0, 1]
inf
n
U(h, P
n
) = inf
n
n + 1
2n
=
1
2
.
So U(h) 1/2 L(h).
On the other hand, L(h) U(h) by Proposition 6.2.
Therefore, U(h) = L(h) = 1/2.
Thus h is integrable on [0, 1] and
_
1
0
h = 1/2.
136 MA3110 MATHEMATICAL ANALYSIS II
Proposition 6.3. Let f : [a, b] R be a bounded
and integrable function and set
m = inff(x) : x [a, b],
M = supf(x) : x [a, b].
Then
(i) m(b a)
_
b
a
f M(b a).
(ii) If f(x) 0 for all x [a, b], then
_
b
a
f 0.
Proof. (i): By Proposition 6.1, U(f, P) M(b a)
for any partition P.
Hence U(f) M(b a).
Similarly, L(f) m(b a).
If f is integrable, then
_
b
a
f = U(f) = L(f) and the
inequalities in (i) hold.
(ii) follows directly from (i) with m 0.
MA3110 MATHEMATICAL ANALYSIS II 137
6.2 Riemann integrability criterion and con-
sequences
Theorem 6.4. (Riemann Integrability Criterion) Let
f : [a, b] R be a bounded function. Then f is
integrable on [a, b] if and only if for any > 0, there
exists a partition P of [a, b] such that
U(f, P) L(f, P) < .
Proof. () Assume that f is integrable.
Let > 0 be given.
Choose a partition Q such that U(f, Q) < U(f) +/2.
Choose another partition R such that
L(f, R) > L(f) /2.
Let P = Q R. By Proposition 6.2,
L(f, R) L(f, P) U(f, P) U(f, Q).
Thus
U(f, P)L(f, P) U(f, Q)L(f, R) < U(f)L(f)+,
which is , as U(f) = L(f) since f is integrable.
138 MA3110 MATHEMATICAL ANALYSIS II
() Given > 0, there eixsts a partition P of [a, b]
such that U(f, P) L(f, P) < .
Since L(f, P) L(f) U(f) U(f, P),
0 U(f) L(f) U(f, P) L(f, P) < .
Since is arbitrary, we have U(f) L(f) = 0.
MA3110 MATHEMATICAL ANALYSIS II 139
Example 6.7. Show that the function f : [0, 1] R,
f(x) = x
3
, is integrable on [0, 1].
Proof. We will use the Riemann Integrability Criterion.
Let > 0 be given. Choose N N such that N > 1/.
Let P be the partition 0,
1
N
,
2
N
, . . . ,
N 1
N
, 1.
Let x
k
=
k
N
for k = 0, 1, . . . , N. For k = 1, . . . , N, let
m
k
= inff(x) : x [x
k1
, x
k
]
and M
k
= supf(x) : x [x
k1
, x
k
].
Then m
k
= f(x
k1
) = (
k1
N
)
3
and M
k
= f(x
k
) = (
k
N
)
3
.
Hence
U(f, P) L(f, P) =
N

k=1
M
k
N

N

k=1
m
k
N
=
N

k=1
M
k
m
k
N
=
N

k=1
k
3
(k 1)
3
N
4
=
N
3
0
3
N
4
=
1
N
< .
Thus f satises the Riemann Integrability Criterion
and hence is integrable on [0, 1].
140 MA3110 MATHEMATICAL ANALYSIS II
Proposition 6.5. Let f : [a, b] R be a bounded
monotone function. Then f is integrable on [a, b].
Proof. Assume that f is increasing.
Let > 0 be given.
Choose N N such that N > (b a)(f(b) f(a))/.
Let P be the partition a = x
0
< x
1
< < x
N
= b,
where x
k
= a + k
ba
N
for k = 0, 1, . . . , N.
For k = 1, . . . , N, let m
k
= inff(x) : x [x
k1
, x
k
]
and M
k
= supf(x) : x [x
k1
, x
k
].
Then m
k
= f(x
k1
) and M
k
= f(x
k
).
Hence
U(f, P) L(f, P) =
N

k=1
M
k
b a
N

N

k=1
m
k
b a
N
=
b a
N
N

k=1
(f(x
k
) f(x
k1
))
=
(b a)
N
(f(b) f(a)) < .
Thus f satises the Riemann Integrability Criterion
and hence is integrable on [a, b].
The proof is similar if f is decreasing.
MA3110 MATHEMATICAL ANALYSIS II 141
Theorem 6.6. Let f : [a, b] R be continuous on
[a, b]. Then f is integrable on [a, b].
Proof. By Theorem 4.9, f is uniformly continuous on
[a, b]. Given > 0, there exists > 0 such that
[f(x) f(y)[ < /(2(b a))
whenever x, y [a, b] and [x y[ < .
Choose N N such that N > (b a)/.
Let P be the partition a = x
0
< x
1
< < x
N
= b,
where x
k
= a + k
ba
N
for k = 0, 1, . . . , N.
For k = 1, . . . , N, let m
k
= inff(x) : x [x
k1
, x
k
]
and M
k
= supf(x) : x [x
k1
, x
k
].
Since f is continuous on [x
k1
, x
k
], by Extreme Value
Theorem, there exist c
k
, d
k
[x
k1
, x
k
] such that
m
k
= f(c
k
) and M
k
= f(d
k
).
Since [d
k
c
k
[ [x
k
x
k1
[ = (b a)/N < ,
by uniform continuity of f, we have
M
k
m
k
= [f(d
k
) f(c
k
)[ < /(2(b a)).
142 MA3110 MATHEMATICAL ANALYSIS II
Therefore,
U(f, P) L(f, P) =
N

k=1
M
k
(b a)
N

N

k=1
m
k
(b a)
N
=
N

k=1
(M
k
m
k
)(b a)
N

k=1

2N
=

2
< .
Thus f satises the Riemann Integrability Criterion
and hence is integrable on [a, b].
MA3110 MATHEMATICAL ANALYSIS II 143
Lemma 6.7. Let f, g : [a, b] R be bounded func-
tions, P a partition of [a, b] and c R. Then
(i) L(cf, P) =
_
_
_
c L(f, P) if c > 0
c U(f, P) if c < 0.
(ii) U(cf, P) =
_
_
_
c U(f, P) if c > 0
c L(f, P) if c < 0.
(iii) L(f, P) + L(g, P) L(f + g, P)
U(f + g, P) U(f, P) + U(g, P).
Proof. (i) and (ii) follow from the fact that for any
nonempty bounded subset S of R,
inf cS =
_
_
_
c inf S if c > 0
c sup S if c < 0,
sup cS =
_
_
_
c sup S if c > 0
c inf S if c < 0.
(iii): Let P = a = x
0
< < x
n
= b.
If 1 k n and x [x
k1
, x
k
], then
m
k
(f) f(x) M
k
(f)
m
k
(g) g(x) M
k
(g).
Hence m
k
(f) +m
k
(g) (f +g)(x) M
k
(f) +M
k
(g).
Thus m
k
(f) + m
k
(g) m
k
(f + g),
144 MA3110 MATHEMATICAL ANALYSIS II
and M
k
(f + g) M
k
(f) + M
k
(g).
Therefore,
L(f, P) + L(g, P) =
n

k=1
(m
k
(f) + m
k
(g))(x
k
x
k1
)

k=1
m
k
(f + g)(x
k
x
k1
)
= L(f + g, P) U(f + g, P)
=
n

k=1
M
k
(f + g)(x
k
x
k1
)

k=1
(M
k
(f) + M
k
(g))(x
k
x
k1
)
= U(f, P) + U(g, P).

MA3110 MATHEMATICAL ANALYSIS II 145


Lemma A. Let S be a nonempty bounded subset of
R and K a positive constant such that [s t[ K
whenever s, t S. Then sup S inf S K.
Proof. Let > 0. Then there exist s, t S such that
s > sup S

2
and t < inf S +

2
.
Hence
sup S inf S < (s +

2
) (t

2
)
= s t +
[s t[ +
K + .
Since is arbitrary, we have sup S inf S K.
146 MA3110 MATHEMATICAL ANALYSIS II
Proposition 6.8. Suppose that f, g : [a, b] R are
integrable on [a, b] and c R. Then
(i) cf is integrable on [a, b] and
_
b
a
cf = c
_
b
a
f.
(ii) f + g is integrable on [a, b] and
_
b
a
(f + g) =
_
b
a
f +
_
b
a
g.
(iii) If f(x) g(x) for all x [a, b], then
_
b
a
f
_
b
a
g.
(iv) [f[ is integrable on [a, b] and [
_
b
a
f[
_
b
a
[f[.
(v) fg is integrable on [a, b].
MA3110 MATHEMATICAL ANALYSIS II 147
Proof. (ii): Let > 0 be given.
There exist partitions P and Q such that
U(f, P) < U(f) +

2
and U(g, Q) < U(g) +

2
.
Let R = P Q. By Lemma 6.7 and Proposition 6.2,
U(f + g, R) U(f, R) + U(g, R)
U(f, P) + U(g, Q)
< U(f) +

2
+ U(g) +

2
= U(f) + U(g) + .
Hence for any > 0,
U(f + g) U(f + g, R) < U(f) + U(g) + .
Therefore, U(f + g) U(f) + U(g).
Similarly, L(f) + L(g) L(f + g). (Ex.)
Since f and g are integrable on [a, b],
L(f) = U(f) =
_
b
a
f and L(g) = U(g) =
_
b
a
g.
Thus
_
b
a
f +
_
b
a
g L(f +g) U(f +g)
_
b
a
f +
_
b
a
g.
So f + g is integrable and
_
b
a
(f + g) =
_
b
a
f +
_
b
a
g.
148 MA3110 MATHEMATICAL ANALYSIS II
(iii): Suppose f(x) g(x) for all x [a, b]. Then
h(x) := g(x) f(x) 0 for all x [a, b]. By (i) and
(ii), h is integrable on [a, b] and
_
b
a
h =
_
b
a
g
_
b
a
f.
This value is 0 by Proposition 6.3.
(iv): Let > 0 be given.
Choose a partition P = a = x
0
< x
1
< < x
n
=
b such that U(f, P) L(f, P) < .
Let m
k
(f, P) = inff(x) : x [x
k1
, x
k
],
M
k
(f, P) = supf(x) : x [x
k1
, x
k
].
Similarly dene m
k
([f[, P) and M
k
([f[, P).
If u, v [x
k1
, x
k
], then
[[f(u)[[f(v)[[ [f(u)f(v)[ M
k
(f, P)m
k
(f, P).
By Lemma A,
sup[f(x)[ : x [x
k1
, x
k
]inf[f(x)[ : x [x
k1
, x
k
]
M
k
(f, P) m
k
(f, P),
i.e., M
k
([f[, P) m
k
([f[, P) M
k
(f, P) m
k
(f, P).
Thus
U([f[, P) L([f[, P)
=

n
k=1
(M
k
([f[, P) m
k
([f[, P))(x
k
x
k1
)
MA3110 MATHEMATICAL ANALYSIS II 149


n
k=1
(M
k
(f, P) m
k
(f, P))(x
k
x
k1
)
= U(f, P) L(f, P) < .
Hence [f[ is integrable on [a, b] by the Riemann Inte-
grability Criterion.
By (i), [f[ is also integrable on [a, b].
Since [f(x)[ f(x) [f(x)[ for all x [a, b],

_
b
a
[f[
_
b
a
f
_
b
a
[f[ by (iii).
Hence [
_
b
a
f[
_
b
a
[f[.
(v): Since f and g are integrable, they are bounded.
Let K be a constant such that [f(x)[ K and
[g(x)[ K for all x [a, b].
Let > 0 be given.
Choose partitions P and Q such that
U(f, P) L(f, P) <

2K
, U(g, Q) L(g, Q) <

2K
.
Let R = P Q = a = x
0
< x
1
< < x
n
= b, and
x
k
= x
k
x
k1
for k = 1, . . . , n. By Proposition 6.2,
U(f, R) L(f, R) U(f, P) L(f, P) <

2K
,
U(g, R) L(g, R) U(g, Q) L(g, Q) <

2K
.
Let m
k
(f, R) = inff(x) : x [x
k1
, x
k
],
150 MA3110 MATHEMATICAL ANALYSIS II
M
k
(f, R) = supf(x) : x [x
k1
, x
k
].
Similarly, dene m
k
(g, R), M
k
(g, R)), m
k
(fg, R), M
k
(fg, R)).
If u, v [x
k1
, x
k
], then
[(fg)(u) (fg)(v)[
[f(u)g(u) f(v)g(u)[ + [f(v)g(u) f(v)g(v)[
K[f(u) f(v)[ + K[g(u) g(v)[
K(M
k
(f, R)m
k
(f, R))+K(M
k
(g, R)m
k
(g, R))
:= d
k
.
By Lemma A, M
k
(fg, R) m
k
(fg, R) d
k
. Thus
U(fg, R) L(fg, R) =
n

k=1
(M
k
(fg, R) m
k
(fg, R)) x
k
K
n

k=1
(M
k
(f, R) m
k
(f, R)) x
k
+ K
n

k=1
(M
k
(g, R) m
k
(g, R)) x
k
= K(U(f, R) L(f, R)) + K(U(g, R) L(g, R))
< K(

2K
+

2K
) = .
Thus fg is integrable by the Riemann Integrability Cri-
terion.
MA3110 MATHEMATICAL ANALYSIS II 151
Proposition 6.9. Suppose that f is integrable on
both of the intervals [a, b] and [b, c]. Then f is inte-
grable on [a, c] and
_
c
a
f =
_
b
a
f +
_
c
b
f.
Proof. First observe that if P is a partition of [a, b] and
Q is a partition of [b, c], then P Q is a partition of
[a, c] and
U(f, P Q) = U(f, P) + U(f, Q),
L(f, P Q) = L(f, P) + L(f, Q).
Given > 0, choose a partition P of [a, b] and a parti-
tion Q of [b, c] such that
U(f, P) <
_
b
a
f +

2
and U(f, Q) <
_
c
b
f +

2
.
Then
U(f)(over the interval [a, c])
U(f, P Q) = U(f, P) +U(f, Q) <
_
b
a
f +
_
c
b
f +
for any > 0.
Hence U(f)(over the interval [a, c])
_
b
a
f +
_
c
b
f.
Similarly, L(f)(over the interval [a, c])
_
b
a
f +
_
c
b
f.
Thus f is integrable on [a, c] and
_
c
a
f =
_
b
a
f +
_
c
b
f.

152 MA3110 MATHEMATICAL ANALYSIS II


Proposition 6.10. Suppose that f is integrable on
the interval [a, b]. For any c (a, b), f is integrable
on [a, c] and [c, b].
Proof. Given > 0, there exists a partition P of [a, b]
such that U(f, P) L(f, P) < .
Let Q = P c. By Proposition 6.2,
U(f, Q) L(f, Q) U(f, P) L(f, P) < .
Let R = Q [a, c]. Then R is a partition of [a, c].
Also U(f, R) L(f, R) U(f, Q) L(f, Q). (Ex.)
Thus U(f, R) L(f, R) < .
So f is integrable on [a, c] by the Riemann Integrability
Criterion.
The proof for the interval [c, b] is similar.
MA3110 MATHEMATICAL ANALYSIS II 153
Example 6.8. The results from Proposition 6.5 to
Proposition 6.10 give us a wide variety of integrable
functions. For example, all of the following functions
are integrable on the indicated intervals. (Ex.)
(1) Any polynomial function on any interval [a, b].
(2) f
1
(x) = x|, on any interval [a, b].
(3) f
2
(x) = [x|[, on any interval [a, b].
(4) f
3
(x) = x
2
x| +
1
x
2
+ 1
, on any interval [a, b].
(5)
f
4
(x) =
_
_
_
[x[ if [x[ 1
2 if 1 < x 2,
on the interval [1, 2].
154 MA3110 MATHEMATICAL ANALYSIS II
6.3 Fundamental Theorem of Calculus
Proposition 6.11. Let f be integrable on [a, b]. De-
ne F : [a, b] R by F(x) =
_
x
a
f, where we inter-
pret
_
a
a
f to be 0. Then F is continuous on [a, b].
Proof. Let > 0.
Since f is bounded, there exists M > 0 such that
[f(x)[ M for all x [a, b].
Choose = /M. If x, c [a, b] and [xc[ < , then
[F(x) F(c)[ = [
_
x
a
f
_
c
a
f[
=
_

_
[
_
x
c
f[ if c < x
[
_
c
x
f[ if x < c
0 if x = c.
(Proposition 6.9)

_
_
x
c
[f[ if c < x
_
c
x
[f[ if x < c
0 if x = c.
(Proposition 6.8(iv))
M[x c[ < M = .
Hence F is (uniformly) continuous on [a, b].
MA3110 MATHEMATICAL ANALYSIS II 155
Theorem 6.12. (Fundamental Theorem of Calculus
I) Let f be integrable on [a, b]. Dene F : [a, b] R
by F(x) =
_
x
a
f, where we interpret
_
a
a
f to be 0.
If f is continuous at a point c [a, b], then F is
dierentiable at c and F
/
(c) = f(c).
156 MA3110 MATHEMATICAL ANALYSIS II
Proof. Suppose > 0 is given.
Since f is continuous at c, there exists > 0 such that
[f(x)f(c)[ < /2 whenever x [a, b] and [xc[ < .
If x [a, b] and c < x < c, then for all u [x, c],
f(c)

2
f(u) f(c) +

2
.
Since f is integrable on [x, c] (Proposition 6.10), and
constant functions are integrable (Example 6.5), we
have from Propositions 6.8 and 6.9 that
_
c
x
(f(c)

2
)
_
c
x
f
_
c
x
(f(c) +

2
)
(c x)(f(c)

2
) F(c) F(x) (c x)(f(c) +

2
)
f(c)

2

F(x) F(c)
x c
f(c) +

2

F(x) F(c)
x c
f(c)



2
< .
The proof is similar if x [a, b] and c < x < c + .
Thus
F
/
(c) = lim
xc
F(x) F(c)
x c
= f(c).

MA3110 MATHEMATICAL ANALYSIS II 157


Example 6.9. Let F : [0, 1] R, F(x) =
_
x
0
1
t
3
+1
dt.
Then F is dierentiable on [0, 1] and
F
/
(x) =
Example 6.10. Let f : [0, 1] R be continuous on
[0, 1] and dene G : [0, 1] R, G(x) =
_
1
x
f(t
2
) dt.
Then G is dierentiable on [0, 1] and
G
/
(x) =
158 MA3110 MATHEMATICAL ANALYSIS II
Example 6.11. Dene H : [0, 1] R by
H(x) =
_
x
2
0
sin(t
3
) dt.
Then H is dierentiable on [0, 1] and
H
/
(x) =
MA3110 MATHEMATICAL ANALYSIS II 159
Theorem 6.13. (Fundamental Theorem of Calculus
II) Let g be dierentiable on [a, b] and assume that
g
/
is continuous on [a, b]. Then
_
b
a
g
/
= g(b) g(a).
Proof. Since g
/
is continuous on [a, b], it is integrable
on [a, b] by Theorem 6.6.
Dene F : [a, b] R by F(x) =
_
x
a
g
/
.
By Theorem 6.12, F is dierentiable at any point c
[a, b] and F
/
(c) = g
/
(c).
By Proposition 5.12, there is a constant C such that
g(x) = F(x) + C for all x [a, b].
Then
g(b) g(a) = F(b) F(a)
=
_
b
a
g
/

_
a
a
g
/
=
_
b
a
g
/
.

160 MA3110 MATHEMATICAL ANALYSIS II


Example 6.12. Let f : [0, 1] R, f(x) = x
3
/3.
Then f is dierentiable on [0, 1].
f
/
(x) = x
2
is continuous on [0, 1].
By the Fundamental Theorem of Calculus II,
_
1
0
x
2
dx = f(1) f(0) = 1/3 0 = 1/3.
MA3110 MATHEMATICAL ANALYSIS II 161
The following result is stronger than Theorem 6.13:
Theorem 6.14. (Cauchys Fundamental Theorem of
Calculus) Let f be integrable on [a, b]. Suppose F :
[a, b] R is dierentiable on [a, b] and F
/
(x) = f(x)
for all x [a, b]. Then
_
b
a
f = F(b) F(a).
Remark. Roughly speaking, the Fundamental Theo-
rems of Calculus say that under appropriate conditions,
the operations of dierentiation and integration are in-
verses of each other:
d
dx
_
x
a
f(t) dt = f(x),
_
x
a
g
/
(t) dt = g(x) g(a).
162 MA3110 MATHEMATICAL ANALYSIS II
Proof. Let P = a = x
0
< x
1
< . . . < x
n
= b be a
partition of [a, b].
For k = 1, . . . , n, set m
k
= inff(x) : x [x
k1
, x
k
],
M
k
= supf(x) : x [x
k1
, x
k
], x
k
= x
k
x
k1
.
For each k, F is dierentiable on (x
k1
, x
k
) and con-
tinuous on [x
k1
, x
k
].
By the Mean Value Theorem, there exists c
k
(x
k1
, x
k
)
such that
F(x
k
) F(x
k1
) = f(c
k
)(x
k
x
k1
) = f(c
k
)x
k
.
Note that m
k
f(c
k
) M
k
.
Hence
m
k
x
k
F(x
k
) F(x
k1
) M
k
x
k
,
n

k=1
m
k
x
k

n

k=1
(F(x
k
) F(x
k1
))
n

k=1
M
k
x
k
,
L(f, P) F(b) F(a) U(f, P).
Thus L(f) F(b) F(a) U(f).
Since f is integrable, U(f) = L(f) =
_
b
a
f.
Thus
_
b
a
f = F(b) F(a).
MA3110 MATHEMATICAL ANALYSIS II 163
Theorem 6.15. (Integration by parts) Suppose the
functions u, v : [a, b] R are dierentiable on [a, b],
and their derivatives u
/
, v
/
are integrable on [a, b].
Then
_
b
a
uv
/
= u(b)v(b) u(a)v(a)
_
b
a
vu
/
.
Proof. Let F(x) = u(x)v(x) for x [a, b]. Then F is
dierentiable on [a, b]. By product rule and Proposi-
tion 6.8, F
/
= u
/
v + uv
/
is integrable on [a, b].
By Cauchys Fundamental Theorem of Calculus,
_
b
a
F
/
= F(b) F(a).
By Proposition 6.8 again, we have
_
b
a
uv
/
+
_
b
a
vu
/
= u(b)v(b) u(a)v(a),
which gives the desired result.
164 MA3110 MATHEMATICAL ANALYSIS II
Theorem 6.16. (Integration by substitution) Sup-
pose : [a, b] R is continuous on [a, b],
/
exists
and is integrable on [a, b]. If f : I R is continu-
ous on an interval I containing ([a, b]), then
_
b
a
f((t))
/
(t) dt =
_
(b)
(a)
f(x) dx .
Proof. Let F(x) =
_
x
(a)
f(u) du for x I. Since f is
continuous on I, by Theorem 6.12, F is dierentiable
on I and F
/
(x) = f(x), x I.
By Chain Rule, F is dierentiable on [a, b] and
(F)
/
(t) = F
/
((t))
/
(t) = f((t))
/
(t), t [a, b].
By Proposition 4.3, f is continuous on [a, b] and so
by Theorem 6.6 it is integrable on [a, b]. Therefore by
Proposition 6.8, (F)
/
is integrable on [a, b]. It follows
from Cauchys Fundamental Theorem of Calculus that
_
b
a
f((t))
/
(t) dt =
_
b
a
(F )
/
(t) dt
= (F )(b) (F )(a) =
_
(b)
(a)
f(u) du .

MA3110 MATHEMATICAL ANALYSIS II 165


We have the following Mean Value Theorems of Integral
Calculus.
Theorem 6.17. Suppose that f is continuous on
[a, b]. Then there exists c (a, b) such that
_
b
a
f = f(c)(b a).
Proof. Let F(x) =
_
x
a
f(u) du for x [a, b]. Since f
is continuous on [a, b], by Theorem 6.12, F is dieren-
tiable on [a, b] and F
/
(x) = f(x), x [a, b].
By Mean Value Theorem for derivatives (Theorem 5.9),
there exists c (a, b) such that
F(b) F(a) = F
/
(c)(b a),
i.e.,
_
b
a
f = f(c)(b a).
166 MA3110 MATHEMATICAL ANALYSIS II
Theorem 6.18. Suppose that f is continuous on
[a, b], g is integrable on [a, b] and g(x) has constant
sign on [a, b]. Then there exists c [a, b] such that
_
b
a
fg = f(c)
_
b
a
g.
Proof. We may assume that g(x) 0 for all x [a, b].
Let m = minf(x) : x [a, b]
and M = maxf(x) : x [a, b].
Then for all x [a, b], we have
mg(x) f(x)g(x) Mg(x).
Hence by Proposition 6.8,
(6.1) m
_
b
a
g
_
b
a
fg M
_
b
a
g.
Suppose
_
b
a
g > 0. Then m
_
b
a
fg/
_
b
a
g M.
By Intermediate Value Theorem, there exists c [a, b]
such that f(c) =
_
b
a
fg/
_
b
a
g, i.e.,
_
b
a
fg = f(c)
_
b
a
g.
If
_
b
a
g = 0, by (6.1)
_
b
a
fg = 0. In this case, the desired
result holds with any choice of c [a, b].
MA3110 MATHEMATICAL ANALYSIS II 167
Theorem 6.19. (Taylors Theorem with integral form
of remainder) Let f be a function such that f, f
/
, . . . , f
(n+1)
exist on [a, x] and f
(n+1)
is integrable on [a, x]. Then
f(x) =
n

k=0
f
(k)
(a)
k!
(xa)
k
+
1
n!
_
x
a
f
(n+1)
(t)(xt)
n
dt .
168 MA3110 MATHEMATICAL ANALYSIS II
6.4 Riemann sums
Denition. Let P = a = x
0
< x
1
< . . . < x
n
= b
be a partition of [a, b], and let x
k
= x
k
x
k1
for
k = 1, . . . , n. The norm |P| of P is given by
|P| = maxx
k
: k = 1, . . . , n.
Remark. If P and Q are partitions of [a, b] and P is
a renement of Q (i.e., P Q), then |P| |Q|.
Lemma 6.20. Let f : [a, b] R be a bounded func-
tion. Then for every > 0, there exists > 0 such
that
U(f, P) < U(f) + and L(f, P) > L(f)
for any partition P of [a, b] with |P| < .
Proof. We will prove the rst inequality. The proof of
the second inequality is similar.
Let > 0. Then there exists a partition Q of [a, b] such
that U(f, Q) < U(f) +

2
.
Let Q = a = y
0
< y
1
< . . . < y
N
= b and
= miny
j
: j = 1, . . . , N, where y
j
= y
j
y
j1
.
Choose > 0 such that < min,

6(N1)M
, where
MA3110 MATHEMATICAL ANALYSIS II 169
M = sup[f(x)[ : x [a, b].
Consider a partition P = a = x
0
< x
1
< . . . <
x
n
= b of [a, b] with |P| < (< ). Denoting
x
k
= x
k
x
k1
, since |P| < we have
(6.2) x
k
< y
j
for k = 1, . . . , n, j = 1, . . . , N.
Let R = P Q = a = z
0
< z
1
< . . . < z
r
= b and
z

= z

z
1
for = 1, . . . , r.
Then |R| |P| < < , and by Proposition 6.2,
(6.3) U(f, R) U(f, Q) < U(f) +

2
.
By (6.2), there is at most one point of Q between any
two consecutive points of P. Thus we separate the
subintervals of P into two classes: (P1) consisting of
those which contain a point of Q as its interior point;
and (P2) consisting of those which do not.
Similarly, the subintervals of R are separated into two
classes: (R1) for those which arise from a subinterval
of P by its containing a point of Q as its interior point;
and (R2) for all others, these being identical to those
of class (P2) above.
170 MA3110 MATHEMATICAL ANALYSIS II
In the dierence U(f, P) U(f, R), all contributions
from the subintervals of classes (P2) and (R2) cancel
each other. The remaining terms are from subintervals
of classes (P1) and (R1). From the denition of Q,
there are N 1 points of Q in (a, b). Hence there are
at most N 1 subintervals of class (P1) from P and
at most 2(N 1) subintervals of class (R1) from R.
In the expression U(f, P) U(f, R), the contribution
from each of these subintervals is of the form
M
k
(f, P)x
k
or M

(f, R)z

. Obviously we have
[M
k
(f, P)x
k
[ M and [M

(f, R)z

[ M.
Hence by triangle inequality
[U(f, P) U(f, R)[ 3(N 1)M <

2
.
Since P R, we have
(6.4)
0 U(f, P) U(f, R) = [U(f, P) U(f, R)[ <

2
.
By (6.3) and (6.4), we get
U(f, P) = (U(f, P)U(f, R))+U(f, R) <

2
+U(f)+

2
,
i.e., U(f, P) < U(f) + .
MA3110 MATHEMATICAL ANALYSIS II 171
Denition. Let f : [a, b] R be a bounded func-
tion.
(1) Let P = a = x
0
< x
1
< . . . < x
n
= b be a
partition of [a, b] and let
k
be a point in the k-th
subinterval [x
k1
, x
k
] for k = 1, . . . , n. The sum
S(f, P)() =
n

k=1
f(
k
)(x
k
x
k1
) =
n

k=1
f(
k
) x
k
is called the Riemann sum of f with respect to P
and = (
1
, . . . ,
n
).
(2) If there exists a real number A such that for every
> 0, there exists > 0 such that
[S(f, P)() A[ <
for any partition P of [a, b] with |P| < and any
choice of = (
1
, . . . ,
n
), then we say that A is the
limit of these Riemann sums as |P| 0 and we
denote this by
lim
|P|0
S(f, P)() = A.
172 MA3110 MATHEMATICAL ANALYSIS II
Remarks.
(1) For any partition P of [a, b] and any choice of
= (
1
, . . . ,
n
), we have
L(f, P) S(f, P)() U(f, P).
(2) Using the above notation of limit, we can rephrase
Lemma 6.20 as follows:
Let f : [a, b] R be a bounded function. Then
lim
|P|0
U(f, P) = U(f) and lim
|P|0
L(f, P) = L(f).
MA3110 MATHEMATICAL ANALYSIS II 173
The next result shows that the above notion of integra-
bility is equivalent to that dened in section 6.1
Theorem 6.21. Let f : [a, b] R be a bounded
function. Then f is integrable on [a, b] and
_
b
a
f = A
if and only if
lim
|P|0
S(f, P)() = A.
Proof. if: Suppose the limit exists and equals A.
Then for every > 0, there exists > 0 such that
A < S(f, P)() < A +
for any partition P of [a, b] with |P| < and any
choice of = (
1
, . . . ,
n
).
For each k = 1, . . . , n, choose
k
in [x
k1
, x
k
] such that
f(
k
) > M
k


ba
,
where M
k
= supf(x) : x [x
k1
, x
k
].
Letting = (
1
, . . . ,
n
), then
S(f, P)() =
n

k=1
f(
k
) x
k
>
n

k=1
M
k
x
k
= U(f, P).
Hence
A + > S(f, P)() > U(f, P) U(f) ,
174 MA3110 MATHEMATICAL ANALYSIS II
i.e., A > U(f) 2 for any > 0.
Therefore A U(f).
By a similar argument, we have A L(f)
and so U(f) A L(f).
Since L(f) U(f), we have L(f) = U(f) = A.
only if: Suppose that f is integrable on [a, b] and
L(f) = U(f) =
_
b
a
f = A.
By Lemma 6.20, for every > 0, there exists > 0
such that
U(f, P) < A + and L(f, P) > A
for any partition P of [a, b] with |P| < .
By a previous remark, for any such P and any choice
of = (
1
, . . . ,
n
), we have
A < L(f, P) S(f, P)() U(f, P) < A + ,
i.e., [S(f, P)() A[ < .
Hence lim
|P|0
S(f, P)() = A.
MA3110 MATHEMATICAL ANALYSIS II 175
6.5 Improper integrals
In dening the Riemann (or denite) integral
_
b
a
f(x) dx,
the function f is assumed to be bounded on the bounded
closed interval [a, b]. An improper integral
_
b
a
f(x) dx
is one whose
(I) integrand f is unbounded on (a, b), or
(II) interval of integration is unbounded.
Denition. Let b R , and f be a real-
valued function dened on a neighbourhood of b, except
possibly at b. The point b is a singularity of f if
b = , or f is unbounded on every neighbourhood
of b.
176 MA3110 MATHEMATICAL ANALYSIS II
Denition 1. Suppose that f is dened on [a, b),
b is a singularity of f and f is integrable on [a, c] for
every c (a, b). Then the improper integral
_
b
a
f(x) dx
converges and
_
b
a
f(x) dx = lim
cb

_
c
a
f(x) dx
if this limit exists in R. The improper integral
_
b
a
f(x) dx
diverges if the above limit does not exist in R.
Denition 2. We have a similar denition if f is
dened on (a, b], where a is a singularity of f:
_
b
a
f(x) dx = lim
ca
+
_
b
c
f(x) dx.
MA3110 MATHEMATICAL ANALYSIS II 177
Denition 3. If f is dened on [a, b] c, where
a < c < b and c is the only singularity of f on [a, b],
then the improper integral
_
b
a
f(x) dx converges and
_
b
a
f(x) dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx
= lim
tc

_
t
a
f(x) dx + lim
sc
+
_
b
s
f(x) dx
if both limits exist in R. The improper integral
_
b
a
f(x) dx
diverges if (at least) one of these limits does not exist
in R.
Remark. The above right hand side is stronger than
the existence of the limit
lim
0
+
_
_
c
a
f(x) dx +
_
b
c+
f(x) dx
_
,
called the Cauchy principal value of
_
b
a
f(x) dx.
178 MA3110 MATHEMATICAL ANALYSIS II
Example 6.13. f(x) =
1
x
2
has a singularity at 0, and
_
1
1
dx
x
2
=
_
0
1
dx
x
2
+
_
1
0
dx
x
2
= lim
t0

_
t
1
dx
x
2
+ lim
s0
+
_
1
s
dx
x
2
= lim
t0

(
1
t
1) + lim
s0
+
(1 +
1
s
).
Both limits do not exist in R. Hence
_
1
1
dx
x
2
diverges.
Remarks.
(1) If we try to evaluate the integral
_
1
1
dx
x
2
by the
Fundamental Theorem of Calculus, we have
_
1
1
dx
x
2
=
_

1
x
_
1
1
= 1 1 = 2.
This is a contradiction, since the integrand is non-
negative on [1, 1] 0.
(2) The Cauchy principal value of
_
1
1
dx
x
2
is
lim
0
+
_
_

1
dx
x
2
+
_
1

dx
x
2
_
= lim
0
+
_
(
1

1) + (1 +
1

)
_
,
which does not exist in R.
MA3110 MATHEMATICAL ANALYSIS II 179
Denition 4. Suppose that f is dened on (a, b),
a and b are the only singularities of f on [a, b] and
f is integrable on every closed bounded subinterval of
(a, b). Take any c in (a, b). Then the improper integral
_
b
a
f(x) dx converges and
_
b
a
f(x) dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx
= lim
ta
+
_
c
t
f(x) dx + lim
sb

_
s
c
f(x) dx
if both limits exist in R.
Remarks.
(1) The above right hand side is independent of the
choice of c (Ex.)
(2) When a = and b = , the above becomes
_

f(x) dx = lim
t
_
c
t
f(x) dx + lim
s
_
s
c
f(x) dx.
Existence of the limits on the right hand side
is stronger than existence of lim
s
_
s
s
f(x) dx
(the Cauchy principal value of
_

f(x) dx ).
180 MA3110 MATHEMATICAL ANALYSIS II
Example 6.14. (Singularities at 0 and )
_

0
dx

x
=
_
1
0
dx

x
+
_

1
dx

x
= lim
0
+
_
1

dx

x
+ lim
b
_
b
1
dx

x
= lim
0
+
_
2

+ lim
b
_
2

b
1
= 2 + (limit does not exist).
Hence
_

0
dx

x
diverges.
Example 6.15. (Singularities at and )
_

x dx = lim
t
_
0
t
x dx + lim
s
_
s
0
x dx
= lim
t
_
x
2
2
_
0
t
+ lim
s
_
x
2
2
_
s
0
.
Both limits do not exist in R. Hence
_

x dx diverges.
Note: The Cauchy principal value of
_

x dx is
lim
s
_
s
s
x dx = lim
s
_
x
2
2
_
s
s
= lim
s
(0) = 0.
MA3110 MATHEMATICAL ANALYSIS II 181
7. Sequences and Series of Functions
7.1 Pointwise and uniform convergence
Throughout this section, let E be a non-empty set and
let f
n

n=1
be a sequence of real-valued functions de-
ned on E, E
f
n
R.
Denition. If f
n
(x)

n=1
converges in R for every x
in E, then we say that f
n
converges pointwise on
E to a limit function f, dened by
f(x) = lim
n
f
n
(x), x E.
Equivalently, f
n
converges to f pointwise on E if
> 0, x E, N(, x) N such that
[f
n
(x) f(x)[ < n N.
In this case, we also write f
n
f pointwise on E.
182 MA3110 MATHEMATICAL ANALYSIS II
Example 7.1. Let f
n
: [0, 1] R be dened by
f
n
(x) = x
n
.
Then
(i)
f(x) = lim
n
f
n
(x) =
_
_
_
0 if 0 x < 1,
1 if x = 1.
(ii) Each f
n
is continuous on [0,1], but f is not contin-
uous on [0, 1].
MA3110 MATHEMATICAL ANALYSIS II 183
Example 7.2. Let f
n
: [0, 1] R be dened by
f
n
(x) =
_

_
2n
2
x if 0 x
1
2n
2n 2n
2
x if
1
2n
x
1
n
0 if
1
n
x 1.
Then
(i) f(x) = lim
n
f
n
(x) = 0, x [0, 1].
(If x = 0, then f
n
(0) = 0 n N.
If 0 < x 1, then n >
1
x
,
1
n
< x,

. . f
n
(x) = 0.)
(ii) All f
n
and f are continuous on [0, 1].
(iii) lim
n
_
1
0
f
n
(x) dx ,=
_
1
0
f(x) dx, since
_
1
0
f
n
(x) dx =
1
2

1
n
n =
1
2
n N,
_
1
0
f(x) dx = 0.
184 MA3110 MATHEMATICAL ANALYSIS II
Example 7.3. For n N, let
f
n
(x) =
1

n
sin nx, x R.
Then
(i)
f(x) = lim
n
f
n
(x) = 0, x R.
(ii) f
/
n
(x) =

ncos nx, f
/
n
continuous and
f
/
= 0 continuous, but f
/
n
does not converge to f
/
.
E.g., f
/
n
(0) =

n as n , but f
/
(0) = 0.
MA3110 MATHEMATICAL ANALYSIS II 185
Pointwise convergence of f
n
to f does not necessarily
mean that f
n
approaches f in a global manner.
Denition. The sequence f
n
converges uniformly
to f on E if > 0, N() N such that x E,
[f
n
(x) f(x)[ < n N;
equivalently: > 0, N() N such that
sup
xE
[f
n
(x) f(x)[ < n N;
equivalently:
sup
xE
[f
n
(x) f(x)[ 0 as n .
In this case, we also write f
n
f uniformly on E.
(Geometrically, given any > 0, the graph of f
n
is
bounded by those of f and f + for all large n.)
186 MA3110 MATHEMATICAL ANALYSIS II
Remarks.
(1) If f
n
converges uniformly on E,
then f
n
converges pointwise on E.
The converse is not true (Example 7.1, Example 7.2).
(2) f
n
does not converge uniformly to f on E if and
only if > 0, N N, x E and n N
such that
[f
n
(x) f(x)[ ;
equivalently:
sup
xE
[f
n
(x) f(x)[ , 0 as n .
MA3110 MATHEMATICAL ANALYSIS II 187
In Example 7.1: (1) f
n
f pointwise on [0, 1].
(2) Let =
1
2
. n N, let x
n
= (
1
2
)
1
n
< 1. Then
[f
n
(x
n
) f(x
n
)[ = [x
n
n
0[ =
1
2
= .

. . f
n
, f uniformly on [0, 1].
(3) Let 0 < r < 1. Since r
n
0 as n , given
> 0, N N such that r
n
< n N. Hence
[f
n
(x) f(x)[ = [x
n
[ r
n
< , 0 x r n N.

. . f
n
f uniformly on [0, r] 0 < r < 1.
(4) Note that f
n
, f uniformly on [0, 1). (Ex)
n (0.1)
n
(0.9)
n
1 0.1 0.9
2 0.01 0.81
.
.
.
.
.
.
.
.
.
15 10
15
0.205
16 10
16
0.185
.
.
.
.
.
.
.
.
.
22 10
22
0.098
188 MA3110 MATHEMATICAL ANALYSIS II
(5) Let = 0.2. Then
[(0.1)
n
0[ < 0.2 n N(0.2, 0.1) = 1.
[(0.9)
n
0[ < 0.2 n N(0.2, 0.9) = 16.
Let 0 < < 1, 0 < x < 1 and N(, x) =
_
log
log x
_
+ 1
(an increasing function of x for each xed ).
If n N(, x), then n >
log
log x
.

. . nlog x < log (since log x < 0).

. .
[f
n
(x) f(x)[ = x
n
< .
Note that for each 0 < < 1, we have sup
0<x<1
N(, x) = .
Hence there is no positive integer N() that works for
all x.
MA3110 MATHEMATICAL ANALYSIS II 189
In Example 7.2: (1) f
n
f = 0 pointwise on [0, 1].
(2) f
n
, f uniformly on [0, 1], as
sup
x[0,1]
[f
n
(x) f(x)[ = sup
x[0,1]
f
n
(x) = n , 0.
(3) If 0 < r < 1, then f
n
0 uniformly on [r, 1]. (Ex).
190 MA3110 MATHEMATICAL ANALYSIS II
In Example 7.3: (1) f
n
f = 0 pointwise on R.
(2) f
n
0 uniformly on R, since
sup
xR
[f
n
(x)[ = sup
xR
1

n
[ sin nx[
1

n
0 as n .
(3) f
/
n
and f
/
continuous on R but lim
n
f
/
n
(x) ,= f
/
(x).
MA3110 MATHEMATICAL ANALYSIS II 191
Example 7.4. For n N, let
f
n
(x) = x
n
(1 x
n
), x [0, 1].
Then
(i) f(x) = lim
n
f
n
(x) = 0, x [0, 1]. (Ex)
(ii)
f
/
n
(x) = nx
n1
(1 2x
n
)
= 0 if x = 0, (
1
2
)
1
n
.
Thus f
n
attains absolute maximum at x
n
= (
1
2
)
1
n
.
Hence for every n N,
sup
x[0,1]
[f
n
(x) f(x)[ = f
n
_
(
1
2
)
1
n
_
=
1
4
, 0 as n .
Therefore f
n
does not converge uniformly to f.
192 MA3110 MATHEMATICAL ANALYSIS II
Theorem 7.1. (Cauchys criterion) Let f
n
be a se-
quence of functions dened on E.
(i) f
n
converges pointwise on E if and only if
(p)
_
_
_
x E, > 0, N(, x) N such that
[f
n
(x) f
m
(x)[ < m, n N.
(ii) f
n
converges uniformly on E if and only if
(u)
_
_
_
> 0, N() N such that x E,
[f
n
(x) f
m
(x)[ < m, n N.
Proof. (i) follows from Cauchys criterion for sequences
of constants applied to f
n
(x)

n=1
, for all x in E.
(ii) () Suppose f
n
f uniformly on E:
> 0, N() N such that x E,
[f
n
(x) f(x)[ <

2
n N.
Hence if m, n N, then
[f
n
(x) f
m
(x)[ [f
n
(x) f(x)[ + [f(x) f
m
(x)[
<

2
+

2
= .
Therefore condition (u) holds.
MA3110 MATHEMATICAL ANALYSIS II 193
() Conversely, suppose that
(u)
_
_
_
> 0, N() N such that x E,
[f
n
(x) f
m
(x)[ < m, n N.
This implies that condition (p) holds.
Therefore f
n
converges pointwise on E.
Let
f(x) = lim
n
f
n
(x) x E.
Letting m in (u), we have
> 0, N() N such that x E,
[f
n
(x) f(x)[ n N.
Hence f
n
f uniformly on E.

194 MA3110 MATHEMATICAL ANALYSIS II


Example 7.5. Let
f
n
(x) =
1
1 + x
2n
, x R, n N.
(1) Let r > 1. Then x r, n N,
0 f
n
(x)
1
1 + r
2n
0 as n .

. .
sup
x[r,)
f
n
(x) 0 as n .
Hence f
n
0 uniformly on [r, ) for every r > 1.
Similarly, f
n
0 uniformly on (, r] r < 1.
Hence f
n
0 pointwise on (, 1) (1, ).
(2) Let 0 < r < 1. Then x [r, r], n N,
[f
n
(x) 1[ =
x
2n
1 + x
2n
r
2n
0 as n .
Hence f
n
1 uniformly on [r, r] for every 0 < r < 1.
Therefore, f
n
1 pointwise on (1, 1).
MA3110 MATHEMATICAL ANALYSIS II 195
(3) Let
f(x) = lim
n
f
n
(x) =
_

_
1 if [x[ < 1
1
2
if [x[ = 1
0 if [x[ > 1.
For every n N, let x
n
= (
1
2
)
1
2n
. Then 0 < x
n
< 1 and
[f
n
(x
n
) f(x
n
)[ =
1
3
.
Hence f
n
, f uniformly on (1, 1).
Similarly, f
n
, f uniformly on (1, ) nor on (, 1).
(Ex)
196 MA3110 MATHEMATICAL ANALYSIS II
7.2 Properties preserved by uniform conver-
gence
Throughout this section, let f
n
: [a, b] R, n N.
Theorem 7.2. If f
n
converges uniformly to f on
[a, b] and each f
n
is continuous at x
0
[a, b] (resp.
continuous on [a, b]), then f is continuous at x
0
(resp. continuous on [a, b]).
Proof. Let f
n
f uniformly on [a, b]. Let > 0.
Then there exists n N such that for every x [a, b],
[f
n
(x) f(x)[ <

3
.
For this n, since f
n
is continuous at x
0
, there exists
> 0 such that for every x [a, b] with [x x
0
[ < ,
[f
n
(x) f
n
(x
0
)[ <

3
.
Hence if x [a, b] and [x x
0
[ < , then
[f(x) f(x
0
)[ [f(x) f
n
(x)[ + [f
n
(x) f
n
(x
0
)[
+ [f
n
(x
0
) f(x
0
)[
<

3
+

3
+

3
= .
Hence f is continuous at x
0
.
MA3110 MATHEMATICAL ANALYSIS II 197
Remarks.
(1) The conclusion means that
lim
xx
0
f(x) = f(x
0
)
lim
xx
0
lim
n
f
n
(x) = lim
n
f
n
(x
0
)
= lim
n
lim
xx
0
f
n
(x),
i.e., we can interchange the order of these two limit
operations in this case.
(2) Suppose f
n
f pointwise on [a, b] and each f
n
is
continuous on [a, b]. Then we have the implication:
f
n
f uniformly on [a, b]
= f continuous on [a, b].
(3) Pointwise limit of continuous functions may or may
not be continuous. In Example 7.1, each f
n
is con-
tinuous on [0, 1], f
n
f pointwise on [0, 1] but f
is not continuous on [0, 1]. (Hence by (2), f
n
, f
uniformly on [0, 1].)
(4) Converse of the implication = in (2) is not true
(see Example 7.2).
198 MA3110 MATHEMATICAL ANALYSIS II
Theorem 7.3. Suppose f
n
f uniformly on [a, b]
and each f
n
is integrable on [a, b]. Then f is in-
tegrable on [a, b] and for each x
0
[a, b], the se-
quence of functions F
n
(x) :=
_
x
x
0
f
n
(t) dt converges
uniformly to the function F(x) :=
_
x
x
0
f(t) dt on [a, b].
Hence
lim
n
_
x
x
0
f
n
(t) dt =
_
x
x
0
lim
n
f
n
(t) dt.
In particular,
lim
n
_
b
a
f
n
(t) dt =
_
b
a
f(t) dt.
Proof. (i) Let f
n
f uniformly on [a, b]:

n
= sup
x[a,b]
[f
n
(x) f(x)[ 0 as n .
Then
[f
n
(x) f(x)[
n
, x [a, b] n.

. . f
n
(x)
n
f(x) f
n
(x) +
n
, x [a, b] n.
MA3110 MATHEMATICAL ANALYSIS II 199
Since f
n

n
are integrable on [a, b],
_
b
a
(f
n

n
) dx =
_
b
a
(f
n

n
) dx

_
b
a
f dx

_
b
a
f dx

_
b
a
(f
n
+
n
) dx =
_
b
a
(f
n
+
n
) dx.
Hence
0
_
b
a
f dx
_
b
a
f dx
_
b
a
2
n
dx = 2
n
(b a).
Since
n
0 as n , we have
_
b
a
f dx =
_
b
a
f dx.
Therefore f is integrable on [a, b].
(ii) Let > 0. Then N N such that for every
x [a, b],
[f
n
(x) f(x)[ <

(b a)
if n N.
200 MA3110 MATHEMATICAL ANALYSIS II
For every x [a, b], if n N, we have
[F
n
(x) F(x)[ =

_
x
x
0
(f
n
(t) f(t)) dt

_
x
x
0
[f
n
(t) f(t)[ dt
(or
_
x
0
x
[f
n
(t) f(t)[ dt)

_
x
x
0

b a
dt
(or
_
x
0
x

b a
dt)
=

b a
[x x
0
[


b a
(b a) = .
Hence F
n
F uniformly on [a, b].

MA3110 MATHEMATICAL ANALYSIS II 201


Remarks.
(1) If f
n
f pointwise on [a, b], and all f
n
and f are
continuous on [a, b], it is not necessarily true that
then
lim
n
_
b
a
f
n
(t) dt =
_
b
a
f(t) dt.
E.g., see Example 7.2.
(2) Theorem 7.3 (and its proof) depends on [a, b] being
a bounded interval.
202 MA3110 MATHEMATICAL ANALYSIS II
Theorem 7.4. Suppose f
n
is a sequence of dif-
ferentiable functions on [a, b] such that
(i) f
n
(x
0
)

n=1
converges for some x
0
[a, b],
(ii) f
/
n
converges uniformly on [a, b], and
(iii) each f
/
n
is continuous on [a, b].
Then f
n
converges uniformly on [a, b] to a dier-
entiable function f, and
lim
n
f
/
n
(x) = f
/
(x), a x b,
i.e.,
lim
n
_
d
dx
f
n
(x)
_
=
d
dx
_
lim
n
f
n
(x)
_
.
Proof. Recall the Fundamental Theorem of Calculus
(Theorem 6.13): if g is dierentiable on [a, b] and g
/
is
continuous on [a, b], then
g(x) = g(x
0
) +
_
x
x
0
g
/
(t) dt, x, x
0
[a, b].
Since each f
/
n
is continuous on [a, b],
f
n
(x) = f
n
(x
0
) +
_
x
x
0
f
/
n
(t) dt, x [a, b].
Let f
/
n
G uniformly on [a, b] and f
n
(x
0
) L as
n .
MA3110 MATHEMATICAL ANALYSIS II 203
By Theorem 7.2, G is continuous on [a, b], and by The-
orem 7.3,
_
x
x
0
f
/
n
(t) dt
_
x
x
0
G(t) dt
uniformly on [a, b]. Hence f
n
converges uniformly on
[a, b] to a function f, dened by
f(x) = L +
_
x
x
0
G(t) dt.
Then by Theorem 6.12, f is dierentiable on [a, b] and
f
/
(x) =
d
dx
_
x
x
0
G(t) dt = G(x) = lim
n
f
/
n
(x),
for all x [a, b].

204 MA3110 MATHEMATICAL ANALYSIS II


Remarks.
(1) Condition (iii) in Theorem 7.4 can be omitted (see
Rudin, p.152, Theorem 7.17, or B & S, Theorem 8.2.3).
The proof is more dicult.
(2) f
n
f uniformly on [a, b] and all f
/
n
and f
/
are
continuous on [a, b]
,= f
/
n
f
/
.
(See Example 7.3.)
(3) Theorem 7.4 (including the stronger version in Re-
mark 1 above) needs the domain [a, b] to be bounded.
MA3110 MATHEMATICAL ANALYSIS II 205
7.3 Innite series of functions
Denition. If f
n

n=1
is a sequence of functions on
E, then

n=1
f
n
is an innite series of functions.
(1) For each n N, the nth partial sum of

n=1
f
n
is
the function S
n
=
n

i=1
f
i
dened on E.
(2) If the sequence S
n

n=1
of functions converges point-
wise (resp. uniformly) to a function S on a subset
E
0
of E, we say that the series

n=1
f
n
converges
pointwise (resp. uniformly) to the sum function S
on E
0
.
Remarks.
(1)

n=1
f
n
converges pointwise on E if and only if

n=1
f
n
(x) converges x E.
(2)

n=1
f
n
converges uniformly on E
=

n=1
f
n
converges pointwise on E.
The converse is false in general.
206 MA3110 MATHEMATICAL ANALYSIS II
Theorem 7.5. (Cauchys criterion for series of func-
tions)
(i)

n=1
f
n
converges pointwise on E if and only if
> 0, x E, N(, x) N such that

i=n
f
i
(x)

< m n N.
(ii)

n=1
f
n
converges uniformly on E if and only if
> 0, N() N such that x E,

i=n
f
i
(x)

< m n N;
equivalently: > 0, N() N such that
sup
xE

i=n
f
i
(x)

< m n N.
Proof. Apply Cauchys criterion for sequences of func-
tions (Theorem 7.1) to the sequence S
n
of partial
sums.
MA3110 MATHEMATICAL ANALYSIS II 207
Corollary 7.6. If

f
n
converges uniformly on E,
then
sup
xE
[f
n
(x)[ 0 as n ,
i.e., f
n
0 uniformly on E.
Proof. Let m = n in Theorem 7.5(ii).
Remark.
If f
n
does not converge to 0 uniformly on E,
then

f
n
does not converge uniformly on E.
208 MA3110 MATHEMATICAL ANALYSIS II
Theorem 7.7. (Weierstrass M-test) If M
n
is a
sequence of positive constants such that the series

M
n
converges and
[f
n
(x)[ M
n
x E n N,
then

f
n
converges uniformly on E (in fact,

[f
n
[
converges uniformly on E).
Proof. Let > 0. Since

M
n
converges, N() N
such that
m

i=n
M
i
< m n N.
Then x E and m n N,

i=n
f
i
(x)

i=n
[f
i
(x)[
m

i=n
M
i
< .
By Theorem 7.5(ii),

f
n
converges uniformly on E
(in fact, so does

[f
n
[).
MA3110 MATHEMATICAL ANALYSIS II 209
Example 7.6. Consider the series of functions

n=1
sin nx
n
2
, x R.
Since

sin nx
n
2


1
n
2
x R n N
and

1
n
2
converges, by Weierstrass M-test,

n=1
sin nx
n
2
converges uniformly on R.
210 MA3110 MATHEMATICAL ANALYSIS II
Example 7.7. Let

n=1
f
n
(x) =

n=1
_
x
1 + x
_
n
, x ,= 1.
(1) For every 0 < r < 1, the series

n=1
r
n
converges,
and n N,
[f
n
(x)[ r
n
if

x
1 + x

r.
By Weierstrass M-test,

f
n
converges uniformly
on
_
x R [

x
1 + x

r
_
=
_
r
1 + r
,
r
1 r
_
for every 0 < r < 1. Since as r 1

,
r
1 + r

_

1
2
_
+
and
r
1 r
,

f
n
converges uniformly on every closed bounded
subinterval of (
1
2
, ) and so converges pointwise
on (
1
2
, ).
MA3110 MATHEMATICAL ANALYSIS II 211
(2) However,
1
2
< b < , since f
n
is continuous,
sup
x(
1
2
,b)
[f
n
(x)[ = sup
x[
1
2
,b]
[f
n
(x)[
[f
n
(
1
2
)[ = 1.

. . f
n
, 0 uniformly on (
1
2
, b).

. .

f
n
does not converge uniformly on (
1
2
, b).
(3) For x
1
2
and x ,= 1, we have

x
1+x

1.

. .

f
n
(x) diverges.

. .

f
n
converges pointwise exactly on (
1
2
, ).
212 MA3110 MATHEMATICAL ANALYSIS II
Theorem 7.8. If

f
n
converges uniformly to f on
[a, b] and each f
n
is continuous at x
0
[a, b] (resp.
continuous on [a, b]), then f is continuous at x
0
(resp. continuous on [a, b]).
Proof. Since
S
n
=
n

n=1
f
i
f
uniformly on [a, b], and each S
n
is continuous at x
0
, by
Theorem 7.2, f is continuous at x
0
.
In Example 7.7: the series of functions

n=1
_
x
1 + x
_
n
converges uniformly on every closed bounded subinter-
val of (
1
2
, ) and each function
_
x
1 + x
_
n
is con-
tinuous on (
1
2
, ). By Theorem 7.8,

n=1
_
x
1 + x
_
n
is continuous on every closed bounded subinterval of
(
1
2
, ) and hence continuous on (
1
2
, ).
MA3110 MATHEMATICAL ANALYSIS II 213
Theorem 7.9. If

n=1
f
n
converges uniformly to f
on [a, b] and each f
n
is integrable on [a, b], then f is
integrable on [a, b] and for every x [a, b],

n=1
_
x
a
f
n
(t) dt =
_
x
a
f(t) dt =
_
x
a

n=1
f
n
(t) dt,
where the convergence is uniform on [a, b].
Proof. By the hypothesis,
S
n
=
n

i=1
f
i
f
uniformly on [a, b], and each S
n
is integrable on [a, b].
By Theorem 7.3, f is integrable on [a, b], and for every
x in [a, b],
lim
n
n

i=1
_
x
a
f
i
(t) dt = lim
n
_
x
a
S
n
(t) dt
=
_
x
a
f(t) dt,
i.e.,

i=1
_
x
a
f
i
(t) dt =
_
x
a
f(t) dt,
and the convergence is uniform on [a, b].
214 MA3110 MATHEMATICAL ANALYSIS II
Example 7.8. For every r > 0,

n=0
(1)
n
x
2n
n!
= e
x
2
,
where the convergence is uniform on [r, r]. (Ex)
By Theorem 7.9,
_
x
0
e
t
2
dt =

n=0
(1)
n
_
x
0
t
2n
n!
dt
=

n=0
(1)
n
n! (2n + 1)
x
2n+1
, x [r, r],
and the convergence is uniform on [r, r].
MA3110 MATHEMATICAL ANALYSIS II 215
Theorem 7.10. Suppose f
n
is a sequence of dif-
ferentiable functions on [a, b] such that
(i)

n=1
f
n
(x
0
) converges for some x
0
[a, b], and
(ii)

n=1
f
/
n
converges uniformly on [a, b].
Then

n=1
f
n
converges uniformly on [a, b] to a dif-
ferentiable function f, and

n=1
f
/
n
(x) = f
/
(x), a x b
i.e.,

n=1
d
dx
f
n
(x) =
d
dx
_

n=1
f
n
(x)
_
.
Proof. Apply Theorem 7.4 (and the Remark) to
S
n
=
n

i=1
f
i
.
(Ex)
Remark. Theorem 7.10 requires the domain [a, b] to
be a bounded interval (because Theorem 7.4 does).
216 MA3110 MATHEMATICAL ANALYSIS II
Example 7.9. The series

n=1
f
n
(x) =

n=1
(1)
n
1

n
cos
x
n
converges at x = 0 (since

n=1
(1)
n
1

n
converges), and
the series

n=1
f
/
n
(x) =

n=1
(1)
n+1
n
3/2
sin
x
n
converges uniformly on R (by Weierstrass M-test).
By Theorem 7.10,

f
n
converges uniformly on [r, r]
for every r > 0, and its derivative at x is

n=1
f
/
n
(x), x R.
Remark. We cannot apply Theorem 7.10 directly to
the entire real line R.
MA3110 MATHEMATICAL ANALYSIS II 217
7.4 More tests of uniform convergence
Denition. A sequence f
n
of functions dened on
a set E is uniformly bounded on E if there exists a
positive constant K such that
[f
n
(x)[ K, x E and n.
Theorem 7.11. (Dirichlets test) The series

a
n
b
n
of functions converges uniformly on E if
(i) the sequence of partial sums of

b
n
is
uniformly bounded on E,
(ii) a
n
0 uniformly on E, and
(iii) x E, the sequence a
n
(x)

n=1
is
monotone.
Corollary 7.12. (Dirichlets test for series of constant
terms) Given two sequences a
n
and b
n
of con-
stant terms, the series

a
n
b
n
converges if
(i) the sequence of partial sums of

b
n
is bounded,
(ii) a
n
0, and
(iii) a
n
is monotone.
218 MA3110 MATHEMATICAL ANALYSIS II
Proof. For every n N, let B
n
=

n
i=1
b
i
and S
n
=

n
i=1
a
i
b
i
be the partial sums of the series

b
i
and

a
i
b
i
respectively. Then we have the partial sum-
mation formula (Ex)
S
n
=
n

i=1
a
i
b
i
=
n1

i=1
(a
i
a
i+1
)B
i
+ a
n
B
n
.
By (i), there exists a positive constant K such that
[B
n
(x)[ K, x E n.
By (ii), given > 0, N() N such that
[a
n
(x)[ <

4K
, x E n N().
By (iii), for every x E, the sequence a
n
(x)

n=1
is
monotone. If a
n
(x)

n=1
is decreasing, then
m

i=n
[a
i+1
(x) a
i
(x)[ =
m

i=n
(a
i
(x) a
i+1
(x))
= a
n
(x) a
m+1
(x).
If a
n
(x)

n=1
is increasing, then
m

i=n
[a
i+1
(x) a
i
(x)[ =
m

i=n
(a
i+1
(x) a
i
(x))
= a
m+1
(x) a
n
(x).
MA3110 MATHEMATICAL ANALYSIS II 219
Therefore
m

i=n
[a
i+1
(x) a
i
(x)[ = [a
m+1
(x) a
n
(x)[, x E.
Hence x E and m n N(),
m

i=n
[a
i+1
(x) a
i
(x)[ [a
m+1
(x)[ + [a
n
(x)[ <

2K
.
Therefore by partial summation formula, x E and
m > n N(),
[S
m
(x) S
n
(x)[ = [
m1

i=n
(a
i
(x) a
i+1
(x))B
i
(x)
+ a
m
(x)B
m
(x) a
n
(x)B
n
(x)[

m1

i=n
[a
i
(x) a
i+1
(x)[[B
i
(x)[
+[a
m
(x)[[B
m
(x)[ + [a
n
(x)[[B
n
(x)[
<

2K
K +

4K
K +

4K
K = ,
By Cauchys criterion, S
n
converges uniformly on E.

220 MA3110 MATHEMATICAL ANALYSIS II


Corollary 7.13. (Alternating series test) Suppose that
a
n
is a sequence of functions on E,
(i) a
n
0 uniformly on E, and
(ii) x E, a
n
(x)

n=1
is monotone.
Then the series

(1)
n
a
n
converges uniformly on
E.
Proof. Let b
n
(x) = (1)
n
, x E. Apply Dirichlets
test. (Ex)
MA3110 MATHEMATICAL ANALYSIS II 221
Theorem 7.14. (Abels test) The series

a
n
b
n
of
functions converges uniformly on E if
(i)

b
n
converges uniformly on E,
(ii) a
n
is uniformly bounded on E, and
(iii) x E, a
n
(x)

n=1
is monotone.
Corollary 7.15. (Abels test for series of constant
terms) Given two sequences a
n
and b
n
of con-
stant terms, the series

a
n
b
n
converges if
(i)

b
n
converges,
(ii) a
n
is bounded, and
(iii) a
n
is monotone.
222 MA3110 MATHEMATICAL ANALYSIS II
Proof. For every n N, let B
n
=

n
i=1
b
i
B and
let
S
n
=
n

i=1
a
i
b
i
=
n1

i=1
(a
i
a
i+1
)B
i
+ a
n
B
n
=
n1

i=1
(a
i
a
i+1
)(B
i
B) + a
n
(B
n
B) + a
1
B.
Since a
n
is uniformly bounded on E, there exists a
positive constant K such that
[a
n
(x)[ K, x E n.
Since B
n
B uniformly on E, given > 0, there
exists N() N such that
[B
n
(x) B(x)[ <

4K
x E n N.
By (iii), as in the proof of Dirichlets Test, x E and
m n,
m

i=n
[a
i+1
(x) a
i
(x)[ = [a
m+1
(x) a
n
(x)[ 2K.
MA3110 MATHEMATICAL ANALYSIS II 223
Hence x E and m > n N,
[S
m
(x) S
n
(x)[ = [
m1

i=n
(a
i
(x) a
i+1
(x))(B
i
(x) B(x))
+ a
m
(x)(B
m
(x) B(x))
a
n
(x)(B
n
(x) B(x))[

m1

i=n
[a
i
(x) a
i+1
(x)[[B
i
(x) B(x)[
+[a
m
(x)[[B
m
(x) B(x)[
+[a
n
(x)[[B
n
(x) B(x)[
2K

4K
+ K

4K
+ K

4K
= ,
By Cauchys criterion, S
n
converges uniformly on E.

224 MA3110 MATHEMATICAL ANALYSIS II


Example 7.10. Consider the series

n=1
(1)
n
n + x
2
=

n1
(1)
n
a
n
(x),
where a
n
(x) =
1
n + x
2
.
(i) For every x R and for every n N,
[a
n
(x)[ =
1
n + x
2

1
n
0 as n .
Hence a
n
0 uniformly on R.
(ii) Obviously, a
n
(x)

n=1
is decreasing for every x
in R. By alternating series test,

(1)
n
a
n
converges
uniformly on R. This result cannot be obtained from
Weierstrass M-test, since

n=1
1
n + x
2
diverges for
every x R.
(iii) This example also shows that a series of functions
may converge uniformly on R without converging
absolutely for any x R.
MA3110 MATHEMATICAL ANALYSIS II 225
We have seen (Theorem 7.2) that if f
n
f pointwise
on [a, b] and each f
n
is continuous on [a, b], then
f
n
f uniformly on [a, b] = f continuous on [a, b].
The converse is not true in general (Example 7.2), but
it is true if f
n
is monotone.
Denition. The sequence f
n
is monotone if it is
increasing (f
n
(x) f
n+1
(x) n N x [a, b]), or
it is decreasing (f
n
(x) f
n+1
(x) n N x [a, b]).
Theorem 7.16. (Dinis Theorem) Suppose that
(i) f
n
f pointwise on [a, b],
(ii) all f
n
and f are continuous on [a, b], and
(iii) the sequence f
n
is monotone.
Then f
n
f uniformly on [a, b].
Proof. Without loss of generality, we may assume that
f
n
is decreasing. Let g
n
= f
n
f ( 0) n N.
Then g
n
0 pointwise on [a, b], each g
n
is continuous
on [a, b] and g
n
is decreasing. We want to show that
g
n
0 uniformly on [a, b].
226 MA3110 MATHEMATICAL ANALYSIS II
Since each g
n
is continuous on [a, b], by Extreme Value
Theorem there exists x
n
[a, b] such that
g
n
(x
n
) = sup
x[a,b]
g
n
(x).
Suppose g
n
(x
n
) , 0 as n . Then there ex-
ist > 0 and a subsequence x
n
k

k=1
of x
n
such
that g
n
k
(x
n
k
) k N (Ex). Since x
n
k

k=1
is
a bounded sequence in [a, b], by Bolzano-Weierstrass
Theorem, there exists a subsequence of x
n
k
converg-
ing to some c [a, b]. For simplicity of notation, in-
stead of considering a convergent subsequence of x
n
k
,
let x
n
k
c as k . Fix m N. Since g
n
is
decreasing, for every k N with m < n
k
,
g
m
(x
n
k
) g
n
k
(x
n
k
) .
Since g
m
is continuous on [a, b],
g
m
(c) = lim
k
g
m
(x
n
k
) m N.
In this case, g
m
(c) , 0 as m . This contradicts
that g
m
0 pointwise on [a, b]. Hence g
n
(x
n
) 0 as
n , and so g
n
0 uniformly on [a, b].

MA3110 MATHEMATICAL ANALYSIS II 227


Remark. It is crucial that the domain of the functions
is a closed and bounded interval [a, b] in Dinis theorem.
Example 7.11. For each n N, let f
n
: (0, 1) R
be dened by
f
n
(x) =
1
nx + 1
, 0 < x < 1.
Then f
n
0 pointwise on (0, 1), all f
n
are continuous
on (0, 1) and f
n
is decreasing. For each n N, by
continuity of f
n
,
sup
x(0,1)
f
n
(x) = sup
x[0,1]
f
n
(x) = f
n
(0) = 1.
Hence f
n
, 0 uniformly on (0, 1).
228 MA3110 MATHEMATICAL ANALYSIS II
Corollary 7.17. (Dinis Theorem for series of func-
tions) Suppose that
(i)

n=1
f
n
(x) = f(x) x [a, b],
(ii) all f
n
and f are continuous on [a, b], and
(iii) f
n
(x) 0 x [a, b] and n N.
Then

n=1
f
n
converges uniformly to f on [a, b].
Proof. Apply Theorem 7.16 to S
n
=
n

i=1
f
i
. (Ex)
MA3110 MATHEMATICAL ANALYSIS II 229
8. Power Series
8.1 Introduction to power series
A series of functions of the form

n=0
a
n
(x x
0
)
n
= a
0
+ a
1
(x x
0
) + a
2
(x x
0
)
2
+ + a
n
(x x
0
)
n
+ ,
where x
0
, a
1
, a
2
, . . . are constants, is a power series in
x x
0
.
If x
0
= 0, the power series becomes

n=0
a
n
x
n
= a
0
+ a
1
x + a
2
x
2
+ + a
n
x
n
+ .
Question: On what sets does a power series converge
(uniformly)?
Denition. A series

f
n
of functions converges
absolutely-uniformly on a set E if

[f
n
[ converges
uniformly on E.
(=

f
n
converges absolutely on E and

f
n
converges uniformly on E (Ex).)
230 MA3110 MATHEMATICAL ANALYSIS II
Theorem 8.1. Given a power series

n=0
a
n
(x x
0
)
n
,
let
R =
1
lim
n
[a
n
[
1
n
R
+
.
(If lim[a
n
[
1
n
= 0, dene R = ;
if lim[a
n
[
1
n
= , dene R = 0.)
(i) If 0 < R < , then the power series con-
verges absolutely-uniformly on [x
0
r, x
0
+r]
for every 0 < r < R (hence converges abso-
lutely on (x
0
R, x
0
+ R)), and diverges for
every x with [x x
0
[ > R.
(ii) If R = 0, then the series converges at x
0
only.
(iii) If R = , then the series converges absolutely-
uniformly on [x
0
r, x
0
+ r] for every r > 0
(hence converges absolutely on R).
The number R is called the radius of convergence of
the power series, and (x
0
R, x
0
+R) the interval of
convergence.
MA3110 MATHEMATICAL ANALYSIS II 231
Proof. If 0 < r < R, then
lim
n
([a
n
[r
n
)
1
n
= r lim
n
[a
n
[
1
n
< 1.
By root test,

[a
n
[r
n
converges.
If [x x
0
[ r, then [a
n
(x x
0
)
n
[ [a
n
[r
n
, n.
By Weierstrass M-test,

[a
n
(xx
0
)
n
[ converges uni-
formly on [x
0
r, x
0
+r]. This proves the rst part of
(i) and (iii).
If [x x
0
[ > R, then
lim
n
[a
n
(x x
0
)
n
[
1
n
= lim
n
[a
n
[
1
n
[x x
0
[ > 1.
By root test,

a
n
(x x
0
)
n
diverges. This proves the
2nd part of (i) and (ii).
232 MA3110 MATHEMATICAL ANALYSIS II
Remarks.
(1) Suppose

a
n
(x x
0
)
n
converges at some a.
Then

a
n
(x x
0
)
n
converges absolutely at every
x with [x x
0
[ < [a x
0
[ (since [a x
0
[ R).
Also

a
n
(x x
0
)
n
converges uniformly on
x R [ [x x
0
[ r for every r < [a x
0
[.
(2) If

a
n
(x x
0
)
n
diverges at some b, then

a
n
(x x
0
)
n
diverges at every x with
[x x
0
[ > [b x
0
[ (since [b x
0
[ R).
(3) The power series may converge or diverge at the end-
points x
0
R, x
0
+R of the interval of convergence
(x
0
R, x
0
+ R).
MA3110 MATHEMATICAL ANALYSIS II 233
Corollary 8.2. If a
n
,= 0 n, then the radius of
convergence of

a
n
(x x
0
)
n
is given by
1
lim
n
[
a
n+1
a
n
[
if this limit exists in R
+
.
Proof. This follows from the inequalities (Ex.)
lim [
a
n+1
a
n
[ lim [a
n
[
1
n
lim [a
n
[
1
n
lim [
a
n+1
a
n
[.

234 MA3110 MATHEMATICAL ANALYSIS II


Example 8.1. For

n=0
x
n
n!
,
lim
n

a
n+1
a
n

= lim
n
1
(n+1)!
1
n!
= 0.

. . R = .

. .

x
n
n!
converges x R and
absolutely-uniformly on [r, r] r > 0.
Example 8.2. For

n=0
n! x
n
,
lim
n

a
n+1
a
n

= lim
n
(n + 1)!
n!
= .

. . R = 0.

. . the series converges only at 0.
MA3110 MATHEMATICAL ANALYSIS II 235
Example 8.3. For

n=1
(1)
n+1
n
(x 1)
n
,
a
n
=
(1)
n+1
n
,
lim
n

a
n+1
a
n

= lim
n
1
n+1
1
n
= 1.

. . R = 1, the interval of convergence is (0, 2).


At x = 0, the series is

n=1
1
n
, divergent.
At x = 2, the series is

n=1
(1)
n+1
n
, convergent
(by alternating series test).

. . the series converges on (0, 2].


236 MA3110 MATHEMATICAL ANALYSIS II
Example 8.4. For

n=0
sin
n
4
x
n
=

n=0
a
n
x
n
,
lim
n

a
n+1
a
n

, lim
n
[a
n
[
1
n
= lim
n

sin
n
4

1
n
= 1.

. . R = 1, the series converges absolutely on (1, 1)


and absolutely-uniformly on [r, r] 0 < r < 1.
The series diverges at x = 1,
since sin
n
4
, 0 and (1)
n
sin
n
4
, 0.
MA3110 MATHEMATICAL ANALYSIS II 237
Example 8.5. For the power series

n=0
x
2n
2
n
,
we cant apply the formula
lim
n

a
n+1
a
n

to nd the radius of convergence R,


since a
2n+1
= 0, a
2n
=
1
2
n
, n. However,
lim
n
[a
n
[
1
n
= lim
n
[a
2n
[
1
2n
= lim
n
_
1
2
n
_ 1
2n
=
1

2
.
Hence R =

2.
[ OR : Letting y = x
2
, the series becomes

n=0
y
n
2
n
=

n=0
b
n
y
n
.
Then
lim
n

b
n+1
b
n

= lim
n
2
n
2
n+1
=
1
2
.
238 MA3110 MATHEMATICAL ANALYSIS II

. . radius of convergence for

y
n
2
n
is 2,
i.e.,

n=0
y
n
2
n
converges absolutely if [y[ < 2 and diverges
if [y[ > 2.

. .

n=0
x
2n
2
n
converges absolutely if [x[ <

2 and di-
verges if [x[ >

2. ]
At x =

2, the series is

n=0
1, which diverges.
MA3110 MATHEMATICAL ANALYSIS II 239
8.2 Properties of functions dened by power
series
A power series

n=0
a
n
(x x
0
)
n
denes a function
f(x) =

n=0
a
n
(x x
0
)
n
on the domain
_
x R [

n=0
a
n
(x x
0
)
n
converges
_
.
Theorem 8.3. If
f(x) =

n=0
a
n
(x x
0
)
n
with radius of convergence R > 0, then f is inn-
itely dierentiable on [x x
0
[ < R,
f
/
(x) =

n=1
na
n
(x x
0
)
n1
for [x x
0
[ < R,
and for every positive integer k and [x x
0
[ < R,
f
(k)
(x) =

n=k
n(n 1) (n k + 1)a
n
(x x
0
)
nk
,
where the radius of convergence of each of these de-
rived series is R.
240 MA3110 MATHEMATICAL ANALYSIS II
Proof. The two series

n=1
na
n
(x x
0
)
n1
,

n=1
na
n
(x x
0
)
n
,
have the same radii of convergence, and
lim[na
n
[
1
n
= (limn
1
n
) lim[a
n
[
1
n
= lim[a
n
[
1
n
.
Hence

n=1
na
n
(x x
0
)
n1
has radius of convergence R, and by Theorem 8.1, this
power series converges uniformly on [x
0
r, x
0
+r] for
every 0 < r < R. By dierentiation theorem for series
of functions,
f
/
(x) =

n=1
na
n
(x x
0
)
n1
on [x
0
r, x
0
+ r]
for every 0 < r < R and so on (x
0
R, x
0
+ R).
Use induction (or repeat this argument) to obtain the
results for f
(k)
(x), k 1.
MA3110 MATHEMATICAL ANALYSIS II 241
Example 8.6.

n=0
x
n
has radius of convergence 1, and
1
1 x
=

n=0
x
n
, [x[ < 1.
(Note that the domain of
1
1 x
is R1, much larger
than x R [ [x[ < 1.)
By Theorem 8.3,
1
(1 x)
2
=

n=1
nx
n1
if [x[ < 1,
2
(1 x)
3
=

n=2
n(n 1)x
n2
if [x[ < 1.
242 MA3110 MATHEMATICAL ANALYSIS II
Remark. Although a power series and its derived
series have the same radii of convergence, they may
converge on dierent sets.
Example 8.7.
f(x) =

n=1
x
n
n
2
converges on [1, 1]. (Since lim
n
n
2
(n + 1)
2
= 1, radius
of convergence is 1.)
By Theorem 8.3,
f
/
(x) =

n=1
x
n1
n
, [x[ < 1,
and this derived series converges on [1, 1).
MA3110 MATHEMATICAL ANALYSIS II 243
Corollary 8.4. If
f(x) =

n=0
a
n
(x x
0
)
n
, [x x
0
[ < r,
for some r > 0, then
a
k
=
f
(k)
(x
0
)
k!
k N.
Proof. Note that r R, the radius of convergence of

n=0
a
n
(x x
0
)
n
,
and f is the restriction of the sum function of the power
series to [x x
0
[ < r. Hence by Theorem 8.3,
f
(k)
(x) =

n=k
n(n 1) . . . (n k + 1)a
n
(x x
0
)
nk
for [x x
0
[ < r.

. . f
(k)
(x
0
) = k(k 1) . . . 1 a
k
= k! a
k
.
244 MA3110 MATHEMATICAL ANALYSIS II
Corollary 8.5. (Uniqueness of power series) If

n=0
a
n
(x x
0
)
n
=

n=0
b
n
(x x
0
)
n
on [x x
0
[ < r
for some r > 0, then a
n
= b
n
n N.
Proof. Let f be the sum function of these series on
[x x
0
[ < r. By Corollary 8.4,
a
k
=
f
(k)
(x
0
)
k!
= b
k
k N.

MA3110 MATHEMATICAL ANALYSIS II 245


Example 8.8. Consider the expansion
(1)
1
1 x
=

n=0
x
n
, [x[ < 1.
Replacing x in (1) by x,
(2)
1
1 + x
=

n=0
(1)
n
x
n
, [x[ < 1.
(the power series representation of
1
1 + x
about 0 on
[x[ < 1).
Replacing x in (1) by x
2
(respectively by x
2
),
(3)
1
1 x
2
=

n=0
x
2n
, [x[ < 1.
(4)
1
1 + x
2
=

n=0
(1)
n
x
2n
, [x[ < 1.
246 MA3110 MATHEMATICAL ANALYSIS II
Theorem 8.6. Let
f(x) =

n=0
a
n
(x x
0
)
n
, [x x
0
[ < r,
for some r > 0. Then f is integrable on every
[a, b] (x
0
r, x
0
+ r) and
_
b
a
f(x) dx =

n=0
_
b
a
a
n
(x x
0
)
n
dx.
If particular, if x
0
= 0, then
_
x
0
f(t) dt =

n=0
a
n
n + 1
x
n+1
, [x[ < r.
Proof. Use theorem on integration of series of functions,
since

a
n
(x x
0
)
n
converges uniformly on [a, b].
MA3110 MATHEMATICAL ANALYSIS II 247
Example 8.9. Power series representation of the func-
tion log(1 + x)
log(1 + x) = ?

1
1 + x
=

n=0
(1)
n
x
n
if [x[ < 1.
For 1 < x < 1, by Theorem 8.6, we have
log(1 + x) =
_
x
0
dt
1 + t
=

n=0
(1)
n
_
x
0
t
n
dt
=

n=0
(1)
n
x
n+1
n + 1
. ()
When x = 1, the series is

n=0
1
n + 1
, divergent.
When x = 1, the series is

n=0
(1)
n
n + 1
, convergent.
Question: Is () still valid for x = 1? (to be answered
later)
248 MA3110 MATHEMATICAL ANALYSIS II
log(1 + x) =

n=0
(1)
n
x
n+1
n + 1
, [x[ < 1.

. . log(1 x) =

n=0

x
n+1
n + 1
, [x[ < 1.

. . log
_
1 + x
1 x
_
=

n even
2
x
n+1
n + 1
= 2

n=0
x
2n+1
2n + 1
, [x[ < 1.
MA3110 MATHEMATICAL ANALYSIS II 249
Theorem 8.7. (Abel) Let
f(x) =

n=0
a
n
(x x
0
)
n
for [x x
0
[ < R, where R is the radius of conver-
gence of the power series.
(i) If

n=0
a
n
R
n
converges (i.e., if the power series
converges at x
0
+ R), then
lim
x(x
0
+R)

f(x) =

n=0
a
n
R
n
.
(ii) If

n=0
(1)
n
a
n
R
n
converges (i.e., if the power
series converges at x
0
R), then
lim
x(x
0
R)
+
f(x) =

n=0
(1)
n
a
n
R
n
.
250 MA3110 MATHEMATICAL ANALYSIS II
Proof. Consider rst the case x
0
= 0 and R = 1. Let
g(y) =

n=0
b
n
y
n
,
where the radius of convergence is 1. Suppose

n=0
b
n
converges. Then by Abels test,
g(y) =

n=0
b
n
y
n
converges uniformly on [0, 1]. (Ex.)
Since each b
n
y
n
is continuous on [0, 1], the sum function
g is continuous on [0, 1]. Therefore
lim
y1

g(y) = g(1) =

n=0
b
n
.
Let
f(x) =

n=0
a
n
(x x
0
)
n
, R = radius of convergence.
(i) Suppose

n=0
a
n
R
n
converges. Let
g(y) =

n=0
a
n
R
n
y
n
,
MA3110 MATHEMATICAL ANALYSIS II 251
radius of convergence = lim[a
n
R
n
[
1
n
= 1.
By above,
lim
y1

g(y) =

n=0
a
n
R
n
.
Since
f(x) = g
_
x x
0
R
_
, [x x
0
[ < R,
lim
x(x
0
+R)

f(x) = lim
x(x
0
+R)

g
_
x x
0
R
_
= lim
y1

g(y)
=

n=0
a
n
R
n
.
(ii) Suppose

n=0
(1)
n
a
n
R
n
converges. Let
g(y) =

n=0
(1)
n
a
n
R
n
y
n
,
radius of convergence = 1.
By above,
lim
y1

g(y) =

n=0
(1)
n
a
n
R
n
.
252 MA3110 MATHEMATICAL ANALYSIS II
Since
f(x) = g
_
x
0
x
R
_
, [x x
0
[ < R,
lim
x(x
0
R)
+
f(x) = lim
x(x
0
R)
+
g
_
x
0
x
R
_
= lim
y1

g(y)
=

n=0
(1)
n
a
n
R
n
.

MA3110 MATHEMATICAL ANALYSIS II 253


Remarks.
(1) If R < radius of convergence of

n=0
a
n
(x x
0
)
n
,
the statement in Abels Theorem is obvious since by
Theorem 8.3, f is continuous on [x
0
R, x
0
+R] in
this case.
(2) Converse of Abels Theorem is false, i.e.,
lim
x(x
0
+R)

f(x)
,=

n=0
a
n
(x x
0
)
n
converges at x
0
+ R.
Example 8.10. Consider
f(x) =
1
1 + x
=

n=0
(1)
n
x
n
, 1 < x < 1,
radius of convergence is 1.
lim
x1

f(x) =
1
2
but

n=0
(1)
n
x
n
divergent at x = 1.
254 MA3110 MATHEMATICAL ANALYSIS II
An application: In Example 8.9, we have
() log(1 + x) =

n=0
(1)
n
x
n+1
n + 1
, 1 < x < 1,
where radius of convergence = 1.
The power series converges at x = 1. By Abels Theo-
rem and continuity of the function log at 2,
log 2 = lim
x1

log(1 + x) =

n=0
(1)
n
n + 1

. . () is valid for 1 < x 1.


MA3110 MATHEMATICAL ANALYSIS II 255
8.3 Taylor series and Maclaurin series
Let f be innitely dierentiable on (x
0
r, x
0
+r), i.e.,
f
(n)
(x) exists x (x
0
r, x
0
+ r) and n 0.
The power series

n=0
f
(n)
(x
0
)
n!
(x x
0
)
n
is called the Taylor series of f about x
0
.
If x
0
= 0, the series becomes

n=0
f
(n)
(0)
n!
x
n
and is called the Maclaurin series of f.
Questions:
1. Find the radius of convergence of the Taylor series
of f.
2. Is f represented by its Taylor series? i.e., is
f(x) =

n=0
f
(n)
(x
0
)
n!
(x x
0
)
n
on (x
0
r, x
0
+ r),
or, is
T
n
(x) =
n

k=0
f
(k)
(x
0
)
k!
(x x
0
)
k
f(x)
pointwise (respectively uniformly) on (x
0
r, x
0
+r)?
NO in general!
256 MA3110 MATHEMATICAL ANALYSIS II
Example 8.11. Let
f(x) =
_
_
_
e
1/x
2
if x ,= 0
0 if x = 0.
The function f is innitely dierentiable on R.
Proof: By induction and denition,
f
(n)
(x) = e
1/x
2
P
n
_
1
x
_
for x ,= 0,
where P
n
_
1
x
_
is a polynomial in
1
x
, and
f
(n)
(0) = 0 n 1
(apply L Hospital rule to get lim
x0
e
1/x
2
x
n
= 0). (Ex)
Hence the Maclaurin series of f is

n=0
f
(n)
(0)
n!
x
n
0.
Therefore f(x) equals

f
(n)
(0)
n!
x
n
only at x = 0.
MA3110 MATHEMATICAL ANALYSIS II 257
Theorem 8.8. If f has a power series representa-
tion
f(x) =

n=0
a
n
(x x
0
)
n
on (x
0
r, x
0
+ r),
then

n=0
a
n
(x x
0
)
n
is the Taylor series of f about x
0
.
Proof. By Corollary 8.4,
a
n
=
f
(n)
(x
0
)
n!
n.

. .

n=0
a
n
(x x
0
)
n
=

n=0
f
(n)
(x
0
)
n!
(x x
0
)
n
.
258 MA3110 MATHEMATICAL ANALYSIS II
Example 8.12. By Example 8.9,
log(1 + x) =

n=0
(1)
n
n + 1
x
n+1
, 1 < x < 1.
To nd the Taylor series of f(x) = log x about x
0
> 0,
i.e.,

n=0
f
(n)
(x
0
)
n!
(x x
0
)
n
,
f(x) = log x = log
_
x
0
_
1 +
x x
0
x
0
__
= log x
0
+ log
_
1 +
x x
0
x
0
_
= log x
0
+

n=0
(1)
n
n + 1
_
x x
0
x
0
_
n+1
if 1 <
x x
0
x
0
< 1
= log x
0
+

n=0
(1)
n
(n + 1)x
n+1
0
(x x
0
)
n+1
if 0 < x < 2x
0
This is a power series in (x x
0
) representing f on
(0, 2x
0
). Hence it is the Taylor series of f about x
0
.
MA3110 MATHEMATICAL ANALYSIS II 259
Recall
Taylor formula with remainder (cf. Theorem 5.16)
Suppose f
(n+1)
exists on I = (x
0
r, x
0
+ r). Then
for every x I, there exists c
n
(depending on n, x, x
0
)
between x and x
0
such that
f(x) =
n

k=0
f
(k)
(x
0
)
k!
(x x
0
)
k
. .
T
n
(x)
+
f
(n+1)
(c
n
)
(n + 1)!
(x x
0
)
n+1
. .
R
n
(x)
.
Theorem 8.9. Let f be innitely dierentiable on
I = (x
0
r, x
0
+ r), x I. Then
f(x) =

n=0
f
(n)
(x
0
)
n!
(x x
0
)
n
if and only if
lim
n
R
n
(x) = lim
n
f
(n+1)
(c
n
)
(n + 1)!
(x x
0
)
n+1
= 0,
where each c
n
(depending on n, x, x
0
) is between x
and x
0
.
260 MA3110 MATHEMATICAL ANALYSIS II
Example 8.13. The Maclaurin series of f(x) = e
x
is

n=0
f
(n)
(0)
n!
x
n
=

n=0
x
n
n!
= 1 + x +
x
2
2!
+ .
If x = 0, f(0) = e
0
= 1.
For x ,= 0, f(x) = T
n
(x) + R
n
(x), where
R
n
(x) =
f
(n+1)
(c
n
)
(n + 1)!
x
n+1
=
e
c
n
(n + 1)!
x
n+1
(c
n
between 0 and x). Since

n=0
r
n
n!
converges for all
r > 0 (by ratio test), we have lim
n
r
n
n!
= 0.

. . [R
n
(x)[ =
e
c
n
(n + 1)!
[x[
n+1
e
[x[
[x[
n+1
(n + 1)!
0
as n .

. . e
x
= f(x) =

n=0
x
n
n!
x R.
(We can modify this proof to get T
n
f uniformly on
[r, r] for every r > 0.)
MA3110 MATHEMATICAL ANALYSIS II 261
Example 8.14. The Taylor series of f(x) = sin x
about a is

n=0
sin
(n)
(a)
n!
(x a)
n
.
Let r > 0. For every x with [x a[ r,
sin x = T
n
(x) +
sin
(n+1)
(c
n
)
(n + 1)!
(x a)
n+1
,
where c
n
(depending on n, x, a) is between a and x.
Since [ sin
(n+1)
(c
n
)[ 1 n,
if [x a[ r, then
[ sin xT
n
(x)[
[x a[
n+1
(n + 1)!

r
n+1
(n + 1)!
0 as n .

. . sin x =

n=0
sin
(n)
(a)
n!
(x a)
n
x R,
and the convergence is uniform on [ar, a+r] r > 0.
In particular, taking a = 0, we have
sin x =

n=0
(1)
n
x
2n+1
(2n + 1)!
x R
= x
x
3
3!
+
x
5
5!

x
7
7!
+ ,
and the convergence is uniform on [r, r] r > 0.
262 MA3110 MATHEMATICAL ANALYSIS II
Example 8.15.
cos x =

n=0
(1)
n
x
2n
(2n)!
x R
= 1
x
2
2!
+
x
4
4!

x
6
6!
+ ,
and the convergence is uniform on [r, r] r > 0.
MA3110 MATHEMATICAL ANALYSIS II 263
8.4 Arithmetic operations with power series
Product of two series of constant terms
Denition. The Cauchy product of

n=0
a
n
and

n=0
b
n
is

n=0
c
n
, where
c
n
=
n

k=0
a
k
b
nk
=
n+1 terms
..
a
0
b
n
+ a
1
b
n1
+ + a
n
b
0
(i.e., c
n
is the sum of all products a
i
b
j
, where i, j 0
and i + j = n).
Motivation: Multiplying formally,
_

n=0
a
n
x
n
_

n=0
b
n
x
n
_
= (a
0
+ a
1
x + a
2
x
2
+ )(b
0
+ b
1
x + b
2
x
2
+ )
= a
0
b
0
+ (a
1
b
0
+ a
0
b
1
)x + (a
2
b
0
+ a
1
b
1
+ a
0
b
2
)x
2
+
= c
0
+ c
1
x + c
2
x
2
+
Letting x = 1, we arrive at the above denition.
Questions:
(1) When will the series

c
n
converge?
(2) If

c
n
converges, what is its value?
264 MA3110 MATHEMATICAL ANALYSIS II
Theorem 8.10. (Mertens) If

n=0
a
n
= A,

n=0
b
n
= B,
and if

a
n
or

b
n
converges absolutely, then the
Cauchy product

c
n
converges to AB.
Proof. Suppose

a
n
converges absolutely, and let
=

n=0
[a
n
[ < .
Let
A
n
=
n

i=0
a
i
, B
n
=
n

i=0
b
i
, C
n
=
n

i=0
c
i
,

n
= B
n
B( 0 as n ).
Then
C
n
= a
0
b
0
+ (a
0
b
1
+ a
1
b
0
) +
+(a
0
b
n
+ a
1
b
n1
+ + a
n
b
0
)
= a
0
B
n
+ a
1
B
n1
+ + a
n
B
0
= a
0
(B +
n
) + a
1
(B +
n1
) + + a
n
(B +
0
)
= A
n
B +
n
,
MA3110 MATHEMATICAL ANALYSIS II 265
where

n
=
n

i=0
a
i

ni
.
Claim:
n
0 as n .
Since
n
0, there exists M > 0 such that
[
n
[ M n.
Given > 0, there exists N such that
[
n
[ <

2 + 1
n N
and
n

i=m
[a
i
[ <

2M
if n m N.
(Cauchys criterion)
For n 2N,
[
n
[ [a
0

n
+ + a
nN1

N+1
[
+[a
nN

N
+ + a
n

0
[
< ([a
0
[ + + [a
nN1
[)

2 + 1
+
n

i=nN
[a
i
[M
<

2 + 1
+

2M
M .
Hence
n
0. Since A
n
A, we have C
n
AB.

266 MA3110 MATHEMATICAL ANALYSIS II


Remarks
(1) Under the hypotheses of Theorem 8.10,

[c
n
[ may
or may not converge.
(2) The Cauchy product may or may not converge if
both

a
n
and

b
n
converge conditionally.
Corollary 8.11. If both

a
n
and

b
n
converge
absolutely to A and B respectively, then the Cauchy
product

c
n
converges absolutely to AB.
Proof. (Exercise)
Example 8.16. Let [r[ < 1. The series

n=0
r
n
con-
verges absolutely to
1
1r
. Consider the Cauchy product
of

r
n
with itself: a
n
= b
n
= r
n
,
c
n
=
n

k=0
a
k
b
nk
=
n

k=0
r
k
r
nk
= (n + 1)r
n
.
By Corollary 8.11, the Cauchy product

n=0
(n+1)r
n
converges absolutely to
1
(1 r)
2
:

n=0
(n + 1)r
n
=
1
(1 r)
2
, [r[ < 1.
MA3110 MATHEMATICAL ANALYSIS II 267
Operations on power series
Theorem 8.12. Let
f(x) =

n=0
a
n
(x x
0
)
n
, [x x
0
[ < R
1
,
g(x) =

n=0
b
n
(x x
0
)
n
, [x x
0
[ < R
2
,
and , are constants. Then
(i) f(x) + g(x) =

n=0
(a
n
+ b
n
)(x x
0
)
n
for [x x
0
[ < minR
1
, R
2
;
(ii) f(x)g(x) =

n=0
c
n
(xx
0
)
n
, [xx
0
[ < minR
1
, R
2
,
where c
n
=
n

r=0
a
r
b
nr
(the nth term of the Cauchy
product of

n=0
a
n
and

n=0
b
n
).
268 MA3110 MATHEMATICAL ANALYSIS II
Proof. (i) is obvious.
(ii) Let R
0
= minR
1
, R
2
. Since

n=0
a
n
(x x
0
)
n
and

n=0
b
n
(x x
0
)
n
converge absolutely to f(x) and g(x) respectively if
[xx
0
[ < R
0
, by Corollary of Mertens Theorem, their
Cauchy product converges absolutely to f(x) g(x) if
[x x
0
[ < R
0
. The nth term of this Cauchy product
is
n

r=0
a
r
(x x
0
)
r
b
nr
(x x
0
)
nr
=
n

r=0
a
r
b
nr
. .
(x x
0
)
n
= c
n
(x x
0
)
n
.

. . f(x)g(x) =

n=0
c
n
(x x
0
)
n
if [x x
0
[ < R
0
.
MA3110 MATHEMATICAL ANALYSIS II 269
Example 8.17. Let
f(x) =
1
1 x
=

n=0
x
n
, [x[ < 1,
g(x) =

n=0
b
n
x
n
, [x[ < R,
f(x)g(x) =
g(x)
1 x
=

n=0
c
n
x
n
, [x[ < min1, R,
where
c
n
=
n

r=0
1 b
r
=
n

r=0
b
r
(nth partial sum of

n=0
b
n
).
Let
h(x) =

n=0
c
n
(x x
0
)
n
, [x x
0
[ < R
1
,
g(x) =

n=0
b
n
(x x
0
)
n
, [x x
0
[ < R
2
,
f(x) =
h(x)
g(x)
, and b
0
= g(x
0
) ,= 0.
270 MA3110 MATHEMATICAL ANALYSIS II
By continuity of g, g(x) ,= 0 x in an open interval I
containing x
0
and so f is innitely dierentiable on I.
Fact: f can be represented by its Taylor series about
x
0
(we omit the proof). Let
f(x) =

n=0
a
n
(x x
0
)
n
, [x x
0
[ < R.
Since f(x)g(x) = h(x), by Theorem 8.12,

n=0
_
n

r=0
a
r
b
nr
_
(x x
0
)
n
=

n=0
c
n
(x x
0
)
n
for [x x
0
[ < minR, R
1
, R
2
.

. .
n

r=0
a
r
b
nr
= c
n
n 0.
Solving recursively, a
0
=
c
0
b
0
and for n 1,
a
n
=
1
b
0
_
c
n

n1

r=0
a
r
b
nr
_
.
MA3110 MATHEMATICAL ANALYSIS II 271
Example 8.18. Find the Maclaurin series of tan x.
Let
tan x =

n=0
a
n
x
n
, x (

2
,

2
).
Since tan x is odd: tan(x) = tan x, x (

2
,

2
),

n=0
a
n
(1)
n
x
n
=

n=0
a
n
x
n
, x (

2
,

2
).

. . a
n
= 0 if n is even.

. . tan x =

n=0
a
2n+1
x
2n+1
= a
1
x + a
3
x
3
+ .
Since tan x cos x = sin x, by Examples 8.14 and 8.15,
_

n=0
a
2n+1
x
2n+1
_

n=0
(1)
n
(2n)!
x
2n
_
=

n=0
(1)
n
(2n + 1)!
x
2n+1
,
(a
1
x + a
3
x
3
+ a
5
x
5
+ )(1
x
2
2!
+
x
4
4!
)
= x
x
3
3!
+
x
5
5!
,
272 MA3110 MATHEMATICAL ANALYSIS II
a
1
x +
_
a
3

a
1
2
_
x
3
+
_
a
5

a
3
2
+
a
1
4!
_
x
5
+
= x
1
3!
x
3
+
1
5!
x
5
.

. . a
1
= 1, a
3

a
1
2
=
1
6
, a
5

a
3
2
+
a
1
4!
=
1
5!
,
a
3
=
1
3
, a
5
=
2
15
,

. . tan x = x +
x
3
3
+
2
15
x
5
+ , x (

2
,

2
).

You might also like