BNU-HKBU United International College

Division of Science and Technology

MATH4090 Real Analysis

Instructor: Dr. Chak-On Chow

Office: E407B-R7, phone no.: 3620645, email: [email protected]
Times and venues: Mon 10:0012:50, D408.
Assessment: Homework assignments (20%) midterm (20%); class participation (5%) and
5 half-hour quizzes (5%); final exam (50%). Homework will be assigned during the 2nd,
5th, 8th and 11th week, 2-hour midterm arranged during the 9th week (tentative), and
final exam during the final examination period.
References:
1. W. Rudin, Principles of Mathematical Analysis, 3rd edition, McGraw-Hill, c1976.
2. W.R. Parzynski, P.W. Zipse, Introduction to Mathematical Analysis, McGraw-Hill,
c1987.
Teaching Assistant: Ms. Jingya Zhao
Office: E409-H10, phone no.: 3620631, email: [email protected]
Course contents.
1. Sets, relations and mappings.
2. The real and complex number systems.
3. Basic topology.
4. Sequences and series.
5. Continuity.
6. Differentiation.
7. The Riemann-Stieltjes integral.
8. Sequences and series of functions.
9. (If time permits) Review of Lebesgue measure and integration.

Sets, relations and mappings

Sets
Definition. A set is a well-defined collection of objects. Objects that constitute the set
are called elements of the set. One writes x A to mean x belongs to A, and writes
x 6 A to mean x does not belong to A.
Example 1. Let A = {1, 3, 5, 7, 9}. Then 3 A but 4 6 A.
Sets can be represented in two ways:
(1) Tabulation form: By listing the elements of the set.
e.g., The set of all prime numbers < 20 is {2, 3, 5, 7, 11, 13, 17, 19}.
(2) Set-builder form: By specifying the properties of elements of the set.
e.g., The set all prime numbers < 20 is {x : x is prime and x < 20}.
Example 2. Let A = {1, 2, 3} and Y = {x : x = 100a 10b + c, with a, b, c A}.
Express the set Y in tabulation form. How many elements does Y have?
Solution. It is easy to see that
Y = {91, 92, 93, 81, 82, 83, 71, 72, 73, 191, 192, 193, 181, 182, 183,
171, 172, 173, 291, 292, 293, 281, 282, 283, 271, 272, 273}.

Since a, b, c each has 3 possible values, so Y has 33 = 27 elements.

Venn diagrams show all possible logical relations between a finite collection of sets.
Definition. If any element x of a set A is also an element of another set B, then the
set A is called a subset of the set B, denoted A B or B A. If A B and A 6= B,
then A is called a proper subset of B, denoted A $ B.
When A B, then B contains A, or A is contained in B. The Venn diagrams representing
A B and A 6 B are as follow:
B

AB

A 6 B

Example 3. Obviously {2, 3, 4} $ {1, 2, 3, 4, 5} (the inclusion is proper).

Example 4. Let N = {natural numbers}, Q = {rational numbers}, R = {real numbers}.

The following inclusions hold:
N $ Q $ R.
n
N Q because for any n N, n = 1 Q N Q; N 6= Q because 12 Q but 12 6 N.

Q R because any fraction is a real number; Q 6= R because 2 R but 2 6 Q.

Definition. Two sets A and B are equal if and only if A B and B A, denoted
A = B. (It means for all x, x A x B).
Let A = {a, a, b, a, b} and B = {a, b}, with a 6= b. Obviously, every element of A is an
element of B, and vice versa, i.e., A = B. So, distinct elements of a set are counted only
once.
Definitions.
(a) A set having only one element is called a singleton.
(b) Let a 6= b. The set {a, b} is an unordered pair of a and b.
(c) A set having no elements is called an empty set (void set,null set), denoted .
Elements of a set are unordered, i.e., {a, b} = {b, a}.
Proposition.
(a) Empty set is a subset of any set.
(b) The empty set is unique.
Proof. (a) Suppose not. Let A be a set such that 6 A. Then there exists x such
that x 6 A, which is a contradiction.
(b) Let be another set having no elements. Then and together imply
= , thus proving the uniqueness.

Definition. The set that includes all subsets of a set A is called the power set of A:
A

It is also denoted 2 .

P(A) = {X : X A}.

Example 5. When A = , {a} and {a, b}, find P(A).

Solution. We have

P() = {},

P({a}) = {, {a}},

P({a, b}) = {, {a}, {b}, {a, b}}.

Observe from example 7, when A has 0, 1, 2 elements, respectively, P(A) has 1 = 20 , 2 =

21 , 4 = 22 elements. One can deduce that when A has n elements, P(A) have 2n elements.
This is why P(A) is called the power set of A.
Exercise. Let A be a set having n elements. Prove by mathematical induction that A
has 2n subsets.

Definition. Let A, B be sets.

(a) The union of A and B is A B = {x : x A or x B}.
(b) The intersection of A and B is A B = {x : x A and x B}.
When A B = , A and B are disjoint.
(c) The complement of A is Ac = {x : x 6 A}, also denoted A .
(d) The relative complement B in A is A \ B = {x : x A and x 6 B} = A B c .
The Venn diagrams of A B, A B and A \ B are as follow: When A B 6= , then

AB

AB

A\B

AB

AB

A\B

When A B = , then

Proposition. Let A and B be sets. Then

(a) A B A and A B B;
(b) A A B and B A B.
Proof. It is clear from the above Venn diagrams.

Example 6. Let A = {1, 2, 3, 4}, B = {3, 4} and C = {5}. Then

A C = {1, 2, 3, 4, 5},

A B = {3, 4},

A C = .

Example 7. Let A be the set of all even numbers, B the set of all multiples of 3, and
C the set of all odd numbers. Prove that
(a) A B = {6n : n Z},
(b) B C = {3k : k Z is odd}.
Solution. The sets A, B, C can be written as
A = {2k : k Z},

B = {3n : n Z},

C = {2m + 1 : m Z}.

(a) For any n Z, 6n = 2(3n) A and 6n = 3(2n) B so that {6n : n Z} A B.

To prove the reverse inclusion, let x A B. Then x = 2k = 3n for some k, n Z.
Since k = 3n
is an integer and 2 3, necessarily 2|n so that n = 2m for some m Z,
2
whence x = 3(2m) = 6m. Thus, A B {6n : n Z}. Combining both inclusions,
A B = {6n : n Z} follows.
(b) follows from the fact that odd odd is odd, and odd even is even.


Example 8. Let A = {1, 2, 3, 4, 5}, B = {2, 4, 6, 8, 10} and C = {1, 3, 5}, then
A \ B = {1, 3, 5}, B \ A = {6, 8, 10},
A \ C = {2, 4},

C \ A = .

Theorem.
(i) (Idempotent laws) A A = A, A A = A.
(ii) (Commutative laws) A B = B A, A B = B A.
(iii) (Associative laws) (A B) C = A (B C), (A B) C = A (B C).
(iv) (Distributive laws) A(B C) = (AB)(AC), A(B C) = (AB)(AC).
(v) (Identity laws) A = A, A = .
(vi) (Complement laws) A Ac = , (Ac )c = A.
(vii) (De Morgans laws) (A B)c = Ac B c , (A B)c = Ac B c .
(viii) (Absorption laws) A B A B = B, A B A B = A.
(ix) (Transitive law) A B and B C A C.
Proof. (iii)

x, x (A B) C x A B or x C

(x A or x B) or x C

x A or (x B or x C)
x A or x B C
x A (B C)

(viii) Obviously A B B. Because A B, so A B B B = B.

Suppose on the contrary that A 6 B. Then x A such that x 6 B. Thus,
A B {x} B % B, which contradicts A B = B.
The proofs of the remaining assertions are left as exercises.

Example 9. Prove that
(a) A \ B = A B,
(b) When A \ B = and B \ A = , so A = B.

Solution. (a) Suppose on the contrary that A 6 B. Then x A and x 6 B so that

x A B c = A \ B = , which is a contradiction.
: If A \ B 6= . Then x A and x 6 B so that A 6 B.
(b) By (a), one obtains
A \ B = and B \ A = A B and B A A = B.

Definition. The symmetric difference of the two sets A and B is defined as

AB = (A \ B) (B \ A).

The Venn diagrams depicting AB in the non-disjoint and disjoint cases are as follow:

A B 6=

AB =

Example 10. Let A = {1, 2, 3, 4} and B = {3, 4, 5}, then

A \ B = {1, 2},

B \ A = {5},

AB = {1, 2} {5} = {1, 2, 5}.

Example 11. Let A and B be sets. Prove that AB = (A B) \ (A B).

Solution. By definition of AB, the distributive laws and the De Morgans law, we have
AB = (A \ B) (B \ A)
= (A B c ) (B Ac )

= (A B) (A Ac ) (B c B) (B c Ac )

= (A B) (B c Ac )
= (A B) (A B)c
= (A B) \ (A B).

Theorem. Let A and B be two sets. Then

(a) (Commutative law) AB = BA,
(b) (Associative law) (AB)C = A(BC).
Proof. Exercise.

Definition. A set having finitely many elements is called a finite set; otherwise, it is
called an infinite set. Denote by n(S) the number of elements of a finite set S, also called
the cardinality of S. Other commonly used notations denoting the number of elements of
a finite set include |A|, #A, card A.
Theorem. (Principle of inclusion-exclusion) For any finite sets A and B,
n(A B) = n(A) + n(B) n(A B).

Theorem. Let A1 , A2 , . . . , Am be finite sets. Then

[
 X
m
m
X
n
Ai =
n(Ai )
n(Ai Aj ) +
i=1

i=1

16i<j6m

+ + (1)

m1

16i<j<k6m

n(Ai Aj Ak )

n(A1 A2 Am ).

Example 12. In a class of 40 students, every student need to take Chemistry or Mathematics. If there are 35 students taking Chemistry and 20 students taking Mathematics,
how many of them are taking Chemistry and Mathematics at the same time?
Solution. Let
C = {Students taking Chemistry},

M = {Students taking Mathematics}.

then n(C) = 35, n(M ) = 20 and n(C M ) = 40. So, there are

n(C M ) = n(C) + n(M ) n(C M )

= 35 + 20 40 = 15

students taking Chemistry and Mathematics at the same time.

Exercise. Let A = {{}, }. Which of the following are true?

(a) {{}} A,
(b) A,
(c) {} A,
(d) {{}} A,
(e) A,
(f) {} A
Exercise. Consider the following sets of geometric figures in the xy-plane:
A = {quadrilaterals}, B = {parallelograms}, C = {rhombuses},
D = {rectangles}, E = {squares}.

(a) Determine which set is a subset of another set.

(b) Find the union and intersection of each set.
Exercise. Find the power set of the following sets:
(a) {a, b, c},
(b) {a, {b, c}}.

Exercise. Let A, B, C be sets. Prove that

(a) If A B and A 6 C, then B 6 C,
(b) If A B = A and A C 6= , then B C 6= .
Exercise. Let A, B E. Prove that
(a) B \ A E \ A,
(b) If A B = , then B (E \ A) = B and A (E \ B) = E \ B,
(c) (E \ A) \ (E \ B) = B \ A,
(d) If A B = E and A B = , then B = E \ A

Exercise. Let E be a set and A, B E. Find the condition under which

holds.

A B = (E \ A) (E \ B) =

Exercise. For any set A, B, C, Prove that

(a) (A B) \ C = (A \ C) (B \ C),
(b) A \ (B \ C) = (A \ B) (A C),
(c) A \ (B C) = (A \ B) \ C

Relations and mappings

Let (x, y) be the coordinates of a point P in the xy-plane. If x 6= y, then (y, x) are the
coordinates of another point P . In this example, the order in which x and y appear in
the coordinates (x, y) is important.
y

P (x, y)

P (y, x)

x
O

Definition. The ordered pair of x and y, (x, y), is defined as the set {{x}, {x, y}}, where
x and y are called the first coordinate and the second coordinate of (x, y), respectively.
Theorem. (x, y) = (s, t) if and only if x = s and y = t.
Proof. Let X = {x}, Y = {x, y}, S = {s} and T = {s, t}. By definition, we have
(x, y) = {X, Y } and (s, t) = {S, T }.
Since x = s and y = t, X = S and Y = T so that (x, y) = {X, Y } = {S, T } = (s, t).
Since (x, y) = (s, t), we have {X, Y } = {S, T }. There are two cases to consider.
Case 1. When X = S and Y = T , then x = s and {x, y} = {s, t}, hence x = s and y = t.
Case 2. When X = T and Y = S, then x = s = t and x = y = s, hence x = s and
y = t.

Definition. The Cartesian product of the two sets A and B is defined as
A B = {(a, b) : a A and b B}.

Generally speaking, A B 6= B A.

Example 13. Let A = {1, 2, 3}, B = {4, 5}. Then

A B = {(1, 4), (1, 5), (2, 4), (2, 5), (3, 4), (3, 5)},

B A = {(4, 1), (4, 2), (4, 3), (5, 1), (5, 2), (5, 3)}.

10

Definition. Let A, B be sets. A relation R from A to B is a subset of the Cartesian

product A B. If a A and b B satisfy (a, b) R, then a is said to be R-related to b,
written as aRb. Sets A and B are called, respectively, the set of departure and the set of
destination of the relation R.
Example 14. Let A = {a, b, c, d, e} and B = {1, 2, 3, 4}. Then
R = {(a, 1), (b, 1), (a, 3), (c, 4)},

are relations from A to B.

S = {(a, 2), (b, 2), (c, 3), (d, 4), (e, 1)}

Some authors define a relation R from A to B as an ordered triple R = (A, B, G) =

((A, B), G), with G A B.
Example 15. Let A = {1, 2, 3, 4} and B = {1, 3, 5}. Define a relation R from A to B as
follows. For any a A and b B,
aRb a < b.

(a) List all the elements of R.

(b) Describe the image of R in the xy-plane.

Definition. Let A, B be sets, and R a relation from A to B. The domain Dom(R) of

R is the set of all first coordinates of (x, y) R; the image Im(R) of R is the set of all
second coordinates of (x, y) R, i.e.,
Dom(R) = {a A : (a, b) R},
Im(R) = {b B : (a, b) R}.

Some authors call Dom(R) the first projection pr1 (R), and Im(R) the second projection
pr2 (R), of the relation R.
Example 16. The domain and image of the relation R in example 2 are
Dom(R) = {a, b, c},

Im(R) = {1, 3, 4}.

Definition. Let R be a relation from A to B. The inverse relation R1 of R is the

relation from B to A such that
R1 = {(b, a) B A : (a, b) R}.

11

Obviously, Dom(R) = Im(R1 ) and Im(R) = Dom(R1 ).

A
Dom(R)

B
Im(R)

R
R1

Example 17. The inverse relation of R in Example 2 is

R1 = {(1, a), (1, b), (3, a), (4, c)}.
Definition. Let A, B, C be sets, R a relation from A to B, S a relation from B to C.
The composite relation S R of R and S is a relation from A to C defined by:
S R = {(a, c) A C : (a, b) R, (b, c) S with b B}.
Example 18. Let A = {a, b, c, d, e}, B = {1, 2, 3, 4}, C = {x, y, z},
S = {(a, 2), (b, 2), (c, 3), (d, 4), (e, 1)}

T = {(2, x), (2, y), (3, z), (4, y)}.

S and T are the relations from A to B, and from B to C, respectively. The composite
relation of S and T is
T S = {(a, x), (a, y), (b, x), (b, y), (c, z), (d, y)}.
Theorem. Let A, B, C, D be sets, R, S, T be relations from A to B, from B to C, and
from C to D, respectively. Then
(a) (R1 )1 = R,
(b) (T S) R = T (S R),
(c) (S R)1 = R1 S 1 .
Proof. (a) Obviously (R1 )1 is a relation from A to B. Since
R1 = {(b, a) B A : (a, b) R},
(R1 )1 = {(a, b) A B : (b, a) R1 }
= {(a, b) A B : (a, b) R}
= R.

(b) Obviously, both T (S R) and (T S) R are relations from A to D. Since

S R = {(a, c) A C : (a, b) R and (b, c) S},

T S = {(b, d) B D : (b, c) S and (c, d) T },

12

T (S R) = {(a, d) A D : (a, c) S R and (c, d) T }

= {(a, d) A D : (a, b) R, (b, c) S and (c, d) T }

= {(a, d) A D : (a, b) R and (b, d) T S}

(c)

= (T S) R.


Definition. A mapping from a set A to a set B, also called a function, is a relation f
from A to B satisfying the following conditions:
(i) Dom(f ) = A;
(ii) (x, y) f and (x, y ) f y = y .
We write f : A B to mean f being a mapping from A into B. Sets A and B are called
the domain and the codomain of f , respectively.
Condition (i) states that f (x) is defined for each x A; condition (ii) states that f (x)
is well defined for each x A, i.e., f is single-valued at each x A.
Example 19. Let A = {1, 2, 3, 4} and B = {a, b, c, d, e}. The following are mappings
from A to B:
(a) f1 = {(1, a), (2, b), (3, b), (4, c)},
(b) f2 = {(1, d), (2, c), (3, b), (4, a)}.
The following are not mappings from A to B:
(c) g1 = {(1, a), (2, b), (3, c)} (g1 (4) is not defined),
(d) g2 = {(1, a), (1, b), (2, a), (3, a), (4, c)} (g2 (1) has two images in B),
(e) g3 = {(1, d), (2, e), (3, f ), (4, h)} (g3 (3) = f 6 B and g3 (4) = h 6 B).


Definition. Let f : A B be a mapping. When (x, y) f , we write y = f (x) and say

that y is the image of x under f ; x is called the pre-image of y.

13

Definition. Let A B. The mapping f : A B defined by f (x) = x for each x A is

called the inclusion mapping of A into B. The inclusion mapping from A to A is called
the identity mapping of A, denoted by iA .
Definition. Let A, B be two non-empty sets and b B as a fixed element. The mapping
f : A B such that f (x) = b for all x A is called the constant mapping from A to B
with the value b.
Definition. Two mappings f : A B and g : C D are equal if and only if A = C,
B = D and f (x) = g(x) for all x A.
Definition. Let f : A B and g : B C be two mappings. The composite mapping
of f and g are defined as the mapping g f : A C such that g f (x) = g(f (x)) for all
x A.
gf
A
B
C
f
x

g
f (x)

g(f (x))

Example 20.
(a) Let f : R R and g : R R be defined by f (x) = 2x and g(x) = cos x, with
x R. Obviously, g f : R R and
g f (x) = g(f (x)) = cos 2x,

x R.

(b) Let 0 < a 6= 1, f : R+ R and g : R R+ defined as f (x) = loga x and

g(y) = ay , where x R+ = {u R : 0 < u < }, and y R. The composite
mapping g f : R+ R+ is given by
g f (x) = g(f (x)) = aloga x = x,

x R+ .

Definition. Let f : A B, X A and Y B. The set

f [X] = {y B : y = f (x) for some x X} = {f (x) : x X}

is called the direct image of X under f . The direct image of the domain A under f , f [A],
is called the range of f . The set
f 1 [Y ] = {x A : f (x) Y }

is called the inverse image of Y under f .

Obviously, for any X A and Y B, f [X] is a subset of B, and f 1 [Y ] a subset of A.

14

Note. Whether the inverse mapping f 1 exists or not, the inverse image f 1 [Y ] always
exists.
Example 21. Let A = {1, 2, 3, 4, 5, 6}, B = {a, b, c, d, e, f } and g : A B be defined as

If X = {2, 3, 4}, then

If Y = {b, c, f }, then

g(1) = a,

g(2) = g(5) = b,

g(4) = e,

g(6) = f.

g(3) = d,

g[X] = {g(2), g(3), g(4)} = {b, d, e}.

g 1 [Y ] = {2, 5, 6}.
Let a, b R such that a 6 b. Certain intervals of real numbers are defined as follows:
(a, b) = {x R : a < x < b},

(a, b] = {x R : a < x 6 b},

[a, b) = {x R : a 6 x < b},

[a, b] = {x R : a 6 x 6 b},

(a, ) = {x R : a < x < },

[a, ) = {x R : a 6 x < },

(, a) = {x R : < x < a},

(, a] = {x R : < x 6 a},

(, ) = {x R : < x < } = R .
Example 22. Let A = [0, 2], B = [1, 1], and f : A B be the mapping defined by
f (x) = sin 2x for all x A. Let X = [0, 3 ] and Y = ( 21 , 0). Find f [X] and f 1 [Y ].
Theorem. Let f : A B be a mapping, X, X be subsets of A, and Y, Y be subsets of
B. Then
(i) X X f [X] f [X ].
(ii) f [X X ] = f [X] f [X ].
(iii) f [X X ] f [X] f [X ].
(iv) Y Y f 1 [Y ] f 1 [Y ].
(v) f 1 [Y Y ] = f 1 [Y ] f 1 [Y ].
(vi) f 1 [Y Y ] = f 1 [Y ] f 1 [Y ].
(vii) f 1 [B \ Y ] = A \ f 1 [Y ].
(viii) X f 1 [f [X]].
(ix) Y f [f 1 [Y ]].
Proof. (i) Since
y f [X] y = f (x) for some x X

y = f (x) for some x X

f [X] f [X ] follows.

y f [X ],

( X X )

15

(ii) y,

y f [X X ] y = f (x) for some x X X

y = f (x) for some x X or x X

y f [X] or y f [X ]
y f [X] f [X ].

Thus, f [X X ] = f [X] f [X ].
(iv) x,

x f 1 [Y ] f (x) Y

f (x) Y

f 1 [Y ] f 1 [Y ] follows.

( Y Y )

x f 1 [Y ].

Exercise: Prove parts (iii), (vi)(ix) of the previous theorem.

The inclusions of in (iii) and (ix) of the previous theorem can be proper. For example,
In theorem (viii) if a A \ X exists and satisfy f (a) = b f [X], then X $ X {a}
f 1 [f [X]].
f
A
B
f 1 [f [X]]
a

f [X]
b

In theorem (ix), if there exists y Y \ f [A], then f [f 1 [Y ]] $ Y .

f

A
f 1 [Y ]

B
f [A]
Y
y

16

Defintion. A mapping f : A B is called

(i) injective (or, one-to-one) if for any x1 , x2 A, f (x1 ) = f (x2 ) x1 = x2 (or
x1 6= x2 f (x1 ) 6= f (x2 ));
(ii) surjective (or, onto) if f [A] = B (or, for any y B, there exists x A such that
f (x) = y);
(iii) bijective (or, one-to-one correspondence) if f is both injective and surjective.
Injective (resp., surjective, bijective) mappings are also called injections (resp., surjections,
bijections).
Example 23. Let f (x) = x2 and
A = {x R : 1 6 x 6 1},

C = {x R : 0 6 x 6 2}.

B = {x R : 0 6 x 6 1},

then
(a) f : A B is surjective but not injective (because f (1) = f (1) = 1);
(b) f : B C is injective but not surjective (because 2 C has no pre-image);
(c) f : A C is neither injective (f (1) = f (1) = 1) nor surjective (2 C has no
pre-image);
(d) f : B B in a bijective.
Example 24. Let f : A B and g : B C be surjective mappings. Prove that g f
is also surjective.
Solution. It is obvious that g f : A C. Let z C. Since g is surjective, there exists
y B such that g(y) = z. Since f is surjective, there exists x A such that f (x) = y.
So, z = g(y) = g(f (x)) = g f (x). Since z C is arbitrary, g f is surjective.

Example 25.
(a) Prove that for any a, b R, a2 + ab + b2 + 3 > 0.
(b) Let f : R R be the mapping defined by f (x) = x3 + 3x 2, for all x R. Prove
that f is injective.
Solution. (a) By completing the square, we have for any a, b R,

3b2
b
+ 3 > 0.
a2 + ab + b2 + 3 = (a + )2 +
2
4
(b) Let x1 , x2 R be such that f (x1 ) = f (x2 ). Then

x31 + 3x1 2 = x32 + 3x2 2

x31 x32 + 3(x1 x2 ) = 0

(x1 x2 )(x21 + x1 x2 + x22 ) + 3(x1 x2 ) = 0

(x1 x2 )(x21 + x1 x2 + x22 + 3) = 0.

Since x21 + x1 x2 + x32 + 3 > 0 by (a), we must have x1 x2 = 0, i.e., x1 = x2 , thus proving
the injectivity of f .


17

Example 26. Let A, B R. A mapping f : A B is said to be strictly increasing if for

any x, y A, x < y f (x) < f (y). Every strictly increasing function is injective.

Example 27. Let f : [ 21 , 0] [1, 0] be the mapping defined by f (x) =

x [ 21 , 0]. Prove that f is surjective.

x
,
1+x

where

Exercise. Let N = {1, 2, 3, . . .} and N0 = {0, 1, 2, 3, . . .}. Define a mapping f : N0 N0

N by
f ((m, n)) = 2m (2n + 1), for any (m, n) N0 N0 .
Prove that f is bijective.
Example 28. Let A, B, C be sets, f : A B and g : B C be mappings. Prove that
(a) If g f is surjective, then g is surjective;
(b) If g f is injective, then f is injective.

Solution. It is clear that g f : A C.

(a) Let c C. Since g f is surjective, a A such that (g f )(a) = c. Since
(g f )(a) = g(f (a)), b = f (a) B satisfies c = g(b).
(b) Let a, a A satisfy f (a) = f (a ). Then
g(f (a)) = g(f (a ))

(g f )(a) = (g f )(a )
a = a

( g f is injective)

18

Let f : A B be a mapping. Regarding f as a relation, by interchanging the roles of the

sets A and B, we obtain the inverse relation of f 1 .
Question: Under what condition will the inverse relation f 1 be a mapping?
Before answering this question, let us consider the following cases.

Let A = {a, b, c}, B = {p, q, r, s} and f : A B be defined as follows:

f
A

p
q
r
s

a
b
c

Since f maps different x A to different f (x) B, f is injective. Since f [A] = {q, r, s} 6=

B, f is not surjective.
Reverse the directions of the arrows, one obtains f 1 , the inverse relation of f :
f 1
B

A
p
q
r
s

a
b
c

Since f 1 (p) is not defined, f 1 is not a mapping.

Considering now the mapping f : A B, where A = {a, b, c, d}, B = {p, q, r}, and f be
defined as follows:
f
A

B
a
b
c
d

p
q
r

Since f [A] = B, f is surjective. Since f (a) = f (b) = p, f is not injective.

19

Reverse the directions of the arrows, one obtains the relation f 1 :

f 1
B

A
a
b
c
d

p
q
r

Since f 1 (p) = a and f 1 (p) = b, f 1 (p) is not well-defined, so f 1 is not a mapping.

From the above discussion, the bijectivity of the mapping f : A B is necessary and
sufficient for the existence of the inverse mapping f 1 : B A.
Theorem. Let f : A B be a mapping. Then f 1 : B A exists and is bijective if and
only if f is bijective. Besides f f 1 = iB and f 1 f = iA .
Proof. To every y B, there exists exactly one x A such that f (x) = y. One
can define f 1 (y) = x. The mapping f 1 : B A so defined is the inverse of f , which is
bijective.
Let x1 , x2 A be such that f (x1 ) = f (x2 ). Since f 1 : B A is a mapping,
x1 = f 1 f (x1 ) = f 1 f (x2 ) = x2 .

So f is injective. Let y B. Then f 1 (y) = x A so that y = f (f 1 (y)) = f (x). So f

is surjective.
Since f f 1 : B B, for each y B,
f f 1 (y) = f (f 1 (y)) = f (x) = y = iB (y),

So f f 1 = iB . The remaining assertion f 1 f = iA follows from a similar argument. 

Example 29. Let A be a finite set and let f : A A be a mapping. Prove that f is
injective if and only if f is surjective.
Solution. Since f is injective, it follows that |f [A]| = |A|. As f [A] A and |f [A]| =
|A|, necessarily f [A] = A, i.e., f is surjective.
Suppose on the contrary that f is not injective. Then there exist distinct a, a A
such that f (a) = f (a ). But then |f [A]| < |A| so that f [A] $ A, contradicting the
surjectivity of f .

Remark. If A is a finite set and f : A A is bijective, then f is a permutation of A.
Graphs of mappings. Let f : A B as a mapping. Its graph is the set
{(x, f (x)) : x A}.

If f 1 : B A exists as a mapping, its graph is the set

{(f (x), x) : x A},

20

whose elements are obtained by swapping the coordinates of the corresponding points in
the graph of f . Geometrically speaking, the graph of f 1 is readily obtained by reflecting
the graph of f with respect to the diagonal line y = x.
Example 30. The inverse of f : R R+ defined by f (x) = ex/2 for x R+ , is the
mapping g : R+ R defined by g(x) = 2 ln x, where x R+ ; the image of g can be
obtained from the image of f by reflection with respect to the diagonal line y = x:
y
y=x

y=f (x)

y=g(x)
x

Exercise. By considering the graph of y = sin(x), sketch the graph of y = sin1 (x).

21

Equivalence relations
Definition. Let A be a set, R A A a relation from A to A. The relation R is called
(E1) reflexive for any a A, (a, a) R;
(E2) symmetric for any (a, b) R, (b, a) R;
(E3) transitive for any (a, b), (b, c) R, (a, c) R.
If the relation R is reflexive, symmetric and transitive, it is an equivalence relation.
Definition. The diagonal of a set A is a relation DA defined as
DA = {(a, a) : a A}.
The reflexivity, symmetry and transitivity can also be expressed as:
(E1) R is reflexive DA R.
(E2) R is symmetric R = R1 .
(E3) R is transitive R R R.
(E1)(E3) can be represented by the R image:
(E1)

(E2)

(E3)

(b,a)
(a,a)
A

DA

(b,b)

(a,b)

(b,c)

(a,c)

(a,b)

Definition. For any a A, define

a/R = {b A : aRb},
called the equivalence class containing a, also denoted by [a]R . Each element of the
equivalence class a/R is a representative of the class.
Definition. Let R be an equivalence relation of a set A, and let
A/R = {a/R : a A},
called the quotient set of A by R, which is the set of all equivalence classes of R. Since
a/R A for any a A, it follows that A/R P(A).
Theorem. Let R be an equivalence relation from A to A. Then for any a, b A,
(a) a a/R;
(b) a/R = b/R aRb;
(c) either a/R b/R = or a/R = b/R.

22

Proof. (a) Since aRa for any a A, a a/R follows.

(b) : aRb b a/R. Let y b/R. Then aRb and bRy aRy y a/R.
So b/R a/R. Interchanging the roles of a and b, a/R b/R follows. Consequently,
a/R = b/R.
: Using (a) to obtain b b/R = a/R aRb.
(c) If a/R b/R = , then there is nothing to prove. Otherwise, let x a/R b/R.
Since aRx and bRx, the symmetry and transitivity of R imply aRb. By (b), we then get
a/R = b/R.

Definition. Let A be a set. A collection P of subsets of A such that
S
(P1) BP B = A,
(P2) if B, B P are distinct, then B B = ,
is called a partition of A. The elements of P are called blocks.

Theorem. Let R be a equivalence relation defined on set A. Its quotient set A/R is a
partition of the set A.
S
Proof. Since aA a/R = A, (P1) holds. Since distinct equivalence classes are disjoint,
(P2) holds.

From the above theorem, every equivalence relation R defined on a set A gives rise to a
partition of A. The converse of the above theorem is also true.
Theorem. Let P = {S1 , S2 , . . . , Sk } be a partition of the set A. There exists an equivalence relation on A such that A/ = P.
Proof. Let a, b A. Declare a b a, b belong to the same block, i.e., a, b Si , for
some 1 6 i 6 k. For any a, b, c A,
S
(E1) since A = ki=1 Si , a, a Si for some i so that a a.
(E2) if a b, then a, b Si for some i so that b, a Si , i.e., b a.
(E3) if a b and b c, then a, b Si and b, c Sj for some i, j; necessarily i = j so
that a, c Si , i.e., a c.
Thus, is an equivalence relation and the quotient set A/ = P.

Exercise. Let L be the set of all straight lines in the xy-plane. For any 1 , 2 L ,
define 1 R2 1 //2 .
(a) Prove that R is an equivalence relation.
(b) Illustrate the equivalence class containing the straight line y = 2x + 1.

23

Example 31. Let N0 as the set of all non-negative integers. A relation R is defined in
N0 as follows: For any (m, n), (p, q) N0 N0 , (m, n)R(p, q) m + q = n + p.
(a) Prove that R is a equivalence relation.
(b) Illustrate the equivalence class containing (m, n).
Solution. (a) Let (m, n), (p, q), (r, s) N0 N0 .
(E1) Since m + n = n + m, (m, n)R(m, n) so that S is reflexive.
(E2) If (m, n)R(p, q), then
m + q = n + p p + n = q + m (p, q)R(m, n)

so that R is symmetric.
(E3) If (m, n)R(p, q) and (p, q)R(r, s), then m + q = n + p and p + s = q + r so that
(m + s) + q = (m + q) + s = (n + p) + s
= n + (p + s) = n + (q + r)
= (n + r) + q.
Cancelling q from both sides, we get m + s = n + r (m, n)R(r, s). Hence, R
is transitive.
Thus, R is an equivalence relation.
(b) If (p, q) (m, n)/R, then (m, n)R(p, q) m + q = n + p m n = p q. One
can identify (p, q) to the difference p q. Thus, the equivalence class (m, n)/R containing
(m, n) contains all differences which are equivalent to m n.

Example 31 shows how to construct any integer from non-negative integers as equivalence
classes of the relation R.
Example 32. Let Z = Z \{0}. A relation S is defined in Z Z as follow: To any
(m, n), (p, q) Z Z , (m, n)S(p, q) mq = np.
(a) Prove that S is a equivalence relation.
(b) Illustrate the equivalence class containing (m, n).
Solution. (a) Let (m, n), (p, q), (r, s) Z Z .
(E1) Since mn = nm, (m, n)S(m, n) so that S is reflexive.
(E2) If (m, n)S(p, q), then mq = np pn = qm (p, q)S(m, n) so that S is
symmetric.
(E3) If (m, n)S(p, q) and (p, q)S(r, s), then mq = np and ps = qr so that
(ms)q = (mq)s = (np)s = n(ps) = n(qr) = (nr)q.
Since q 6= 0, cancelling q from both sides, we get ms = nr (m, n)S(r, s).
Hence, S is transitive.
= pq . One can
(b) If (p, q) (m, n)/S, then (m, n)S(p, q) mq = np m
n
identify (p, q) to the fraction pq . Thus, the equivalence class (m, n)/S containing (m, n)
contains all fractions which are equivalent to m
.

n
Example 32 shows how to construct rational numbers from integers.

24

Example 33. Let A, S be n n real matrices with S nonsingular. If B = SAS 1 , then A

is similar to B. It is easy to see that similarity is an equivalence relation on matrices. If
A is nonsingular, then so is B. Columns of A and B are linearly independent which form
bases for Rn . Thus, in this case, similarity of matrices corresponds to change of bases of
vector spaces.


25

Real and complex number systems

Reference: W. Rudin, Principles of Mathematical Analysis, 3rd edition, McGraw-Hill,
c1976, Chapter 1.
1.1 Example. We first show that the equation
p2 = 2

(1)

is not satisfied by any p Q. If there were one such p, we may assume that p = m
, where
n
m, n Z are relatively prime (i.e., they have no common factor). Substituting p = m
n
into (1), we have
m2 = 2n2 .
(2)
2
Since m is even, m must be even. Writing m = 2k for some k Z. Then
4k 2 = (2k)2 = 2n2

so that after cancelling 2 from both sides, we get 2k 2 = n2 , which forces n to be even.
But this contradicts m and n being relatively prime.
Now let A = {p Q+ : p2 < 2} and B = {p Q+ : p2 > 2}. We next show that the
set A contains no largest element and the set B contains no smallest element. For every
p Q+ , let
p2 2
2p + 2
q =p
=
.
(3)
p+2
p+2
Then
2(p2 2)
q2 2 =
.
(4)
(p + 2)2
If p A, then p2 2 < 0, (3) shows that q > p, and (4) shows that q 2 < 2. Thus, q A.
If p B, then p2 2 > 0, (3) shows that 0 < q < p, and (4) shows that q 2 > 2. Thus,
q B.

1.2 Remark. The purpose of Example 1.1 is to show that Q has certain gaps, in spite of
the fact that in between any two rational numbers there is another: if r, s Q and r < s,
Q satisfies r < r+s
< s. The real numbers fill these gaps.
then r+s
2
2

26

Ordered sets
1.5 Definition. Let S be an nonempty set. A partial order on S is a relation, denoted
by , satisfying the following properties:
(i) (Antisymmetry) If x, y S are such that x y and y x, then x = y.
(ii) (Transitivity) If x, y, z S such that x y and y z, then x z.
Example. On the set of natural numbers N = {1, 2, 3, . . .}, define by
x y x|y,

x, y N .

1.6 Definition. The nonempty set S together with the order is call a partially ordered
set, abbreviated as poset. Write (S, ) to mean the set S whose partial order is . When
the partial order is understood, we simply write S instead of (S, ).
When any two elements x, y of a poset S is comparable, i.e., either x y or y x is true,
then S is said to be totally ordered. We then say that S is an (totally) ordered set.
The Law of trichotomy. Let S be an ordered set and let x, y S. Then exactly one
of the following holds: x < y, x = y, or x > y.
Since the set of all real numbers R is totally ordered, the Law of Trichotomy holds in R.
That is, given two real numbers, we can always determine whether they are equal, or one
is larger than the other.

1.7 Definition. A subset E of an ordered set S is bounded above (resp., bounded below)
if there exists S (resp., S) such that x 6 (resp., 6 x) for all x E; such a
(resp., ) is called an upper bound (resp., lower bound) of E.
1.8 Definition. Suppose S is an ordered set, E S, and E is bounded above. Suppose
there exists an S with the following properties:
(i) is an upper bound of E.
(ii) If < , then is not an upper bound of E.
Then is called the least upper bound (or supremum) of E and we write
= sup E.
The greatest lower bound (or infimum) of a set E which is bounded below is defined in a
similar manner: The statement
= inf E
means that is a lower bound of E and that no with > is a lower bound of E.

27

1.9 Examples.
(a) Consider the sets A = {p Q+ : p2 < 2} and B = {p Q+ : p2 > 2} of Example
1.1 as subsets of the ordered set Q. The set A is bounded above. In fact, the
upper bounds of A are exactly the members of B. Since B contains no smallest
element, A has no least upper bound in Q. Similarly, B is bounded below: The
set of all lower bounds of B is A {r Q : r 6 0}. Since A has no largest element,
B has no greatest lower bound in Q.
(b) If = sup E exists, then may or may not be a member of E. For instance, let
E1 = {r Q : r < 0} and let E2 = {r Q : r 6 0}. Then
sup E1 = sup E2 = 0,

and 0 6 E1 , 0 E2 .
(c) Let E = { n1 : n = 1, 2, 3, . . .}. Then sup E = 1 E, and inf E = 0 6 E.
1.10 Definition. An ordered set S is said to have the least-upper-bound property if the
following is true: If 6= E S and E is bounded above, then sup E exists in S.
Example 1.9(a) shows that Q does not have the least-upper-bound property.
We shall now show that there is a close relation between greatest lower bounds and least
upper bounds, and that every ordered set with the least-upper-bound property also has
the greatest-lower-bound property.
1.11 Theorem. Suppose S is an ordered set with the least-upper-bound property, 6=
B S, and B is bounded below. Let L = {m : m 6 x, x B}, the set of all lower
bounds of B. Then
= sup L
exists in S, and = inf B. In particular, inf exists in S.
Proof. Since B is bounded below, L 6= . Since L consists of exactly those y S such
that y 6 x for every x B, we see that every x B is an upper bound of L. Thus, L is
bounded above. Our hypothesis about S implies therefore that L has a supremum in S;
call it .
If < then is not an upper bound of L, hence 6 B. It follows that 6 x for
every x B. Thus, L.
If < then 6 L, since is an upper bound of L. We have shown that L but
6 L if > . In other words, is a lower bound of B, but is not if > . This
means that = inf B.


28

Fields
1.12 Definition. A field is a set F with two operations, called addition and multiplication, which satisfy the following axioms: for any x, y, z F ,
(A1) (Closed under addition) x + y F .
(A2) (Commutative law for addition) x + y = y + x.
(A3) (Associative law for addition) (x + y) + z = x + (y + z).
(A4) (Existence of additive identity) 0 F such that 0 + x = x.
(A5) (Existence of additive inverse) To every x F corresponds an element x F
such that x + (x) = 0; x is the additive inverse of x.
(M1) (Closed under multiplication) If x, y F , then xy F .
(M2) (Commutative law for multiplication) xy = yx.
(M3) (Associative law for multiplication) (xy)z = x(yz).
(M4) (Existence of multiplicative identity) 0 6= 1 F such that 1x = x for every x F .
(M5) (Existence of multiplicative inverse) If 0 6= x F , then there exists x1 F such
that x( x1 ) = 1; x1 is the multiplicative inverse of x 6= 0.
(D1) (Distributive law) x(y + z) = xy + xz.
Simply speaking, a field is a set of numbers in which one can add, subtract, multiply and
divide (by nonzero) elements of the set, and the results are still elements of the set.
Our usual experience suggests that the set of all real numbers R, the set of all complex
numbers C, and the set of all rational numbers Q, are fields.
1.13 Remarks.
(a) One usually writes (in any field)
x y,

x
,
y

x + y + z, xyz, x2 , x3 , 2x, 3, . . .

in instead of
x + (y), x( y1 ), (x + y) + z, (xy)z, xx, xxx, x + x, x + x + x, . . .
(b) The field axioms clearly hold in Q, the set of all rational numbers, if addition and
multiplication have their customary meaning. Thus Q is a field.
1.14 Proposition. The axioms for addition imply the following statements.
(a) (Cancellation law) If x + y = x + z then y = z.
(b) (Uniqueness of the additive identity) If x + y = x then y = 0.
(c) (Uniqueness of the additive inverse) If x + y = 0 then y = x.
(d) (x) = x.
Proof. (a) If x + y = x + z, the axioms for addition give
y = 0 + y = (x + x) + y = x + (x + y) = x + (x + z) = (x + x) + z = 0 + z = z.

(b) follows from setting z = 0 in (a).

(c) Setting z = x in (a).
(d) Since x + x = 0, by (c) x = (x).

29

1.15 Proposition. The axioms for multiplication imply the following statements.
(a) If x 6= 0 and xy = xz then y = z.
(b) If x 6= 0 and xy = x then y = 1.
(c) If x 6= 0 and xy = 1 then y = x1 .
1
(d) If x 6= 0 then 1/x
= x.
Proof. Exercise.

1.16 Proposition. The field axioms imply the following statements, for any x, y F ,
(a) 0x = 0.
(b) If x 6= 0 and y 6= 0 then xy 6= 0.
(c) (x)y = (xy) = x(y).
(d) (x)(y) = xy.
Proof. (a) Since 0x + 0x = (0 + 0)x = 0x, Proposition 1.14(b) then implies that 0x = 0.
(b) Suppose on the contrary that xy = 0. Then by (a) and (M5), we have
  
  
1
1
1
1
xy =
0 = 0,
1=
y
x
y
x
which is a contradiction.
(c) Since

(x)y + xy = (x + x)y = 0y = 0,
Proposition 1.14(c) then implies (x)y = (xy). The remaining assertion is proved
similarly.
(d) By (c) and Proposition 1.14(d), we have
(x)(y) = [x(y)] = [(xy)] = xy.


1.17 Definition. An ordered field is a field F which is also an ordered set, such that
(i) x + y < x + z if x, y, z F and y < z,
(ii) xy > 0 if x, y F and x, y > 0.
If x > 0, we call x positive; if x < 0, x is negative.
For example, Q is an ordered field.
1.18 Proposition. The following statements are true in every ordered field.
(a) If x > 0 then x < 0, and vice versa.
(b) If x > 0 and y < z then xy < xz.
(c) If x < 0 and y < z then xy > xz.
(d) If x 6= 0 then x2 > 0. In particular, 1 > 0.
(e) If 0 < x < y then 0 < y1 < x1 .

30

Proof. (a) If x > 0 then 0 = x + x > x + 0, so that x < 0. If x < 0 then

0 = x + x < x + 0, so that x > 0.
(b) Since z > y, we have z y > y y = 0, hence x(z y) > 0, and therefore
xz = x(z y) + xy > 0 + xy = xy.

(c) By (a), (b), and Proposition 1.16(c),

[x(z y)] = (x)(z y) > 0

so that x(z y) < 0, hence xz < xy.

(d) If x > 0, Definition 1.17(ii) gives x2 > 0. If x < 0, then x > 0, hence (x)2 > 0.
But x2 = (x)2 , by Proposition 1.16(d). Since 1 = 12 , 1 > 0.
(e) If y > 0 and v 6 0, then yv 6 0. But y( y1 ) = 1 > 0. Hence y1 > 0. Likewise, x1 > 0. If
we multiply both sides of the inequality x < y by the positive quantity ( x1 )( y1 ), we obtain
1
< x1 .

y

31

The real field

1.19 Theorem. There exists an ordered field R which has the least-upper-bound property. Moreover, R contains Q as a subfield.
Proof. See the Appendix to Chap. 1 of Rudins book.

The second statement means that Q R and that the operations of addition and multiplication in R, when applied to the elements of Q, coincide with the usual operations on
rational numbers. The elements of R are called real numbers.
1.20 Theorem.
(a) (The archimedean property of R) If x, y R and x > 0, then there is a positive
integer n such that nx > y.
(b) If x, y R and x < y, then there exists a p Q such that x < p < y.

Proof. (a) Let A = {nx : n Z+ }. If (a) were false, then y would be an upper bound of
A. But then A has a least upper bound in R. Put = sup A. Since x > 0, x < so
that x is not an upper bound of A. Hence x < mx for some m Z+ . But then
< (m + 1)x A, which is a contradiction.
(b) Since x < y, we have y x > 0, and (a) furnishes a positive integer n such that
n(y x) > 1.

Apply (a) again, to obtain positive integers m1 and m2 such that m1 > nx, m2 > nx.
Then
m2 < nx < m1 .
Hence there is an integer m (with m2 6 m 6 m1 ) such that
m 1 6 nx < m.

Combining these inequalities, we obtain

nx < m 6 1 + nx < ny.

Since n > 0, dividing the above inequality by n we obtain
x < p < y,
where p =

m
.
n

1.21 Theorem. (Existence of nth roots of positive reals) For every real x > 0 and every
integer n >0, there is one and only one positive real y such that y n = x. This number y
is written n x or x1/n .
Proof. (Uniqueness) Let y1 and y2 are both nth roots of x. If y1 6= y2 , then we may
assume that 0 < y1 < y2 . Hence,
x = y1n < y2n = x,
which is a contradiction. Consequently, y1 = y2 .
x
(Existence) Let E = {t R+ : tn < x}. If t = 1+x
then 0 < t < 1 so that tn < t < x.
Thus, t E and E 6= .
If t > 1 + x then tn > t > x so that t 6 E. Thus 1 + x is an upper bound of E.

32

Theorem 1.19 implies the existence of

y = sup E.
n

To prove that y = x we will show that each of the inequalities y n < x and y n > x leads
to a contradiction.
The identity bn an = (b a)(bn1 + bn2 a + + an1 ) yields the inequality
bn an < (b a)nbn1

when 0 < a < b. Assume y n < x. Choose h so that 0 < h < 1 and
x yn
h<
.
n(y + 1)n1
Put a = y, b = y + h. Then
(y + h)n y n < hn(y + h)n1 < hn(y + 1)n1 < x y n .

Thus (y + h)n < x and y + h E. Since y + h > y, this contradicts the fact that y is an
upper bound of E.
Assume y n > x. Put
yn x
k=
.
ny n1
Then 0 < k < y. If t > y k, we conclude that
y n tn 6 y n (y k)n < kny n1 = y n x.

Thus, tn > x and t 6 E. It follows that y k is an upper bound of E.

But y k < y, which contradicts the fact that y is the least upper bound of E.
Hence, y n = x, and the proof is complete.

Corollary. If a, b R+ and n Z+ , then

(ab)1/n = a1/n b1/n .

Proof. Put = a1/n , = b1/n . Then

ab = n n = ()n ,
since multiplication is commutative. The uniqueness assertion of Theorem 1.21 shows
therefore that
(ab)1/n = = a1/n b1/n .

1.22 Decimals. Let x > 0 be real. Let n0 be the largest integer such that n0 < x. (By
the archimedean property of R, n0 exists.) Having chosen n0 , n1 , . . . , nk1 , let nk be the
largest integer such that
nk
n1
+ + k 6 x.
n0 +
10
10
Let E be the set of these numbers
n1
nk
n0 +
+ + k (k = 0, 1, 2, . . .).
(5)
10
10
Then x = sup E. The decimal expansion of x is
n0 .n1 n2 n3 . . . .

(6)

33

Conversely, for any infinite decimal (6) the set E of numbers (5) is bounded above, and
(6) is the decimal expansion of sup E.

The extended real number system

= R {, +} = [, +],
1.23 Definition. The extended real number system is R
which consists of the real field R and two symbols, + and . We preserve the original
order in R, and define
< x < +
for every x R.
and that every nonempty
It is then clear that + is an upper bound of every subset of R,
subset has a least upper bound. If 6= E R is not bounded above in R, then sup E =
Exactly the same remarks apply to lower bounds. The extended real number
+ in R.
system does not form a field, but it is customary to make the following conventions:
(a) If x R then
x
x
=
= 0.
x + = +, x = ,
+

(b) If x > 0 then x (+) = +, x () = .

(c) If x < 0 then x (+) = , x () = +.
Any x R is finite, and are infinite.

34

The complex field

1.24 Definition. A complex number is an ordered pair (a, b) of real numbers. Ordered
means that (a, b) and (b, a) are distinct if a 6= b.
Definition. Two complex numbers z = (a, b), w = (c, d) are equal if and only if a = c
and b = d. Define
z + w = (a + c, b + d),

zw = (ac bd, ad + bc).

Writing z = a + ib and w = c + id, then

z + w = (a + c) + i(b + d),

zw = (a + ib)(c + id) = (ac bd) + i(ad + bc).

Denote by C = {a + ib : a, b R} the set of all complex numbers. C forms a field under

the addition and multiplication of complex numbers.
1.30 Definition. Let z = a + ib C. The complex conjugate of z is z = a ib. The
numbers a and b are the real part and imaginary part of z, denoted Re(z) and Im(z),
respectively.
1.31 Theorem. Let z, w C. Then

(a) z + w = z + w,
(b) zw = zw,

(c) z + z = 2 Re(z), z z = 2i Im(z),

(d) z z > 0 and the equality holds if and only if z = 0.
Proof. Let z = a + ib and w = c + id, where a, b, c, d R. Then

(a) z + w = (a + c) + i(b + d) = (a + c) i(b + d) = (a ib) + (c id) = z + w.

(b) zw = (ac bd) + i(ad + bc) = (ac bd) i(ad + bc) = (a ib)(c id) = zw.

(c) z + z = (a + ib) + (a ib) = 2a = 2 Re(z); z z = (a + ib) (a ib) = 2ib = 2i Im(z).

(d) z z = (a + ib)(a ib) = a2 + b2 > 0; the equality holds iff a = b = 0 iff z = 0.

1.32 Definition. Let z = a + ib C. Its modulus |z| =
root of z z.

z z, the non-negative square

Note that |z| is the complex

analogue of the absolute value of real numbers. When x R,
p
x = x so that |x| = x2 . Thus, |x| = x (resp., |x| = x) when x > 0 (resp., x < 0).
Theorem. Let z, w C. Then
(a) (positive definiteness) |z| > 0; the equality holds if and only if z = 0,
(b) |
z | = |z|,
(c) |zw| = |z||w|,
(d) | Re(z)| 6 |z|; the equality holds if and only if z is real,
(e) (triangle inequality) |z + w| 6 |z| + |w|, the equality holds if and only if z = kw
for some non-negative k R.

35

Proof. Let z = a + ib and w = c + id, with a, b, c, d R.

(a) It is clear that |z| = a2 + b2 > 0; the equality holds iff a = b = 0.

p

(b) |
z | = |a ib| = a2 + (b)2 = a2 + b2 = |z|.
(c) Since zw = (ac bd) + i(ad + bc),

|zw|2 = (ac bc)2 + (ad + bc)2 = (a2 + b2 )(c2 + d2 ) = |z|2 |w|2 .

Taking square roots of both sides, the result follows.

(d) This follows from | Re(z)| = |a| = a2 6 a2 + b2 = |z|; the equality holds b =
0 z is real.
(e)
|z + w|2 = (z + w)(z + w)

= (z + w)(
z + w)

= z z + z w + zw + ww

= |z|2 + z w + z w + |w|2

= |z|2 + 2 Re(z w)
+ |w|2

6 |z|2 + 2|z w|
+ |w|2

= |z|2 + 2|z||w|
+ |w|2

= |z|2 + 2|z||w| + |w|2

= (|z| + |w|)2 .

The equality holds Re(z w)

= |z w|
z w is real and non-negative. Since
z w = (a + ib)(c id) = (ac + bd) + i(bc ad), z w R bc ad = 0 db = ac = k

for some k R; z w = z(kz) = k|z|2 > 0 k > 0.
1.35 Theorem. (Cauchy-Schwarz inequality) If z1 , . . . , zn , w1 , . . . , wn C, then

2
n
n
n
X

X
X


2
zi wi  6
|wi |2 .
|zi |



i=1

i=1

z1
w1

z2
w2

i=1

zn
.
wn

=
= =
The equality holds if and only if
Pn
P
P
Proof. Let A = i=1 |zi |2 , B = ni=1 |wi |2 and C = ni=1 zi wi . If B = 0, then w1 = =
wn = 0 and the inequality is trivial. So, suppose B > 0. By Theorem 1.31, we have
n
X
i=1

|Bzi Cwi |2 =

n
X
i=1

= B2

(Bzi Cwi )(B zi C wi )

n
X
i=1

Pn

|zi |2 B C
2

= B(AB |C| ).

n
X
i=1

zi wi BC

n
X
i=1

zi wi + |C|2

n
X
i=1

|wi |2

Since i=1 |Bzi Cwi |2 > 0 and B > 0, it follows that AB > |C|2 . The equality holds
C
|Bz1 Cw1 | = = |Bzn Cwn | = 0 wz11 = = wznn = B
.


36

Euclidean spaces
1.36 Definitions. For each positive integer k, let Rk be the set of all ordered k-tuples
x = (x1 , x2 , . . . , xk )T ,
where x1 , . . . , xk R, called the coordinates of x. The elements of Rk are called points,
or vectors, especially when k > 1. If y = (y1 , . . . , yk )T and if R, put
x + y = (x1 + y1 , . . . , xk + yk )T

and x = (x1 , . . . , xk )T

so that x+y Rk and x Rk . This defines addition of vectors and scalar multiplication
of a vector by a real scalar. These two operations satisfy the commutative, associative,
and distributive laws and make Rk into a vector space over the real field. The zero element
of Rk (called the zero vector) is the point 0 all of whose coordinates are 0. We also define
the usual inner product (or scalar product) of x and by
xy =
and the norm of x by
kxk2 = (x x)

k
X

1/2

xi yi

i=1

X
k

x2i

i=1

1/2

The structure defined (the vector space Rk with the above inner product and norm) is
called the k-dimensional Euclidean space.
1.37 Theorem. Suppose x, y, z Rk , and R. Then
(a) kxk2 > 0; the equality holds if and only if x = 0.
(b) kxk2 = ||kxk2 ;
(c) |x y| 6 kxk2 kyk2 ;
(d) kx + yk2 6 kxk2 + kyk2 .
Proof. (a) It is clear that
kxk22

k
X

x2i > 0;

i=1

the equality holds x1 = x2 = = xk x = 0. (b) follows from

1/2
1/2
X
1/2  X
X
k
k
k
2
2
2
2
= ||kxk2 .
= ||
xi
kxk2 =
xi
(xi )
=
i=1

i=1

i=1

(c) The inequality is trivial if x = 0. So, we suppose x 6= 0 so that kxk2 6= 0. Since

0 6 ktx + yk2 = t2 kxk22 + 2t(x y) + kyk22

for all t R, its discriminant

4((x y)2 kxk22 kyk22 ) 6 0,

i.e., (x y)2 6 kxk22 kyk22 . The desired inequality then follows upon taking square roots of
both sides.

37

(d) follows from

kx + yk22 = (x + y) (x + y)

= x x + 2(x y) + y y
6 kxk22 + 2|x y| + kyk22

6 kxk22 + 2kxk2 kyk2 + kyk22

(c)

= (kxk2 + kyk2 )2 .

1.38 Remarks. Theorem 1.37 (a) and (d) will allow us to regard Rk as a metric space.
R1 (the set of all real numbers) is the real line; R2 is the xy-plane, or the complex plane.
In these two cases the norm is just the absolute value of the corresponding real or complex
number.

38

Basic Topology
2.3 Definition. Let A and B be sets. If there exists a bijective mapping f : A B,
then A and B have the same cardinal number. This means that if A and B are finite sets,
then A and B have the same number of elements; if A and B are infinite, then they have
the same level of infinity.
Proposition. Declare two sets A and B are equivalent, written A B, if there exists a
bijection f : A B. The relation between sets is an equivalence relation.
Proof. Exercise.

2.4 Definition. For any positive integer n, let Jn be the set whose elements are the
integers 1, 2, . . . , n; let N be the set of all positive integers. For any set A,
(a) A is finite if A Jn for some n (the empty set is also considered to be finite).
(b) A is infinite if A is not finite.
(c) A is countable if A N.
(d) A is uncountable if A is neither finite nor countable.
(e) A is at most countable if A is finite or countable.
Countable sets are sometimes called enumerable, or denumerable.
2.5 Example. Let Z be the set of all integers. Then Z is countable. For, consider the
following arrangement of the sets Z and N:
Z : 0, 1, 1, 2, 2, 3, 3, . . .
N : 1, 2, 3, 4, 5, 6, 7, . . .
An explicit formula for a bijection f : N Z can be given:
(
n
if n even,
f (n) = 2 n1
if n is odd.
2
2.6 Remark. A finite set cannot be equivalent to one of its proper subsets. This is,
however, possible for infinite sets, as is shown by Example 2.5, in which N is a proper
subset of Z.
In fact, we could replace Definition 2.4(b) by the statement: A is infinite if A is equivalent
to one of its proper subsets.

39

2.7 Definition. By a sequence, we mean a function f defined on the set N of all positive
integers. If f (n) = xn for n N, it is customary to denote the sequence f by the symbol
{xn }, or sometimes by x1 , x2 , x3 , . . .. The values of f , that is, the elements xn , are called
the terms of the sequence. If A is a set and if xn A for all n N, then {xn } is said to
be a sequence in A, or a sequence of elements of A.
Note that the terms x1 , x2 , x3 , . . . of a sequence need not be distinct.
Since every countable set is the range of a injection defined on N, we may regard every
countable set as the range of a sequence of distinct terms.
Speaking more loosely, we may say that the elements of any countable set can be arranged
in a sequence. Sometimes it is convenient to replace N in this definition by the set of all
nonnegative integers N0 = {0, 1, 2, . . .}.
2.8 Theorem. Every infinite subset of a countable set A is countable.
Proof. Let E be an infinite subset of A. Arrange the elements x of A in a sequence {xn }
of distinct terms. Construct a sequence {nk } as follows: Let n1 be the smallest positive
integer such that xn1 E. Having chosen n1 , . . . , nk1 (k = 2, 3, 4, . . .), let nk be the
smallest integer greater than nk1 such that xnk E.
Defining f (k) = xnk (k = 1, 2, 3, . . .), we obtain a bijection between E and N.


The theorem shows that, roughly speaking, countable sets represent the smallest infinity: No uncountable set can be a subset of a countable set.
2.9 Definition. Let A and be sets, and suppose that with each element of A there
is associated a subset of which we denote by E .
The set whose elements are the sets E will be denoted by {E }. Instead of speaking
of sets of sets, we shall sometimes speak of a collection of sets, or a family of sets.
2.10 Examples. Let A = {x R : 0 < x 6 1}. For every x A, let Ex = {y R : 0 <
y < x}. Then
(i) E
Sx Ez if and only if 0 < x 6 z 6 1;
(ii) TxA Ex = E1 ;
(iii) xA Ex = .
(i) and
T (ii) are clear. To prove (iii), we note that for every y > 0, y 6 Ex if x < y. Hence,
y 6 xA Ex .

40

2.12 Theorem. Let {En }, n = 1, 2, 3, . . ., be a sequence of countable sets, and put

[
S=
En .
(15)
n=1

Then S is countable.

Proof. Let every set En be arranged in a sequence {xnk }, k = 1, 2, 3, . . ., and consider the
infinite array
x11
x12
x13
x14 . . .

x21
x22
x23
x24 . . .

(16)
x31
x32
x33
x34 . . .

x41
x42
x43
x44 . . .
... ... ... ... ... ... ... ...
in which the elements of En form the nth row. The array contains all elements of S, which
can be arranged in a sequence
x11 ; x21 , x12 ; x31 , x22 , x13 ; x41 , x32 , x23 , x14 ; . . .

(17)

If any two of the sets En have elements in common, these repeated elements will occur in
(17). Hence, there is a subset T of the set of all positive integers such that S T , which
shows that S is at most countable (Theorem 2.8). Since E1 S, and E1 is infinite, S is
infinite, and thus countable.

Corollary. Suppose A is at most countable, and, for every A, B is at most
countable. Put
[
T =
B .
A

Then T is at most countable.

2.13 Theorem. Let A be a countable set. The set of all ordered n-tuples
is countable.

Bn = {(a1 , . . . , an ) : a1 , . . . , an A}

Proof. That B1 is countable is evident, since B1 = A. Suppose Bn1 is countable (n =

2, 3, 4, . . .). The elements of Bn are of the form
(b, a) (b Bn1 , a A).

(18)

For every fixed b, the set of pairs (b, a) is equivalent to A, and hence countable. Thus Bn
is the union of a countable set of countable sets. By Theorem 2.12, Bn is countable. The
theorem follows by induction.


41

Corollary. The set of all rational numbers Q is countable.

Proof. We apply Theorem 2.13, with n = 2, noting that every rational r is of the form ab ,
where a and b are integers. The set of pairs (a, b), and therefore the set of fractions ab , is
countable.

2.14 Theorem. Let A be the set of all sequences whose elements are the digits 0 and 1,
e.g., the elements of A are sequences like 1, 0, 0, 1, 0, 1, 1, 1, . . .. The set A is uncountable.
Proof. Let E be a countable subset of A, and let E consist of the sequences s1 , s2 , s3 , . . ..
We construct a sequence s as follows. If the nth digit in sn is 1, we let the nth digit of s
be 0, and vice versa. Then the sequence s differs from every member of E in at least one
place; hence s 6 E. But clearly s A, so that E is a proper subset of A.
We have shown that every countable subset of A is a proper subset of A. It follows that
A is uncountable (for otherwise A would be a proper subset of A, which is absurd). 
The idea of the above proof was first used by Cantor, and is called Cantors diagonal
process; for, if the sequences s1 , s2 , s3 , . . . are placed in an array like (16), it is the elements
on the diagonal which are involved in the construction of the new sequence.
Since any real number can be represented in base 2, Theorem 2.14 implies that the set of
all real numbers is uncountable.

42

Metric spaces
2.15 Definition. A set X, whose elements we shall call points, is said to be a metric
space if with any two points p, q X there is associated a real number d(p, q), called the
distance from p to q, such that
(a) (positive definiteness) d(p, q) > 0, with the equality holds if and only if p = q;
(b) (symmetry) d(p, q) = d(q, p);
(c) (triangle inequality) d(p, q) 6 d(p, r) + d(r, q), for any r X.
Any function satisfying these three properties is called a distance function, or a metric.

2.16 Examples. The most important examples of metric spaces are the Euclidean spaces
Rk , especially R1 (the real line) and R2 (the complex plane); the distance in Rk is defined
by
d(x, y) = kx yk2 (x, y Rk ).
(19)
By Theorem 1.37, the conditions of Definition 2.15 are satisfied by (19).
It is important to observe that every subset Y of a metric space X is a metric space in
its own right, with the same distance function. For it is clear that if conditions (a)(c)
of Definition 2.15 hold for p, q, r X, they also hold if p, q, r are restricted to lie in Y .
Thus, every subset of a Euclidean space is a metric space.
2.17 Definition.
Segment: (a, b) = {x R : a < x < b}; interval: [a, b] = {x : a 6 x 6 b};
half-open intervals: [a, b) = {x R : a 6 x < b} and (a, b] = {x R : a < x 6 b}.
If ai < bi for i = 1, . . . , k, the set {(x1 , . . . , xk ) : ai 6 xi 6 bi , 1 6 i 6 k} is called a k-cell.
Thus a 1-cell is an interval, a 2-cell is a rectangle, etc.
If x Rk and r > 0, the open (resp., closed) ball B with center at x and radius r is defined
as Br (x) = {y Rk : ky xk2 < r} (resp., ky xk2 6 r).
We call a set E Rk convex if

x + (1 )y E

whenever x, y E, and 0 < < 1. For example, balls are convex. For if ky xk2 < r,
kz xk2 < r, and 0 < < 1, we have
ky + (1 )z x| = k(y x) + (1 )(z x)k2
6 ky xk2 + (1 )kz xk2

< r + (1 )r
= r.

The same proof applies to closed balls. It is also easy to see that k-cells are convex.

43

2.18 Definition. Let X be a metric space. All points and sets mentioned below are
understood to be elements and subsets of X.
(a) A neighborhood of p is a set Nr (p) = {q X : d(p, q) < r} for some r > 0. The
number r is called the radius of Nr (p).
(b) A point p is a limit point of the set E if every neighborhood of p contains a point
q 6= p such that q E.
(c) If p E and p is not a limit point of E, then p is called an isolated point of E.
(d) E is closed if every limit point of E is a point of E.
(e) A point p is an interior point of E if there is a neighborhood N of p such that
N E.
(f) E is open if every point of E is an interior point of E.
(g) The complement of E (denoted by E c ) is the set {p X : p 6 E}.
(h) E is perfect if E is closed and if every point of E is a limit point of E.
(i) E is bounded if there is a real number M and a point q X such that d(p, q) < M
for all p E.
(j) E is dense in X if every point of X is a limit point of E, or a point of E (or both).
Let us note that in R1 neighborhoods are segments, whereas in R2 neighborhoods are
interiors of circles.
Example. The empty set is open and closed. Why?
2.19 Theorem. Every neighborhood is an open set.
Proof. Consider a neighborhood E = Nr (p), and let q E. Then there exists h > 0 such
that
d(p, q) = r h.
For all points s such that d(q, s) < h, we then have
d(p, s) 6 d(p, q) + d(q, s) < r h + h = r,

so that s E. Thus q is an interior point of E.

2.20 Theorem. If p is a limit point of a set E, then every neighborhood of p contains

infinitely many points of E.
Proof. Suppose there is a neighborhood N of p which contains only a finite number of
points of E. Let q1 , . . . , qn be those points of N E, which are distinct from p, and put
r = min d(p, qm ).
16m6n

Then r > 0. The neighborhood Nr (p) contains no point q of E such that q 6= p, which
contradicts p being a limit point of E.

Corollary. A finite point set has no limit points.

44

2.21 Examples. Consider the following subsets of R2 :

(a) {z C : |z| < 1}.
(b) {z C : |z| 6 1}.
(c) A nonempty finite set.
(d) Z, the set of all integers.
(e) { n1 : n = 1, 2, 3, . . .}.
(f) C.
(g) The segment (a, b).
Note that (d), (e), (g) can be regarded also as subsets of R. Some properties of these sets
are tabulated below:
Closed Open Perfect Bounded
(a)
No
Yes
No
Yes
(b)
Yes
No
Yes
Yes
(c)
Yes
No
No
Yes
(d)
Yes
No
No
No
(e)
No
No
No
Yes
(f)
Yes
Yes
Yes
No
(g)
No
No
Yes
In (g), we left the second entry blank. The reason is that the segment (a, b) is not open
if we regard it as a subset of R2 , but it is an open subset of R.
2.23 Theorem. A set E is open if and only if its complement is closed.
Proof. suppose E c is closed. Choose x E. Then x 6 E c , and x is not a limit point
of E c . Hence there exists a neighborhood N of x such that E c N = , that is, N E.
Thus x is an interior point of E, and E is open.
Suppose E is open. Let x be a limit point of E c . Then every neighborhood of x
contains a point of E c , so that x is not an interior point of E. Since E is open, this means
that x E c . It follows that E c is closed.

Corollary. A set F is closed if and only if its complement is open. In particular, every
metric space X is open and closed.
2.24 Theorem.
S
(a) For any collection {G } of open sets, T G is open.
(b) For any collection {F } of closed sets, F is closed.
T
(c) For any finite collection G1 , . . . , Gn of open sets, Sni=1 Gi is open,
(d) For any finite collection F1 , . . . , Fn of closed sets, ni=1 Fi is closed.
S
Proof. Put G = G . If x G, then x G for some . Since x is an interior point of
G , x is also an interior point of G, and G is open. This proves (a).
By Theorem 2.22,
!c
[
\
Fc ,
(21)
F =

45

T
and Fc is open, by Theorem 2.23. Hence (a) implies that (21) is open so that F is
closed.
T
Next, put H = ni=1 Gi . For any x H, there exist neighborhoods Ni of x, with radii
ri such that Ni Gi (i = 1, 2, . . . , n). Put
r = min(r1 , . . . , rn ),

and let N be the neighborhood of x of radius r. Then N Gi for i = 1, 2, . . . , n, so that

N H, thus proving H is open.
By taking complements, (d) follows from (c):
!c
n
n
\
[
Fic .
Fi =
i=1

i=1

2.25 Examples. In parts (c) and (d) of the preceding theorem, the finiteness of the
collections is essential.
For let Gn = ( n1 , n1 ), (n = 1, 2, 3, . . .), which is an open subset
T
of R. Put G = n=1 Gn . Then G = {0}, which is not open in R. Thus the intersection
of an infinite collection of open sets need not be open. Similarly, the union of an infinite
collection of closed sets need not be closed.
2.26 Definition. If X is a metric space, if E X, and if E denotes the set of all limit
points of E in X, then the closure of E is the set E = E E .
2.27 Theorem. If X is a metric space and E X, then
(a) E is closed,
(b) E = E if and only if E is closed,
(c) E F for every closed set F X such that E F .
By (a) and (c), E is the smallest closed subset of X containing E.
then p is neither a point of E nor a limit point of E. Hence
Proof. (a) If p X and p 6 E,
The complement of E is therefore open.
p has a neighborhood which does not intersect E.
Hence E is closed.
(a) implies that E is closed. If E is closed, then E E, hence E = E.
(b) If E = E,

(c) If F is closed and E F , then F F , hence E F . Thus E F .

2.28 Theorem. Let 6= E R which is bounded above. Let y = sup E. Then y E.

Hence y E if E is closed.

Assume y 6 E. For every h > 0, there exists x E such

Proof. If y E then y E.
that y h < x < y, for otherwise y h would be an upper bound of E. Thus y is a limit

point of E. Hence y E.


46

2.29 Remark. Suppose E Y X, where X is a metric space. Saying that E is an

open subset of X means that to each point p E there is associated a positive number r
such that the conditions d(p, q) < r, q X imply that q E.
Since Y is also a metric space, our definitions may equally well be made within Y . The
set E is open relative to Y if to each p E there is associated an r > 0 such that q E
whenever d(p, q) < r and q Y .
Example 2.21(g) showed that a set may be open relative to Y without being open relative
to X.
2.30 Theorem. Suppose Y X. A subset E of Y is open relative to Y if and only if
E = Y G for some open subset G of X.

Proof. Suppose E is open relative to Y . To each p E there is a positive number

rp such that the conditions d(p, q) < rp , q Y imply that q E. Let Vp be the set of all
q X such that d(p, q) < rp , and define
[
G=
Vp .
pE

Then G is an open subset of X, by Theorems 2.19 and 2.24.

Since p Vp for all p E, it is clear that E G Y .

By our choice of Vp , we have Vp Y E for every p E, so that G Y E. Thus

E =GY.
If G is open in X and E = G Y , every p E has a neighborhood Vp G. Then
Vp Y E, so that E is open relative to Y .

Example. Let X = R, Y = [0, 1], E = (0, 1] and G = (0, 2). Then E is neither open nor
close in R. However, since G is open in X and E = G Y , E is open in Y .

47

Compact sets
2.31 Definition. By an open cover of a set
S E in a metric space X we mean a collection
{G } of open subsets of X such that E G .

2.32 Definition. A subset K of a metric space X is said to be compact if every open

cover of K contains a finite subcover. In other words, if {G } is an open cover of K, then
there are finitely many indices 1 , . . . , n such that
K G 1 G n .

Let X be a metric space and E a subset of X. Open covers of E always exists. For a
fixed r > 0, {Nr (x) : x E} is an open cover of E because for each x E, {x} Nr (x)
so that
[
[
E=
{x}
Nr (x).
xE

xE

If E is finite, then {Nr (x) : x E} is a finite subcover, hence E is compact. If E is infinite

and is compact, then there is a finite subcollection of {Nr (x) : x E} which covers E.
2.33 Theorem. Suppose K Y X. Then K is compact relative to X if and only if
K is compact relative to Y .
S
Proof. Let {V } be a collection of sets, open relative to Y , such that K V .
By theorem 2.30, there are sets G , open relative to X, such that V = Y G , for all ;
and since K is compact relative to X, we have
K G 1 G n .

(22)

K V 1 V n .

(23)

for some choice of finitely many indices 1 , . . . , n . Since K Y , (22) implies

This proves that K is compact relative to Y .

Let {G } be a collection of open subsets of X which covers K, and put V = Y G .

Then (23) will hold for some choice of 1 , . . . , n ; and since V G , (23) implies (22). 
2.34 Theorem. Compact subsets of metric spaces are closed.
Proof. Let K be a compact subset of a metric space X. We shall prove that the complement of K is an open subset of X.
Suppose p X \K. If q K, let Vq and Wq be neighborhoods of p and q, respectively, of
radius less than 21 d(p, q). Since K is compact, there are finitely many points q1 , . . . , qn K
such that
K Wq1 Wqn = W.
If V = Vq1 Vqn , then K is a neighborhood of p which does not intersect W . Hence
V K c , so that p is an interior point of K c . The theorem follows.

2.35 Theorem. Closed subsets of compact sets are compact.

48

Proof. Suppose F K X, F is closed (relative to X), and K is compact. Let {V } be

an open cover of F . If F c is adjoined to {V }, we obtain an open cover of K. Since K
is compact, there is a finite subcollection of which covers K, and hence F . If F c is
a member of , we may remove it from and still retain an open cover of F . We have
thus shown that a finite subcollection of {V } covers F .

Corollary. If F is closed and K is compact, then F K is compact.
Proof. Theorems 2.24(b) and 2.34 show that F K is closed; since F K K, Theorem
2.35 shows that F K is compact.

2.36 Theorem. (Finite intersection property) If {K } is a collection of compact subsets
of a metric space
T X such that the intersection of every finite subcollection of {K } is
nonempty, then K is nonempty.

Proof. Fix a member K1 of {K } and put G = Kc . Assume that no point of K1 belongs

to every K . Then the sets G form an open cover of K1 ; and since K1 is compact, there
are finitely many indices 1 , . . . , n such that K1 G1 Gn . But this means that

which is a contradiction.

K1 K1 Kn = ,

Corollary. If {Kn } T
is a sequence of nonempty compact sets such that Kn Kn+1
(n = 1, 2, 3, . . .), then
n=1 Kn is not empty.
2.37 Theorem. If E is an infinite subset of a compact set K, then E has a limit point
in K.
Proof. If no point of K were a limit point of E, then each q K would have a neighborhood Vq which contains at most one point of E (namely, q, if q E). It is clear that no
finite subcollection of {Vq } can cover E; and the same is true of K, since E K. This
contradicts the compactness of K.

2.38 Theorem. (Nested interval property)
If {In } is a sequence of intervals in R such
T
that In In+1 (n = 1, 2, 3, . . .), then
I
n=1 n is not empty.

Proof. If In = [an , bn ], let E = {an : n = 1, 2, 3, . . .}. Then E is nonempty and bounded

above (by b1 ). Let x = sup E. If m and n are positive integers, then
an 6 am+n 6 bm+n 6 bn ,
so that x 6 bm for each m. Since it is obvious that am 6 x, we see that x Im for
m = 1, 2, 3, . . ..


2.39 Theorem. Let k be a positive

integer. If {In } is a sequence of k-cells such that
T
I
In In+1 (n = 1, 2, 3, . . .), then
n=1 n is not empty.

49

Proof. For n = 1, 2, 3, . . ., let In = {x = (x1 , . . . , xk ) : an,j 6 xj 6 bn,j for 1 6 j 6 k}, and

put In,j = [an,j , bn,j ]. For each j, the sequence {In,j } satisfies the hypotheses of Theorem
2.38. Hence there are real numbers xj (1 6 j 6 k) such that
an,j 6 xj 6 bn,j

(1 6 j 6 k; n = 1, 2, 3, . . .).

Setting x = (x1 , . . . , xk ), we see that x In for n = 1, 2, 3, . . .. The theorem follows. 

2.40 Theorem. Every k-cell is compact.
Proof. Let I be a k-cell, consisting of all points x = (x1 , . . . , xk )T such that aj 6 xj 6 bj
(1 6 j 6 k). Put
X
1/2
k
2
=
(bj aj )
.
j=1

Then kx yk2 6 , if x I, y I.

Suppose, on the contrary, that there exists an open cover {G } of I which contains no
a +b
finite subcover of I. Put cj = j 2 j . The intervals [aj , cj ] and [cj , bj ] then determine 2k
k-cells Qi whose union is I.
At least one of these sets Qi , call it I1 , cannot be covered by any finite subcollection of
{G } (otherwise I could be so covered). We next subdivide I1 and continue the process.
We obtain a sequence {In } with the following properties:
(a) I I1 I2 I3 ;
(b) In is not covered by any finite subcollection of {G };
(c) if x In and y In , then kx yk2 6 2n .
By (a) and Theorem 2.39, there is a point x which lies in every In . For some , x G .
Since G is open, there exists r > 0 such that ky x k2 < r implies that y G . If n is
so large that 2n < r, then (c) implies that In G , which contradicts (b).
This completes the proof.

50

The equivalence of (a) and (b) in the next theorem is known as the Heine-Borel theorem.
2.41 Theorem. If a set E in Rk has one of the following three properties, then it has
the other two:
(a) E is closed and bounded.
(b) E is compact.
(c) Every infinite subset of E has a limit point in E.
Proof. If (a) holds, then E I for some k-cell I, and (b) follows from Theorems 2.40 and
2.35. Theorem 2.37 shows that (b) implies (c). It remains to show that (c) implies (a).
If E is not bounded, then E contains points xn such that
kxn k2 > n (n = 1, 2, 3, . . .).

The set S = {xn : n = 1, 2, 3, . . .} is infinite and clearly has no limit point in Rk , hence
has none in E. Thus (c) implies that E is bounded.
If E is not closed, then there is a point x0 Rk which is a limit point of E but not a
point of E. For n = 1, 2, 3, . . ., there are points xn E such that kxn x0 k2 < n1 . Let
S = {xn : n = 1, 2, 3, . . .}. Then S is infinite (otherwise kxn x0 k2 would have a constant
positive value, for infinitely many n), S has x0 as the only limit point in Rk . For if y Rk ,
y 6= x0 , then
kxn yk2 > kx0 yk2 kxn x0 k2
1
> kx0 yk2
n
1
> kx0 yk2
2
for all but finitely many n; this shows that y is not a limit point of S (Theorem 2.20).
Thus, S has no limit point in E; hence E must be closed if (c) holds.

Remark. (b) and (c) are equivalent in any metric space. If the underlying metric space
is finite dimensional (e.g., Rk ), then (a) and (b) are equivalent. In general, a subset of a
metric space is compact if and only if it is complete and totally bounded; this equivalence
holds should the dimension of the metric space be finite or not.
2.42 Theorem (Weierstrass). Every bounded infinite subset of Rk has a limit point
in Rk .
Proof. Let E be a bounded infinite subset of Rk . Then there exists a k-cell I Rk such
that E I. By Theorem 2.40, I is compact, and so E has a limit point in I by Theorem
2.37.


51

Perfect sets
2.43 Theorem. Let P be a nonempty perfect set in Rk . Then P is uncountable.
Proof. Since P has limit points, P must be infinite. Suppose P is countable, and denote
the points of P by x1 , x2 , x3 , . . .. We shall construct a sequence {Vn } of neighborhoods,
as follows.
Let
V1 = {y Rk : ky x1 k2 < r}, and V1 = {y Rk : ky x1 k2 6 r}.
Then V1 is a neighborhood of x1 , and V1 is the closure of V1 .
Suppose Vn has been constructed, so that Vn P is not empty. Since every point of P is
a limit point of P , there is a neighborhood Vn+1 such that (i) Vn+1 Vn , (ii) xn 6 Vn+1 ,
(iii) Vn+1 P is not empty. By (iii), Vn+1 satisfies our induction hypothesis, and the
construction can proceed.

Put Kn = Vn P . Since
T Vn is closed and bounded, Vn is compact.
T Since xn 6 Kn+1 ,
no point of P lies in n=1 Kn . Since Kn P , this implies that n=1 Kn is empty. But
each Kn is nonempty, by (iii), and Kn Kn+1 , by (i); this contradicts the Corollary to
Theorem 2.36.


Corollary. Every interval [a, b] (a < b) is uncountable. In particular, the set of all real
numbers is uncountable.
2.44 The Cantor set. The set which we are now going to construct shows that there
exist perfect sets in R which contain no segment.
Let E0 = [0, 1]. Remove the segment ( 13 , 32 ), and let E1 be the union of the intervals
1
2
[0, ], [ , 1].
3
3
Remove the middle thirds of these intervals, and let E2 be the union of the intervals
1
2 3
6 7
8
[0, ], [ , ], [ , ], [ , 1].
9
9 9
9 9
9
Continuing in this way, we obtain a sequence of compact sets En , such that
(a) E1 E2 E3 ;
(b) En is the union of 2n intervals, each of length 3n .

52

The set
P =

En

n=1

is called the Cantor set. P is clearly compact, and Theorem 2.36 shows that P is not
empty.
No segment of the form
3k + 1 3k + 2
),
(24)
( m ,
3
3m
where k and m are positive integers, has a point in common with P . Since every segment
(, ) contains a segment of the form (24), if

,
3m <
6
P contains no segment.
To show that P is perfect, it is enough to show that P contains no isolated point. Let
x P , and let S be any segment containing x. Let In be that interval of En which
contains x. Choose n large enough, so that In S. Let xn be an endpoint of In such
that xn 6= x.
It follows from the construction of P that xn P . Hence, x is a limit point of P , and P
is perfect.
One of the most interesting properties of the Cantor set is that it provides us with an
example of an uncountable set of measure zero.
P
P
1
1
2 n
n1 1
The length of the removed middle-thirds =
( 3n ) = 13
n=1 2
n=0 ( 3 ) = ( 3 )( 1 2 ) = 1.
3
Thus, the Lebesgue measure (the length) of the Cantor set = 1 1 = 0.

53

Connected sets
2.45 Definition. Two subsets A and B of a metric space X are said to be separated if
= and A B = . A set E X is said to be connected if E is not a union
both A B
of two nonempty separated sets.
2.46 Remark. Separated sets are disjoint, but disjoint sets need not be separated. For
example, the interval [0, 1] and the segment (1, 2) are not separated, since 1 is a limit
point of (1, 2). However, the segments (0, 1) and (1, 2) are separated.
Connected subsets of the line have a particularly simple structure:
2.47 Theorem. A subset E of R1 is connected if and only if it has the following property:
If x, y E, and x < z < y, then z E.

Proof. If there exist x, y E, and some z (x, y) such that z 6 E, then E = Az Bz ,

where
Az = E (, z), Bz = E (z, ).
Since x Az and y Bz , A 6= and B 6= . Since Az (, z) and Bz (z, ), they
are separated. Hence E is not connected.
Suppose E is not connected. Then there exist nonempty separated sets A and B
such that A B = E. Pick x A, y B and assume (without loss of generality) that
x < y. Define
z = sup(A [x, y]).
hence z 6 B. In particular, x 6 z < y. If z 6 A, it follows that
By Theorem 2.28, z A;
hence there exists z1 such that z < z1 < y
x < z < y and z 6 E. If z A, then z 6 B,
and z1 6 B. Then x < z1 < y and z1 6 E.


54

Numerical sequences and series

3.1 Definition. A sequence {pn } in a metric space (X, d) is said to converge if there is
a point p X with the following property: For every > 0, there is an integer N such
that n > N implies that d(pn , p) < .
In this case we also say that {pn } converges to p, or that p is the limit of {pn }, and we
write pn p, or lim pn = p.
n

If {pn } does not converge, it is said to diverge.

The convergence of {pn } depends not only on the sequence itself but also on X. For
instance, the sequence { n1 } converges in (R, d) (to 0), but fails to converge in (R+ , d) with
the usual metric d(x, y) = |x y|.
The sequence {pn } is said to be bounded if its set of all terms {pn : n = 1, 2, 3 . . .} is
bounded.
Examples. Consider the following sequences of complex numbers, that is, X = R2 .
(a) If sn = n1 , then lim sn = 0 and the sequence is bounded.
n

(b)
(c)
(d)
(e)

If
If
If
If

sn
sn
sn
sn

= n2 , the sequence {sn } is unbounded and divergent.

n
= 1 + (1)
, the sequence {sn } converges to 1 and is bounded.
n
n
= i , the sequence {sn } is bounded but divergent.
= 1 for all n, then {sn } converges to 1 and is bounded.

3.2 Theorem. Let {pn } be a sequence in a metric space (X, d).

(a) {pn } converges to p X if and only if every neighborhood of p contains pn for all
but finitely many n.
(b) (Uniqueness of limit) If {pn } converges to p X and to p X, then p = p .
(c) If {pn } converges, then {pn } is bounded.
(d) If E X and if p is a limit point of E, then there is a sequence {pn } in E such
that p = lim pn .
n

Proof. (a) Let V be a neighborhood of p. There exists > 0 such that for q X
and d(q, p) < , q V . There exists N such that n > N implies d(pn , p) < . Thus,
n > N implies pn V .

Fix > 0, and let V = {q X : d(p, q) < }. By assumption, there exists N

(corresponding to this V ) such that pn V if n > N . Thus d(pn , p) < if n > N ; hence
pn p.

(b) Let > 0 be given. There exist integers N, N such that n > N implies d(pn , p) <
and n > N implies d(pn , p ) < 2 . Hence, if n = max(N, N ), we have

d(p, p ) 6 d(p, pn ) + d(pn , p ) < + = .
2 2

Since is arbitrary, we conclude that d(p, p ) = 0.

55

(c) Since pn p, there is an integer N such that n > N implies d(pn , p) < 1. Put
r = max(1, d(p1 , p), . . . , d(pN 1 , p)).

Then d(pn , p) 6 r for n = 1, 2, 3, . . ..

(d) For each positive integer n, there is a point pn E such that d(pn , p) < n1 . Given
> 0, choose N so that N > 1. If n > N , it follows that d(pn , p) < n1 < N1 < . Hence
pn p.

3.3 Theorem. Suppose {sn }, {tn } are complex sequences, and lim sn = s, lim tn = t.
n
n
Then
(a) lim (sn + tn ) = s + t;
n

(b) lim csn = cs, lim (c + sn ) = c + s, for any number c.

n

(c) lim sn tn = st;

n

1
n sn

(d) lim

= 1s , provided sn 6= 0 for all n and s 6= 0.

Proof. (a) Given > 0. There exist integers N1 , N2 such that n > N1 implies |sn s| < 2
and n > N2 implies |tn t| < 2 . If N = max(N1 , N2 ), then n > N implies

|(sn + tn ) (s + t)| 6 |sn s| + |tn t| < + = .
2 2

(b) Given > 0. There exists an integer N such that n > N implies |sn s| < |c|+1
,
hence



< .
|csn cs| = |c||sn s| 6 |c|
|c| + 1
The convergence c + sn c + s follows from (a) by setting tn = c for all n.

(c) The following identity holds:

sn tn st = (sn s)(tn t) + s(tn t) + t(sn s).

Given > 0. There are

integers N1 , N2 such that n > N1 implies |sn s| <
n > N2 implies |tn t| < . If we take N = max(N1 , N2 ), then n > N implies

(1)
and

|(sn s)(tn t)| <

so that lim (sn s)(tn t) = 0. We now apply (a) and (b) to (1), and conclude that
n

lim (sn tn st) = 0.

(d) Choose m such that |sn s| < 12 |s| if n > m. By the triangle inequality, we have
1
1
|s| |sn | 6 |sn s| < |s| |s| < |sn | when n > m.
2
2
Given > 0. There is an integer N > m such that n > N implies
1
|sn s| < |s|2 .
2
Hence, for n > N ,


1

1
  = |sn s| < 2|sn s| < .
 sn s 
|sn ||s|
|s|2

56

3.4 Theorem.
(a) Suppose xn = (x1,n , . . . , xk,n )T Rk (n = 1, 2, 3, . . .). Then {xn } converges to
x = (x1 , . . . , xk )T if and only if
lim xj,n = xj

(1 6 j 6 k).

(2)

(b) Suppose {xn }, {yn } are sequences in Rk , {n } is a sequence of real numbers, and
xn x, yn y, n . Then
lim (xn + yn ) = x + y,

lim xn yn = x y,

lim n xn = x.

Proof. (a) We first note that for j = 1, 2, . . . , k,

1/2
X
k
2
|xj,n xj | 6
|xj,n xj |
= kxn xk2 .
j=1

Since xn x, for j = 1, 2, . . . , k,
|xj,n xj | 6 kxn xk2 0 as n ,
i.e., (2) holds.
To each > 0 there corresponds an integer N such that n > N implies

(1 6 j 6 k).
|xj,n xj | <
k
Hence n > N implies
kxn xk2 =

X
k
j=1

|xj,n xj |

1/2

<

X
k 
j=1

2 1/2

= ,

so that xn x.
(b) follows from (a) and Theorem 3.3.

Subsequences
3.5 Definition. Given a sequence {pn }, consider a sequence {nk } of positive integers,
such that n1 < n2 < n3 < . Then the sequence {pni } is called a subsequence of {pn }.
If {pni } converges, its limit is called a subsequential limit of {pn }.
It is clear that {pn } converges to p every subsequence of {pn } converges to p.
3.6 Theorem.
(a) If {pn } is a sequence in a compact metric space X, then some subsequence of {pn }
converges to a point of X.
(b) Every bounded sequence in Rk contains a convergent subsequence.

57

Proof. (a) Let E be the set of all terms of {pn }. If E is finite then there is a p E and a
sequence {ni } with n1 < n2 < n3 < such that
pn1 = pn2 = pn3 = .

The subsequence {pni } so obtained converges evidently to p.

If E is infinite, Theorem 2.37 shows that E has a limit point p X. Choose n1 so that
d(p, pn1 ) < 1. Having chosen n1 , . . . , ni1 we see from Theorem 2.20 that there is an
integer ni > ni1 such that d(p, pni ) < 1i . Then {pni } converges to p.
(b) This follows from (a), since Theorem 2.41 implies that every bounded subset of Rk
lies in a compact subset of Rk .

3.7 Theorem. The subsequential limits of a sequence {pn } in a metric space X form a
closed subset of X.
Proof. Let E be the set of all subsequential limits of {pn } and let q be a limit point of
E . We have to show that q E . Choose n1 so that pn1 6= q. (If no such n1 exists,
then E has only one point, and there is nothing to prove.) Put = d(q, pn1 ). Suppose
n1 , . . . , ni1 are chosen. Since q is a limit point of E , there is an x E with d(x, q) < 2i .
Since x E , there is an ni > ni1 such that d(x, pni ) < 2i . Thus
d(q, pni ) 6

2i1

for i = 1, 2, 3, . . .. This says that {pni } converges to q. Hence q E .

Cauchy sequences
3.8 Definition. A sequence {pn } in a metric space X is said to be a Cauchy sequence if
for every > 0, there is an integer N such that d(pn , pm ) < if m, n > N .
3.9 Definition. Let E be a nonempty subset of a metric space X, and let
S = {d(p, q) : p, q E}.

sup S is called the diameter of E.

If {pn } is a sequence in X and if EN = {pN , pN +1 , pN +2 , . . .}, it is clear from the two

preceding definitions that {pn } is a Cauchy sequence if and only if
lim diam EN = 0.

3.10 Theorem.
(a) Let E be a subset of a metric space X, then diam E = diam E.
(b) If Kn is a sequence of compact sets in X such that Kn Kn+1 (n = 1, 2, 3, . . .)
and if
lim diam Kn = 0,
n
T
then n=1 Kn consists of exactly one point.

58

it is clear that
Proof. (a) Since E E,

diam E 6 diam E.

By the definition of E,
there are points p , q E such
Fix > 0, and choose p, q E.
that d(p, p ) < , d(q, q ) < . Hence
d(p, q) < d(p, p ) + d(p , q ) + d(q , q)
< + d(p , q ) + 6 2 + diam E.
It follows that
diam E 6 2 + diam E,
and since was arbitrary, (a) is proved.
T
(b) Put K =
n=1 Kn . By Theorem 2.36, K is not empty. If K contains more than one
point, then diam K > 0. But for each n, Kn K so that diam Kn > diam K. This
contradicts the assumption that diam Kn 0.

3.11 Theorem.
(a) In any metric space X, every convergent sequence is a Cauchy sequence.
(b) If X is a compact metric space and if {pn } is a Cauchy sequence in X, then {pn }
converges to some point of X.
(c) In Rk , every Cauchy sequence converges.
The fact (contained in Theorem 3.11) that a sequence converges in Rk if and only if it is
a Cauchy sequence is usually called the Cauchy criterion for convergence.
Proof. (a) Let pn p and let , there is an integer N such that d(p, pn ) < for all n > N .
Hence,
d(pn , pm ) 6 d(pn , p) + d(p, pm ) < 2
as soon as m, n > N . Thus {pn } is a Cauchy sequence.
(b) Let {pn } be a Cauchy sequence in the compact space X. For N = 1, 2, 3, . . ., let
EN = {pN , pN +1 , pN +2 , . . .}. Then
lim diam EN = 0
(3)
N

by Definition 3.9 and Theorem 3.10(a). Being a closed subset of the compact space X,
each EN is compact (Theorem 2.35). Also EN EN +1 so that EN EN +1 .
Theorem 3.10(b) shows now that there is a unique p X which lies in every EN . Let
> 0 be given. By (3) there is an integer N0 such that diam EN < if N > N0 . Since
p EN , it follows that d(p, q) < for every q EN , hence for every q EN . In other
words, d(p, pn ) < if n > N0 . This says precisely that pn p.
(c) Let {xn } be a Cauchy sequence in Rk . Define EN as in (b), with xi in place of pi . For
some N , diam EN < 1. The set of all terms of {xn } is the union EN {x1 , . . . , xN 1 }.
Hence {xn } is bounded. Since every bounded subset of Rk has compact closure in Rk
(Theorem 2.41), (c) follows from (b).


59

3.12 Definition. A metric space in which every Cauchy sequence converges is said to be
complete.
Theorem 3.11 says that all compact metric spaces and all Euclidean spaces are complete.
Theorem 3.11 implies also that every closed subset E of a complete metric space X is
complete. (Every Cauchy sequence in E is a Cauchy sequence in X, hence it converges
to some p X, and actually p E since E is closed.)
An example of a metric space which is not complete is the space of all rational numbers,
with d(x, y) = |x y|.
3.13 Definition. A sequence {sn } of real numbers is said to be
(a) monotonically increasing if sn 6 sn+1 (n = 1, 2, 3, . . .);
(b) monotonically decreasing if sn > sn+1 (n = 1, 2, 3, . . .).
The class of monotonic sequences consists of the increasing and the decreasing sequences.
3.14 Theorem. Suppose that {sn } is monotone. Then {sn } converges if and only if it is
bounded.
Proof. Suppose sn 6 sn+1 (the proof is analogous in the other case). Let E be the set of
all terms of {sn }.
This follows from Theorem 3.2(c).
Let s = sup E. Then
sn 6 s (n = 1, 2, 3, . . .).
For every > 0, there is an integer N such that
Since {sn } is increasing, for n > N ,

s < sN 6 s.

s < sN 6 sn 6 s |sn s| < ,

which shows that {sn } converges to s.

Example. Let a > 0 and let the sequence {yn } be defined by y1 =

for n > 2. Prove that
(a) {yn } is increasing,
(b) lim yn exists and find it.

a, yn =

Solution. (a) By induction on n. When n = 1,

a
y2 y1 = a + a a = p

> 0.
a+ a+ a
It is true for n = 1.
Assume that the result is true for n = k, i.e., yk+1 yk > 0. Then

yk+1 yk

> 0.
yk+2 yk+1 = a + yk+1 a + yk =
a + yk+1 + a + yk

a + yn1

60

It is also true for n = k + 1.

By the principle of mathematical induction, it is true for all n N.
(b) For n > 2, since yn1 6 yn , we have
yn2 = a + yn1 6 a + yn
so that yn2 yn a 6 0, which holds if and only if

1 + 1 + 4a
1 1 + 4a
6 yn 6
.
2
2
Since {yn } is increasing and bounded above, {yn } is convergent. Let lim yn = . Then
n

lim

yn2

= lim (a + yn1 )
n

=a+
0 = 2 a

1 + 1 + 4a
=
2
since {yn } is a sequence of positive numbers.

Example. Given two positive real numbers a and b such that a < b. Let {xn } and {yn }
be two sequences satisfying x1 = a, y1 = b and
p
xn + yn
xn+1 = xn yn , yn+1 =
2
for all positive integers n. Show that both {xn } and {yn } converge and lim xn = lim yn .
n
n

You may use the fact that for any u, v R+ , uv 6 u+v

.
2
Solution. It is clear that {xn } and {yn } are positive sequences such that x1 = a < b = y1 ,
and that for n > 0,
p
xn + yn
xn+1 = xn yn 6
= yn+1 .
2
Since yn > xn for each n > 1, we have
p
p
xn+1 = xn yn > xn xn = xn ,
yn + yn
xn + yn
6
= yn ,
yn+1 =
2
2
so that {xn } is an increasing sequence and {yn } is a decreasing sequence.
From above we have that
x1 6 x2 6 6 xn 6 yn 6 yn1 6 6 y2 6 y1 ,

so that {xn } is bounded above and {yn } is bounded below. Thus lim xn = A and
n
lim yn = B exist and
n

xn1 + yn1
n
2
A+B
B=
2
A = B.

lim yn = lim

61

Upper and lower limits

3.15 Definition. Let {sn } be a sequence of real numbers with the following property:
For every real M there is an integer N such that n > N implies sn > M . We then write
sn +.

Similarly, if for every real M there is an integer N such that n > N implies sn 6 M , we
write
sn .
3.16 Definition. Let {sn } be a sequence of real numbers. Let
: sn x for some subsequence {sn }}.
E = {x R
k
k

This set E contains all subsequential limits as defined in Definition 3.5, plus possibly the
numbers +, . Put
s = sup E, s = inf E,
called the limit superior and limit inferior of {sn }; we use the notation
lim sup sn = s ,
n

lim inf sn = s .
n

Simply speaking, lim sup sn and lim inf sn are, respectively, the largest limit point and the
smallest limit point of {sn }. Alternatively, lim sup sn and lim inf sn can be defined by
lim sup sn = inf sup sm ,
n

n m>n

lim inf sn = sup inf sm .

n

m>n

3.17 Theorem. Let {sn } be a sequence of real numbers. Let E and s be defined as in
Definition 3.16. Then s has the following two properties:
(a) s E.
(b) If x > s , there is an integer N such that n > N implies sn < x.
Moreover, s is the only number with the properties (a) and (b).
Proof. (a) If s = +, then E is not bounded above; hence {sn } is not bounded above,
and there is a subsequence {snk } such that snk +.
If s is real, then E is bounded above, and at least one subsequential limit exists, so
that (a) follows from Theorems 3.7 and 2.28. If s = , then E = {}, and there
is no subsequential limit. Hence, for any real M , sn > M for at most a finite number of
values of n, so that sn .
This establishes (a) in all cases.
(b) Suppose there is a number x > s such that sn > x for infinitely many values of n.
These sn form a subsequence of {sn } such that each sn > x. The limit of this subsequence
y E is such that y > x > s , contradicting the definition of s (what definition?).
Thus s satisfies (a) and (b).
To show the uniqueness, suppose there are two numbers, p and q, which satisfy (a) and
(b), and suppose p < q. Choose x such that p < x < q. Since p satisfies (b), we have
sn < x for n > N . But then q cannot satisfy (a).


62

Theorem. Let {sn } be a sequence of real numbers. Let E and s be defined as in

Definition 3.16. Then s has the following two properties:
(a) s E.
(b) If x < s , there is an integer N such that n > N implies sn > x.

Moreover, s is the only number with the properties (a) and (b).
Proof. Exercise.

3.18 Examples.
(a) Let {sn } be a sequence containing all rationals. Then every real number is a
subsequential limit, and
lim sup sn = +,

lim inf sn = .
n

(b) Let sn =

(1)n
1 .
1+ n

Then
lim sup sn = 1,

lim inf sn = 1.
n

(c) For a real-valued sequence {sn }, lim sn = s if and only if

n

lim sup sn = lim inf sn = s.

n

3.19 Theorem. If sn 6 tn for n > N , where N is fixed, then

lim inf sn 6 lim inf tn ,
n

lim sup sn 6 lim sup tn .

n

Proof. To illustrate the idea of the proof, we assume that all the lim sup and lim inf exist.
For m > n > N , we have
sm 6 tm 6 sup tm .
m>n

Since the right side is independent of m, taking sup of the left side over all m > n, we get
sup sm 6 sup tm .

m>n

m>n

Taking inf of the left side over all n, we have

lim sup sn = inf sup sm 6 sup tm .
n

n m>n

m>n

Since the left side is independent of n, taking the inf of the right side over all n, we finally
obtain
lim sup sn 6 inf sup tm = lim sup tn .
n

n m>n

The proof of lim inf sn 6 lim inf sn is left as an exercise.

n

63

Series
3.21
P Definition. Given a sequence {an }. The nth partial sum of the infinite series
n=1 an is
n
X
sn =
ak = a1 + a2 + + an .
k=1

If {sn } converges to s, we say that the series converges, and write

n

X
X
ak = s.
an = lim
n=1

k=1

The number s is called the sum of the series, which is the nth partial sum.
If {sn } diverges, the series is said to diverge.
The Cauchy criterion (Theorem 3.11) can be restated in the following form:

3.22 Theorem.
that

an converges if and only if for every > 0, there is an integer N such

 m

X 


ak  6
(6)



k=n

if m > n > N .
3.23 Theorem. If

an converges, then lim an = 0.

n

Proof. By taking m = n in Theorem 3.22, (6) becomes

|an | 6 (n > N ).

The condition lim an = 0 is necessary but not sufficient for the convergence of
n P
1
1
instance, the series
n=1 n diverges despite that lim n = 0.
n

an . For

3.24 Theorem. A series of non-negative terms converges if and only if its partial sums
form a bounded sequence.
P
Proof. Let
n=1 an be a series of non-negative terms. Then by Theorem 3.14,

X
n=1

an converges the monotone sequence sn =

n
X

ak is bounded.

k=1

3.25 Theorem. (Comparison test)

P
(a) If |anP
| 6 cn for n > N0 , where N0 is some fixed integer, and if
cn converges,
then
an converges.
P
P
(b) If an > dn > 0 for n > N0 , and if
dn diverges, then
an diverges.

64

Proof. Given > 0, there exists N > N0 such that m > n > N implies
m
X
ck < ,
k=n

by the Cauchy criterion. Hence


 m
m
m
X  X
X


|ak | 6
ck 6
ak  6



k=n

k=n

k=n

and (a) follows.

Next, (b) follows from (a), for if

an converges, so must

dn .

Example. Test the convergence of the following series:

X
X
X
1
n2 + 2
4n2 + n

(a)
.
,
(b)
,
(c)
3 7
3n + n
n4 + 5
n + n3
n=1
n=1
n=1
Solution. (a) Since for n > 1,

the series

X
n=1

3n

1
is convergent by the comparison test.
+n

(b) Since for n > 1,

the series

X
n2 + 2
n=1

1
1
< n and
+n
3

1
X 1
1
3
< ,
=
1 =
n
3
2
1 3
n=1

3n

n4 + 5

n2 + 2
n2 + 2
<
and by the p-test,
n4 + 5
n4

X
X
n2 + 2 X 1
2
=
+
< ,
n4
n2 n=1 n4
n=1
n=1
is convergent by the comparison test.

4n2 + n
4n2
4

(c) Since for n > 1,

>
=
and by the p-test,
3 7
3
3
n + n3
n7 + n7
2n1/3

X
4 X 1
4

= 3
= +,
3
2n1/3
2 n=1 n1/3
n=1

X
4n2 + n

the series
is divergent by the comparison test.
3 7
n + n3
n=1

65

The root and ratio tests

p
P
3.33 Theorem (Root Test). Given
an , put = lim sup n |an |. Then
n
P
(a) if < 1, P an converges;
(b) if > 1,
an diverges;
(c) if = 1, the test is inconclusive.

Proof. If < 1, we can choose so that < < 1, and an integer N such that
p
n
|an | <
for n > N [by Theorem 3.11(b)]. That is, n > N implies
|an | < n .

P n
P
Since 0 < < 1,
converges. Convergence of
an follows now from the comparison
test.
If > 1, then, again by Theorem 3.17, there is a sequence {nk } such that
q
nk
|ank | .

Hence |an | > 1 forP

infinitely many values of n, so that the condition an 0, necessary
for convergence of
an , does not hold (Theorem 3.23).
To prove (c), we consider the series

X
X
1
1
,
.
n
n2
n=1
n=1

For each of these series = 1, but the first diverges, the second converges.

P
3.34 Theorem (Ratio Test). The series
an


 an+1 
 < 1,
(a) converges if lim sup 

a
n
n




> 1 for all n > n0 , where n0 is some fixed integer.
(b) diverges if  an+1
an 


 an+1 
 < 1 holds, we can find < 1, and an integer N , such that
Proof. If lim sup 
an 
n


 an+1 


 an  <
for n > N . In particular,
|aN +1 | < |aN |,
|aN +2 | < |aN +1 | < 2 |aN |,
..
.

That is,

|aN +p | < p |aN |.

|an | < |aN | nN

P n
for n > N , and (a) follows from the comparison test, since
converges.
If |an+1 | > |an | for n > n0 , it is easily seen that an 6 0, and (b) follows.

66

= 1 implies nothing about the convergence of

Note: lim an+1
P 1 an
.
and
n2

an , as exemplified by

1
n

3.35 Examples.
(a) Consider the series
1 1
1
1
1
1
1
1
+ + 2 + 2 + 3 + 3 + 4 + 4 + ,
2 3 2
3
2
3
2
3
for which
 n
2
an+1
= lim
= 0,
lim inf
n 3
n
an
r

1
2n 1
n

lim inf an = lim

,
=
n
n
3n
3
r

1
2n 1
lim sup n an = lim
= ,
n
n
2
n
2
 n
1 3
an+1
= lim
= +.
lim sup
n 2 2
an
n

The root test indicates convergence; the ratio test does not apply.
(b) The same is true for the series

X
n=1

where

an =

1
1 1
1
1
1
1
+1+ + +
+
+
+
+ ,
2
8 4 32 16 128 64

lim inf
n

1
an+1
= ,
an
8

but
lim

and

lim sup
n

an+1
= 2,
an

1
an = .
2

3.37 Theorem. For any sequence {cn } of positive numbers,

cn+1
lim inf
6 lim inf n cn ,
n
n
cn

cn+1
lim sup n cn 6 lim sup
.
cn
n
n
Proof. We shall prove the second inequality; the proof of the first is quite similar. Put
cn+1
.
= lim sup
cn
n
If = +, there is nothing to prove. If is finite, choose > . There is an integer N
such that
cn+1
6
cn
for n > N . In particular, for any p > 0,
cN +p 6 cN +k

(k = 0, 1, . . . , p 1).

67

Multiplying these inequalities, we obtain

cN +p 6 p cN ,
or
cn 6 cN nN
Hence

cn 6

so that
lim sup

(n > N ).

p
n
cN N

cn 6 ,

(18)

by Theorem 3.20(b). Since (18) is true for every > , we have

lim sup n cn 6 .

It follows from Theorem 3.37 that if the ratio test works, then the root test works; if the
ratio test fails, the root test may work. It is in this sense that the root test is superior
than the ratio test.

Integral test
Let

n=1

an be a series of positive terms. Since

an =

n=1

N
1
X

an +

n=1

an ,

n=N

P
where the first sum on the right
is
finite,
a
converges
if
and
only
if
n
n=1
n=N an
P
converges, where N > 0. For
a
to
converge,
necessarily
lim
a
=
0.
So, an
n
n=1 n
n
must decrease to zero from some N onwards. Let f : N R be the function such that
an = f (n) for n = 1, 2, 3, . . ..
y
y = f (x)
aN

N +1

N +2

N +3

...

n+1

68

Interpreting aN as the area of the first rectangle, etc., the following inequalities are immediate:
Z

X
X
an 6
f (x) dx 6
an .
n=N +1

n=N

Thus,

n=N +1

an converges

Example. (p-test) Show that

1
n=1 np

f (x) dx converges.

converges if and only if p > 1.

Solution. Since f (x) = x1p > 0 for all x > 1 and is decreasing, by the integral test, we
have
Z

X
1
1
converges
dx converges.
p
p
n
x
1
n=1
Since

we have

1
dx = lim
R
xp

R
1

1
xp

R
1

 p+1

 p+1 R
R
1
x
1
,
= lim

dx = lim
R p + 1
R p + 1
xp
p + 1
1

dx < p + 1 < 0, i.e., p > 1.

Example. Test the convergence of the series

1
k=3 k2 3k+2 .

Solution. Since x2 3x + 2 = (x 2)(x 1) > 0 for x > 3, the series

sum of positive terms. Since
Z
Z R
dx
dx
= lim
2
x 3x + 2 R 3 (x 2)(x 1)
3

Z R
1
1

dx
= lim
R 3
x2 x1
h x 2 iR
= lim ln
R
x1 3
32
R2
ln
= lim ln
R
R1
31
= ln 2 < ,
P
1
the series
k=3 k2 3k+2 converges by the integral test.

1
k=3 k2 3k+2

is a

Power series
3.38 Definition. Given a sequence {cn } of complex numbers, the series

cn z n

(19)

n=0

is called a power series. The numbers cn are called the coefficients of the series; z is a
complex number.

69

Every power series there is associated a circle, the circle of convergence, such that (19)
converges if z is in the interior of the circle and diverges if z is in the exterior.
cn z n , put
p
1
= lim sup n |cn |, R = .

n
P
(If = 0, then R = +; if = +, R = 0.) Then
cn z n converges if |z| < R, and
diverges if |z| > R.
3.39 Theorem. Given the power series

Proof. Put an = cn z n and apply the root test:

p
p
|z|
.
lim sup n |an | = |z| lim sup n |cn | =
R
n
n
P
The series
cnP
z n converges |z|
< 1 |z| < R. R is called the radius of
R
n
convergence of
cn z .


3.40 Examples.
P
n n
(a) The series P
n=1 nn z has R = 0.
z
(b) The series
n=0 n! has R = +. (In this case the ratio test is easier to apply
than the root
P test.)n
n
(c) The series
n=0 z has R = 1. If |z| = 1, the series diverges, since z 6 0 as
n . P
zn
(d) The series
n=1 n2 has R = 1. It converges for all z with |z| = 1, by the comparison test.
Absolute convergence
P
P
Definition. The series
an is said to converge absolutely if the series
|an | converges.
P
P
3.45 Theorem. If
an converges absolutely, then
an converges.
Proof. The assertion follows from the inequality
 m

m
X  X


ak  6
|ak |



k=n

k=n

plus the Cauchy criterion.

3.46 Remarks. For series of positive terms, absolute convergence is the same as convergence.
P
P
P
If
an converges, but
|an | diverges, we say that
an converges non-nonabsolutely or
conditionally. For instance, the series

X
(1)n
n=1

converges nonabsolutely (Theorem 3.43).

70

Addition and multiplication of series

P
P
P
P
3.47 Theorem. If an = A, and bn = B, then (an + bn ) = A + B, and can = cA
for any fixed c.

Proof. Let

An =

n
X

ak

and Bn =

k=1

Then

n
X

bk .

k=1

An + Bn =

n
X

(ak + bk ).

k=1

Since lim An = A and lim Bn = B, we see that

n

lim (An + Bn ) = A + B,

and

lim cAn = c lim An = cA.


Thus, two convergent series may be added term by term, and the resulting series converges
to the sum of the two series.
P
P
3.48 Definition. Given
an and
bn , we put
n
X
cn =
ak bnk (n = 0, 1, 2, . . .)
k=0

and call

cn the Cauchy product of the two given series.

P
This
definition
may
be
motivated
as
follows.
If
we
take
two
power
series
an z n and
P
n
bn z , multiply them term by term, and collect terms containing the same power of z,
we get

X
X
n
an z
bn z n = (a0 + a1 z + a2 z 2 + )(b0 + b1 z + b2 z 2 + )
n=1

n=1

= a0 b0 + (a0 b1 + a1 b0 )z + (a0 b2 + a1 b1 + a2 b0 )z 2 +

= c0 + c1 z + c2 z 2 + .

Setting z = 1, we arrive at the above definition.

3.50 Theorem (Mertens) Suppose

P
(a) P
n=0 an converges absolutely,
(b) P
n=0 an = A,
(c)
bn = B,
n=0P
(d) cn = nk=0 ak bnk , (n = 0, 1, 2, . . .).
P
Then
n=0 cn = AB. That is, the product of two convergent series converges, and to the
right value, if at least one of the two series converges absolutely.
P
P
P
3.51 Theorem (Abel) If the series
an ,
bn ,
cn converge to A, B, C, respectively,
and cn = a0 bn + a1 bn1 + + an b0 , then C = AB.

71

Continuity
Reference: W. Rudin, Principles of Mathematical Analysis, 3rd edition, McGraw-Hill,
c1976, Chapter 4.
4.1 Definition. Let (X, dX ) and (Y, dY ) be metric spaces; suppose E X, f maps E
into Y , and p is a limit point of E. Write
lim f (x) = q

(1)

xp

if there is a point q Y with the following property: For every > 0, there exists a > 0
such that
dY (f (x), q) <
(2)
for all points x E for which

0 < dX (x, p) < .

(3)

It should be noted that p X, but that p need not be a point of E in the above definition.
Moreover, even if p E, we may very well have f (p) 6= lim f (x).
xp

4.2 Theorem. Let X, Y, E, f , and p be as in Definition 4.1. Then

lim f (x) = q

(4)

lim f (pn ) = q

(5)

xp

if and only if
n

for every sequence {pn } in E such that

pn 6= p,

lim pn = p.

(6)

Proof. Let {pn } be a sequence in E satisfying (6). Let > 0 be given. Then there
exists > 0 such that dY (f (x), q) < if x E and 0 < dX (x, p) < . Also, there exists N
such that n > N implies 0 < dX (pn , p) < . Thus, for n > N , we have dY (f (pn ), q) < ,
which shows that (5) holds.
Conversely, suppose (4) is false. Then there exists some > 0 such that for every
> 0 there exists a point x E (depending on ), for which dY (f (x), q) > but
0 < dX (x, p) < . Taking = n1 (n = 1, 2, 3, . . .), we thus find a sequence {xn } in E
satisfying (6) for which (5) is false.


Corollary. If f has a limit at p, this limit is unique.

Proof. Let lim f (x) = q. Then by Theorem 4.2, for every sequence {pn } such that pn 6= p
xp

for all n, and pn p, lim f (pn ) = q. The uniqueness of q then follows from Theorems
n

3.2(b).

72

Example. Consider two sequences {pn } and {qn } defined by

1
1
and qn =
pn =
1
(2n 2 )
(2n + 21 )

for n = 1, 2, . . .. It is clear that lim pn = lim qn = 0. However, since

n

1
1
1
1
lim sin
= lim sin(2n ) = 1 6= 1 = lim sin(2n + ) = lim sin ,
n
n
n
pn n
2
2
qn
by Theorem 4.2, lim sin x1 does not exist.
x0

4.3 Definition. Let f, g : E C. Define

(f g)(x) = f (x) g(x),
(f g)(x) = f (x)g(x),

f (x)
f
provided g(x) 6= 0.
( )(x) =
g
g(x)
If f (x) = c for all x E, where c C is fixed, then f is called a constant function, and
write f = c. If f, g : E R and f (x) > g(x) for all x E, then write f > g.
Similarly, if f , g : E Rk , then define f + g and f g by
(f + g)(x) = f (x) + g(x),

and if R, define (f )(x) = f (x).

(f g)(x) = f (x) g(x);

4.4 Theorem. Let X be a metric space, E X, p a limit point of E, f and g are

complex functions on E, and
lim f (x) = A,

xp

lim g(x) = B.

xp

Then
(a) lim (f + g)(x) = A + B;
xp

(b) lim (f g)(x) = AB;

xp

(c) lim ( fg )(x) =

xp

A
,
B

if B 6= 0.

Proof. In view of Theorem 4.2, these assertions follow immediately from the analogous
properties of sequences (Theorem 3.3).

Remark. If f , g : E Rk , then (a) remains true, and (b) becomes
(b ) lim (f g)(x) = A B.
xp

(Compare Theorem 3.4.)

73

Continuous functions
4.5 Definition. Let (X, dX ) and (Y, dY ) be metric spaces, E X, p E, and f : E Y .
Then f is said to be continuous at p if for every > 0, there exists a > 0 such that
dY (f (x), f (p)) <
for all x E such that dX (x, p) < . If f is continuous at every point of E, then f is said
to be continuous on E.
Note that f has to be defined at the point p in order to be continuous at p.
f is continuous at every isolated point p of E because there exists > 0 such that
N (p) = {q E : dX (p, q) < } = {p}; x E and dX (x, p) < x = p so that
dY (f (x), f (p)) = 0 <

for any > 0.

4.6 Theorem. In Definition 4.5, assume also that p is a limit point of E. Then f is
continuous at p if and only if lim f (x) = f (p).
xp

Proof. This is clear if we compare Definitions 4.1 and 4.5.

The next theorem states that a continuous function of a continuous function is continuous.
4.7 Theorem. Suppose X, Y, Z are metric spaces, E X, f : E Y , g : f (E) Z,
and h : E Z defined by
h(x) = g(f (x)) (x E).
If f is continuous at p E and g is continuous at f (p), then h is continuous at p. The
function h is called the composition or the composite of f and g, denoted by
h = g f.

Proof. Let > 0 be given. Since g is continuous at f (p), there exists > 0 such that
dZ (g(y), g(f (p))) < if dY (y, f (p)) < and y f (E).

Since f is continuous at p, there exists > 0 such that

dY (f (x), f (p)) <
It follows that

if dX (x, p) < and x E.

dZ (h(x), h(p)) = dZ (g(f (x)), g(f (p))) <

if dX (x, p) < and x E. Thus h is continuous at p.

74

4.8 Theorem. Let (X, dX ) and (Y, dY ) be metric spaces. A mapping f : X Y is

continuous on X if and only if f 1 (V ) is open in X for every open set V in Y .
Proof. Let V be an open set in Y . We have to show that every point of f 1 (V ) is an
interior point of f 1 (V ). Let p X be such that f (p) V . Since V is open, there exists
> 0 such that y V if dY (f (p), y) < ; and since f is continuous at p, there exists > 0
such that dY (f (x), f (p)) < if dX (x, p) < . Thus, x f 1 (V ) as soon as dX (x, p) < .

Suppose f 1 (V ) is open in X for every open set V in Y . Fix p X and > 0, let
V = {y Y : dY (y, f (p)) < }. Then V is open; hence f 1 (V ) is open; hence there exists
> 0 such that x f 1 (V ) as soon as dX (p, x) < . But if x f 1 (V ) then f (x) V ,
so that dY (f (x), f (p)) < .


Corollary. A mapping f : (X, dX ) (Y, dY ) is continuous if and only if f 1 (C) is closed

in X for every closed set C in Y .
Proof. This follows from Theorem 4.8, and a set is closed if and only if its complement is
open, and f 1 (E c ) = (f 1 (E))c for every E Y .

4.9 Theorem. Let (X, dX ) be a metric space and f, g : X C be continuous. Then
f g, f g, and f /g are continuous on X.
Proof. This follows from the fact that each of f g, f g and f /g is a continuous function
of the continuous functions f and g.


4.10 Theorem.
(a) Let (X, dX ) be a metric space, f1 , . . . , fk : X R, and let f : X Rk defined by
f (x) = (f1 (x), . . . , fk (x)) (x X);
then f is continuous if and only if each of the functions f1 , . . . , fk is continuous.
(b) If f , g : X Rk , then f + g : X Rk and f g : X R are continuous on X.
The functions f1 , . . . , fk are called the components of f .
Proof. Given > 0.
(a) For j = 1, 2, . . . , k, by virtue of the following inequalities
X
1/2
k
2
|fj (x) fj (y)| 6 kf (x) f (y)k2 =
|fi (x) fi (y)|
,
i=1

there exists > 0 such that dX (x, y) < implies kf (x) f (y)k2 < . Thus, when
dX (x, y) < then |fj (x) fj (y)| < .

75

For j = 1, 2, . . . , k, there exist j > 0 such that dX (x, y) < j |fj (x) fj (y)| <
. Choose = min(1 , . . . , k ) > 0. Then dX (x, y) < |fj (x) fj (y)| < for
k
k
j = 1, 2, . . . , k so that
X
1/2 X
1/2
k
k
2
2
kf (x) f (y)k2 =
|fi (x) fi (y)|
<
( )
= .
k
i=1
i=1

(b) follows from (a) and Theorem 4.9.

4.11 Examples.
(a) If x1 , . . . , xk are the coordinates of the point x Rk , the functions i defined by
i (x) = xi

(x Rk )

(8)

are continuous on Rk , since the inequality

|i (x) i (y)| 6 kx yk2

shows that we may take = in Definition 4.5. The functions i are sometimes
called the coordinate functions.
(b) Repeated application of Theorem 4.9 then shows that every monomial
xn1 1 xn2 2 xnk k

(9)

where n1 , . . . , nk are nonnegative integers, is continuous on Rk . The same is true

of constant multiples of (9), since constants are evidently continuous. It follows
that every polynomial P in k variables, given by
X
P (x) =
cn1 ,...,nk xn1 1 xnk k (x Rk )
(10)

is continuous on Rk . Here the coefficients cn1 ,...,nk are complex numbers, n1 , . . . , nk

are nonnegative integers, and the sum in (10) has finitely many terms.
(c) Every rational function in x1 , . . . , xk , that is, every quotient of two polynomials
of the form (10), is continuous on Rk wherever the denominator is different from
zero.
(d) From the triangle inequality one sees easily that


kxk2 kyk2  6 kx yk2 (x, y Rk ).
(11)
Hence the mapping x kxk2 is a continuous real function on Rk .
(e) If f : X Rk is continuous, and if : X R is defined by (p) = kf (p)k2 , it
follows, by Theorem 4.7, that is continuous on X.

76

Continuity and compactness

4.13 Definition. A mapping f : E Rk is said to be bounded if there is a real number
M such that kf (x)k2 6 M for all x E.
4.14 Theorem. Suppose f is a continuous mapping of a compact metric space X into
a metric space Y . Then f (X) is compact. (In other word, the continuous image of a
compact set is compact.)
S
Proof. Let {V } be an open cover of f (X), i.e., f (X) V . Since f is continuous,
Theorem 4.8 shows that each of the sets f 1 (V ) is open. Also,
[  [
1
Xf
V =
f 1 (V ),

i.e., {f (V )} is an open cover of X. Since X is compact, there are finitely many indices
1 , . . . , n such that
X f 1 (V1 ) f 1 (Vn ).
(12)
1
Since f (f (E)) E for every E Y , (12) implies that
f (X) f (f 1 (V1 ) f 1 (Vn )) V1 Vn .

This completes the proof.

(13)

4.15 Theorem. If f is a continuous mapping of a compact metric space X into Rk , then

f (X) is closed and bounded. Thus, f is bounded.
4.16 Theorem. (Extreme value theorem) Suppose f : X R is continuous, where X is
compact metric space X. Let
M = sup f (t) and m = inf f (t).
tX

tX

(14)

Then there exist points p, q X such that f (p) = M and f (q) = m. In other words, f
attains its maximum (at p) and its minimum (at q).
Proof. By Theorem 4.15, f (X) is a closed and bounded set of real numbers; hence f (X)
contains
M = sup f (X) and m = inf f (X),
by Theorem 2.28.

4.17 Theorem. Suppose f is a continuous 1-1 mapping of a compact metric space X
onto a metric space Y . Then the inverse mapping f 1 defined on Y by
f 1 (f (x)) = x (x X)

is a continuous mapping of Y onto X.

Proof. Applying Theorem 4.8 to in place of f 1 , we see that it suffices to prove that f (V )
is an open set in Y for every open set V in X. Fix such a set V . The complement V c of
V is closed in X, hence compact (Theorem 2.35); hence f (V c ) is a compact subset of Y
(Theorem 4.14) and so is closed in Y (Theorem 2.34). Since f is bijective, f (V ) is the
complement of f (V c ). Hence f (V ) is open.


77

4.18 Definition. Let f be a mapping of a metric space X into a metric space Y . We

say that f is uniformly continuous on X if for every > 0, there exists > 0 such that
dY (f (p), f (q)) <

(15)

for all p, q X for which dX (p, q) < .

Let us consider the differences between the concepts of continuity and of uniform continuity. First, uniform continuity is a property of a function on a set, whereas continuity
can be defined at a single point. Second, if f is continuous on X, then it is possible to
find, for each > 0 and for each x X, a number > 0 having the property specified
in Definition 4.5. This depends on and on p. If f is, however, uniformly continuous
on X, then it is possible, for each > 0, to find one number > 0 which will do for all
points p X.
Evidently, every uniformly continuous function is continuous. The next theorem shows
that these two notions of continuities are equivalent on compact sets.
4.19 Theorem. Let f be a continuous mapping of a compact metric space X into a
metric space Y . Then f is uniformly continuous on X.
Proof. Let > 0 be given. Since f is continuous, we can associate to each p X a
positive number (p) such that

q X, dX (p, q) < (p) implies dY (f (p), f (q)) < .

(16)
2
Let
1
(17)
J(p) = {q X : dX (p, q) < (p)}.
2
Since p J(p), {J(p)}pX is an open cover of X; since X is compact, there is a finite set
of points p1 , . . . , pn X such that
X J(p1 ) J(pn ).

We put

(18)

1
min((p1 ), . . . , (pn )).
(19)
2
Then > 0. Now let p, q X such that dX (p, q) < . By (18), there is an integer m,
1 6 m 6 n, such that p J(pm ); hence
1
dX (p, pm ) < (pm ),
(20)
2
and we also have
1
dX (q, pm ) 6 dX (p, q) + dX (p, pm ) < + (pm ) 6 (pm ).
2
Finally, (16) shows that therefore
=

dY (f (p), f (q)) 6 dY (f (p), f (pm )) + dY (f (pm ), f (q)) < .

This completes the proof.

78

The next theorem shows that compactness is essential in the hypotheses of Theorems
4.14, 4.15, 4.16, and 4.19.
4.20 Theorem. Let E be a noncompact set in R. Then
(a) there exists a continuous function on E which is not bounded;
(b) there exists a continuous and bounded function on E which has no maximum.
If, in addition, E is bounded, then
(c) there exists a continuous function on E which is not uniformly continuous.
Proof. Suppose first that E is bounded, so that there exists a limit point x0 of E which
is not a point of E.
Consider

1
(x E).
(21)
x x0
This is continuous on E (Theorem 4.9), but evidently unbounded. To see that (21) is
not uniformly continuous, let > 0 and > 0 be arbitrary, and choose a point x E
such that |x x0 | < . Taking t close enough to x0 , we can then make the difference
|f (t)f (x)| > , although |xt| < . Since this is true for every > 0, f is not uniformly
continuous on E. This proves (a) when E is bounded.
f (x) =

The function g given by

1
(x E)
1 + (x x0 )2
is continuous on E, and is bounded, since 0 < g(x) < 1. It is clear that
g(x) =

(22)

sup g(x) = 1,
xE

whereas g(x) < 1 for all x E. Thus g has no maximum on E. This proves (b) when E
is bounded.
Having proved the theorem for bounded sets E, let us now suppose that E is unbounded.
Then f (x) = x establishes (a), whereas
h(x) =

x2
1 + x2

(x E)

(23)

establishes (b), since

sup h(x) = 1
xE

and h(x) < 1 for all x E.

Assertion (c) would be false if boundedness were omitted from the hypotheses. For, let
E be the set of all integers. Then every function defined on E is uniformly continuous on
E. To see this, we need merely take < 1 in Definition 4.18.


79

Continuity and connectedness

4.22 Theorem. If f is a continuous mapping of a metric space X into a metric space
Y , and if E is a connected subset of X, then f (E) is connected.
Proof. Assume, on the contrary, that f (E) = A B, where A and B are nonempty
separated subsets of Y . Put G = E f 1 (A), H = E f 1 (B). Then E = G H, and
neither G nor H is empty.
we have G f 1 (A);
the latter set is closed, since f is continuous; hence
Since A A,
1

Since f (H) = B and A B = , we conclude

G f (A). It follows that f (G) A.
that G H = .
= . Thus G and H are separated. This is
The same argument shows that G H
impossible if E is connected.

4.23 Theorem. (Intermediate-value property) Let f be a continuous real function on
the interval [a, b]. If f (a) < f (b) and if c is a number such that f (a) < c < f (b),
then there exists x (a, b) such that f (x) = c. A similar result holds, of course, if
f (a) > f (b). Roughly speaking, a continuous real function assumes all intermediate
values on an interval.
Proof. By Theorem 2.47, [a, b] is connected; hence Theorem 4.22 shows that f ([a, b]) is a
connected subset of R, and the assertion follows if we appeal once more to Theorem 2.47.
(How?)

Theorem. (Fixed point property) Let f : [a, b] [a, b] be continuous. Then there exists
x0 [a, b] such that f (x0 ) = x0 .
Proof. If f (a) = a or f (b) = b, there is nothing to prove.

Otherwise, define h(x) = f (x) x, where x [a, b]. Then h : [a, b] [a, b] is continuous.
Since h(a) = f (a) a > 0 and h(b) = f (b) b < 0. By the Intermediate-value property
of continuous functions, there exists x0 (a, b) such that h(x0 ) = f (x0 ) x0 = 0, i.e.,
f (x0 ) = x0 .


80

Discontinuities
If x is a point in the domain of the function f at which f is not continuous, we say that
f is discontinuous at x, or that f has a discontinuity at x.
4.25 Definition. Let f be defined on (a, b). Consider any point x (a, b). We write the
right-hand limit
f (x+) = q
if f (tn ) q as n , for all sequences {tn } in (x, b) such that tn x. To obtain the
definition of the left-hand limit f (x), we restrict ourselves to sequences {tn } in (a, x).
It is clear that for any x (a, b), lim f (t) exists if and only if
tx

f (x+) = f (x) = lim f (t).

tx

4.26 Definition. Let f be defined on (a, b). If f is discontinuous at a point x, and if

f (x+) and f (x) exist, then f is said to have a discontinuity of the first kind, or a simple
discontinuity, at x, i.e.,
both f (x+) and f (x) exist and either f (x+) 6= f (x), or f (x+) = f (x) 6= f (x).

Otherwise, the discontinuity is said to be of the second kind, i.e., not both f (x+) and
f (x) exist.
4.27 Examples.
(a) Define
f (x) =

1 if x Q,
0 if x R \ Q.

Then f has a discontinuity of the second kind at every point x, since neither f (x+)
nor f (x) exists.
(b) Define
(
x if x Q,
f (x) =
0 if x R \ Q.
Then f is continuous at x = 0 and has a discontinuity of the second kind at every
other point (i.e., x 6= 0).
(c) Define

if 3 < x < 2,
x + 2
f (x) = x 2 if 2 6 x < 0,

x + 2
if 0 6 x < 1.
Then f has a simple discontinuity at x = 0 and is continuous at every other point
of (3, 1).
(d) Define
(
sin x1 if x 6= 0,
f (x) =
0
if x = 0.

Since neither f (0+) nor f (0) exists, f has a discontinuity of the second kind at
x = 0. Assuming that sin x is a continuous function, then Theorem 4.7 implies
that f is continuous at every point x 6= 0.

81

Example. Consider the function

for x < 1,
x
f (x) = 0
for x = 1,

2 x for x > 1.

We clearly have

lim f (x) = lim x2 = 1 and

x1

x1

lim f (x) = lim (2 x) = 1.

x1+

x1+

Thus, lim f (x) = 1. Since lim f (x) 6= f (1), f has a simple discontinuity at x = 1. f can
x1

x1

be made continuous at x = 1 by redefining f (1) = 1. The point x0 = 1 is a removable

discontinuity.
In general, if lim f (x) = lim f (x) = L and f (x0 ) 6= L, then by redefining f at x = x0 ,
i.e., defining

xx0

xx0 +

g(x) =

f (x) for x 6= x0 ,
,
L
for x = x0

then the discontinuity at x = x0 is removed, and the resulting function g is continuous at

x = x0 .

Example. (Thomaes function) Let f : (0, 1) R be defined as
(
1
if x = pq Q (0, 1), p, q relatively prime,
q
f (x) =
0 otherwise.
Show that f (x) is continuous at every x (0, 1) \ Q and discontinuous at every x
(0, 1) Q.

Solution. Let a (0, 1) Q. Then a = pq for some p, q relatively prime so that f (a) = 1q .
There exists {xn } (0, 1)\Q such that xn a as n . But lim f (xn ) = 0 < 1q = f (a),
n
proving f being discontinuous at x = a.

Given > 0. Let now a (0, 1) \ Q. Let q N be the smallest such that 0 < 1q < .
There exist finitely many p N, say, p1 , . . . , pk , which are relatively prime to q. Choose
= min |a pqi | > 0. If x (0, 1) \ Q and |x a| < , then |f (x) f (a)| = 0 < ; if
16i6k

x (0, 1) Q and |x a| < , then |f (x) f (a)| = |f (x)| <

at x = a.

1
q

< . Thus, f is continuous



Example. (Dirichlets discontinuous function) Let f : R R be defined by

(
1 if x Q,
f (x) =
0 if x R \ Q.
Show that f is discontinuous at every x R.

82

Solution. Let = 21 . For every x Q and any > 0, there exists y B (x) \ Q such that
|f (x) f (y)| = 1 > 21 = . Thus, f is discontinuous at x; also, for x R \ Q, there exists
y B (x) Q such that |f (x) f (y)| = 1 > 21 = . Consequently, f is discontinous at
every x R.

Monotone functions
4.28 Definition. Let f : (a, b) R. Then f is said to be monotonically increasing (resp.,
decreasing) on (a, b) if a < x < y < b implies f (x) 6 f (y) (resp., f (x) > f (y)). The class
of monotonic functions consists of both the increasing and the decreasing functions.
4.29 Theorem. Let f be monotonically increasing on (a, b). Then f (x+) and f (x)
exist at every x (a, b). More precisely,
sup f (t) = f (x) 6 f (x) 6 f (x+) = inf f (t).
x<t<b

a<t<x

(25)

Furthermore, if a < x < y < b, then

f (x+) 6 f (y).

(26)

Analogous results evidently hold for monotonically decreasing functions.

Proof. By hypothesis, the set
{f (t) : a < t < x}
is bounded above by the number f (x), hence sup f (t) = A exists. Evidently A 6 f (x).
a<t<x

We have to show that A = f (x).

Let > 0 be given. There exists > 0 such that a < x < x and
A < f (x ) 6 A.

(27)

f (x ) 6 f (t) 6 A (x < t < x).

(28)

Since f is monotone increasing, we have

Combining (27) and (28), we see that
Hence f (x) = A.

|f (t) A| < (x < t < x).

The second half of (25) is proved in precisely the same way (Exercise).
Next, if a < x < y < b, we see from (25) that
f (x+) = inf f (t) = inf f (t).
x<t<b

x<t<y

(29)

The last equality is obtained by applying (25) to (a, y) in place of (a, b). Similarly,
f (y) = sup f (t) = sup f (t).
a<t<y

Comparison of (29) and (30) gives (26).

(30)

x<t<y

83

Corollary. Monotonic functions have no discontinuities of the second kind.

This corollary implies that every monotonic function is discontinuous at a countable set
of points at most.
4.30 Theorem. (Frodas theorem) Let f be monotonic on (a, b). Then the set of points
of (a, b) at which f is discontinuous is at most countable.
Proof. Suppose, for the sake of definiteness, that f is increasing, and let
E = {x (a, b) : f is discontinuous at x}.

With every x E we associate a rational number r(x) such that

f (x) < r(x) < f (x+).

Since x1 < x2 implies f (x1 +) 6 f (x2 ), we see that r(x1 ) 6= r(x2 ) if x1 6= x2 . We have
thus established a 1-1 correspondence between the set E and a subset of Q. The latter is
countable.

Infinite limits and limits at infinity
4.32 Definition. For any c R, (c, +) = {x R : x > c} is called a neighborhood of
+. Similarly, (, c) = {x R : x < c} is a neighborhood of .
4.33 Definition. Let E R and let f : E R. We say that
f (t) A as t x,

if for every neighborhood U of A there is a neighborhood V of x such

where A, x R,
that V E 6= , and such that f (t) U for all t V E, t 6= x.
When A, x R, Definition 4.33 coincides with Definition 4.1.
4.34 Theorem. Let E R and let f, g be defined on E. Suppose
f (t) A,

g(t) B

as t x.

Then
(a) f (t) A implies A = A.
(b) (f + g)(t) A + B,
(c) (f g)(t) AB,
(d) (f /g)(t) A/B,
provided the right members of (b), (c), and (d) are defined.
Note that , 0 , /, A/0 are not defined.

84

Differentiation
5.1 Definition. Let f : [a, b] R. For any x [a, b] form the quotient
f (t) f (x)
(a < t < b, t 6= x),
(t) =
tx
and define
f (x) = lim (t)
tx

(1)
(2)

provided this limit exists.

We thus associate with the function f a function f whose domain is the set of points x
at which the limit (2) exists; f is called the derivative of f . If f is defined at a point x,
we say that f is differentiable at x. If f is defined at every point of a set E [a, b], we
say that f is differentiable on E.
It is possible to consider right-hand and left-hand limits in (2):
f+ (x) = lim (t),
tx+

f (x) = lim (t),

tx

called the right-hand derivative and left-hand derivative of f at x, respectively. In particular, at the endpoints a and b, the derivative, if it exists, is a right-hand or left-hand
derivative, respectively.
If f is defined on a segment (a, b) and if a < x < b, then f (x) is defined by (1) and (2),
as above. But f (a) and f (b) are not defined in this case.
5.2 Theorem. Let f be defined on [a, b]. If f is differentiable at a point x [a, b], then
f is continuous at x.
Proof. Since f (x) exists, we have, by Theorem 4.4,


f (t) f (x)
lim(f (t) f (x)) = lim
(t x)
tx
tx
tx
f (t) f (x)
= lim
lim(t x)
tx
tx
tx
= f (x) 0 = 0.

The converse of this theorem is not true. A standard counterexample is f (x) = |x|, which
is continuous at every x R but not differentiable at x = 0.

85

5.3 Theorem. Suppose that f, g : [a, b] R are differentiable at x [a, b]. Then f + g,
f g, f /g are differentiable at x and
(a) (f + g) (x) = f (x) + g (x);

(b) (f
= f (x)g(x) + f (x)g (x);
 g)(x)

f (x)g(x) f (x)g (x)

f
(x) =
.
(c)
g
[g(x)]2
Proof. (a) We have

(f + g)(t) (f + g)(x)
tx
tx
f (t) + g(t) f (x) g(x)
= lim
tx
tx
g(t) g(x)
f (t) f (x)
+ lim
= lim
tx
tx
tx
tx

= f (x) + g (x).

(f + g) (x) = lim

To prove (b), let h(x) = f (x)g(x) so that

h(t) h(x) = f (x)[g(t) g(x)] + g(x)[f (t) f (x)].

Divide both sides by t x, followed by letting t x, we obtain

h(t) h(x)
h (x) = lim
tx
tx
g(t) g(x)
f (t) f (x)
= f (x) lim
+ g(x) lim
tx
tx
tx
tx

= f (x)g (x) + g(x)f (x).

To prove (c), let h(x) = f (x)/g(x) so that






f (t) f (x)
g(t) g(x)
1
h(t) h(x)
g(x)
f (x)
.
=
tx
g(t)g(x)
tx
tx
Letting t x,


h(t) h(x)
1
g(t) g(x)
f (t) f (x)
lim
= lim
f (x) lim
g(x) lim
tx
tx g(t)g(x)
tx
tx
tx
tx
tx
g(x)f (x) f (x)g (x)
.
h (x) =
[g(x)]2

5.4 Examples. The derivative of any constant is clearly zero. If f is defined by f (x) = x,
then f (x) = 1. Repeated application of (b) and (c) then shows that xn is differentiable,
and that its derivative is nxn1 , for any integer n (if n < 0, we have to restrict ourselves to
x 6= 0). Thus every polynomial is differentiable, and so is every rational function, except
at the points where the denominator is zero.

86

The next theorem states that a differentiable function of a differentiable function is differentiable.
5.5 Theorem (Chain rule) Suppose f is continuous on [a, b], f (x) exists at some point
x [a, b], g is defined on an interval I containing f ([a, b]), and g is differentiable at the
point f (x). If
h(t) = g(f (t)) (a 6 t 6 b),
then h is differentiable at x, and
h (x) = g (f (x))f (x).

(3)

Proof. Since f is differentiable at x, f is continuous at x so that f (t) f (x) as t x.

Since g is differentiable at f (x), g is continuous at f (x) so that g(f (t)) g(f (x)) as
t x. By the definition of the derivative, we have
h(t) h(x)
tx
g(f (t)) g(f (x))
= lim
tx
tx



g(f (t)) g(f (x))
f (t) f (x)
= lim
tx
f (t) f (x)
tx



f (t) f (x)
g(f (t)) g(f (x))
lim
= lim
tx
tx
f (t) f (x)
tx

= g (f (x))f (x).

h (x) = lim

tx


5.6 Examples.
(a) Let f be defined by
f (x) =
For x 6= 0,

x sin x1
0

f (x) = sin

when x 6= 0,
when x = 0.

1 1
1
cos .
x x
x

(8)

At x = 0,
f (t) f (0)
1
= lim sin ,
t0
t0
t0
t

which does not exist. So, f (0) does not exist.

(b) Let f be defined by
(
x2 sin x1 when x 6= 0,
f (x) =
0
when x = 0.
lim

As above, we obtain for x 6= 0,

f (x) = 2x sin

1
1
cos .
x
x

(9)

(9)

(10)

87

At x = 0, since






 f (t) f (0) 
1
 = lim t sin  6 lim |t| = 0,
lim 
t0
t 0  t0 
t  t0

we see that

f (0) = 0.
(11)

Thus f is differentiable at all points x, but f is not a continuous function, since

cos x1 in (10) does not tend to a limit as x 0.
Mean value theorems
5.7 Definition. Let f be a real function defined on a metric space X. We say that f
has a local maximum (resp., local minimum) at a point p X if there exists > 0 such
that f (q) 6 f (p) (resp., f (q) > f (p)) for all q X with d(p, q) < .
5.8 Theorem. (Fermat) Let f be defined on [a, b]; if f has a local maximum at a point
x (a, b), and if f (x) exists, then f (x) = 0. The analogous statement for local minima
is also true.
Proof. Choose > 0 in accordance with Definition 5.7, so that
If x < t < x, then

a < x < x < x + < b.

f (t) f (x)
> 0.
tx
Letting t x, we see that f (x) > 0. If x < t < x + , then
f (t) f (x)
6 0.
tx
which shows that f (x) 6 0. Hence f (x) = 0.

5.9 Theorem. (Cauchys mean value theorem) If f and g are continuous real functions
on [a, b] which are differentiable in (a, b), then there is a point x0 (a, b) at which
Proof. Put

[f (b) f (a)]g (x0 ) = [g(b) g(a)]f (x0 ).

h(t) = [f (b) f (a)]g(t) [g(b) g(a)]f (t) (a 6 t 6 b).

Then h is continuous on [a, b], h is differentiable in (a, b), and
h(a) = [f (b) f (a)]g(a) [g(b) g(a)]f (a) = h(b).

(12)

To prove the theorem, we have to show that h (x0 ) = 0 for some x0 (a, b). If h is
constant, this holds for every x (a, b). If h(t) > h(a) for some t (a, b), let x0 [a, b]
at which h attains its maximum (Theorem 4.16). By (12), x0 (a, b), and Theorem 5.8
shows that h (x0 ) = 0. If h(t) < h(a) for some t (a, b), the same argument applies if we
choose for x0 [a, b] where h attains its minimum.


88

5.10 Theorem (Lagranges Mean value theorem) If f is a real continuous function on

[a, b] which is differentiable in (a, b), then there is a point x0 (a, b) at which
f (b) f (a) = (b a)f (x).

Proof. Take g(x) = x in Theorem 5.9.

The geometric interpretation of the mean value theorem is as follows.

y
y = f (x)

x0

Example. Prove that sin x sin y 6 x y for x, y R such that x > y.

Solution. Let f (t) = sin t, where t R. Then f (t) is continuous on [y, x] and differentiable
in (y, x). By mean value theorem, there exists c (y, x) such that
sin x sin y
= f (c) = cos c 6 1.
xy
Hence, sin x sin y 6 x y since x y > 0.

5.11 Theorem. Suppose f is differentiable in (a, b).
(a) If f (x) > 0 for all x (a, b), then f is monotonically increasing.
(b) If f (x) = 0 for all x (a, b), then f is constant.
(c) If f (x) 6 0 for all x (a, b), then f is monotonically decreasing.

Proof. Let x1 , x2 (a, b) be such that x1 < x2 . Then there exists x (x1 , x2 ) such that

> 0 if f (x) > 0,

f (x2 ) f (x1 ) = (x2 x1 )f (x) = 0 if f (x) = 0,

6 0 if f (x) 6 0.

Example. Let f (x) = ex (1 x), which is differentiable at every x R. Then for

x [0, 1],
f (x) = 1 ex > 0
so that f is monotone increasing in [0, 1]. Thus, for x [0, 1], 0 = f (0) 6 f (x) whence
ex > 1 x.

89

The continuity of derivatives

Example 5.6(b) shows that a function f may have a derivative f which exists at every
point, but is discontinuous at some point.
Derivatives which exist at every point of an interval have the intermediate value property.
5.12 Theorem. Suppose f is a differentiable real function on [a, b] and suppose f (a) <
< f (b). Then there is a point x (a, b) such that f (x) = . A similar result holds if
f (a) > f (b).
Proof. Put g(t) = f (t) t. Then g (a) < 0, so that g(t1 ) < g(a) for some t1 (a, b).
Similarly, g (b) > 0, so that g(t2 ) < g(b) for some t2 (a, b). Hence g attains its minimum
on [a, b] (Theorem 4.16) at some point x such that a < x < b. By Theorem 5.8, g (x) = 0.
Hence f (x) = .

Corollary. If f is differentiable on [a, b], then f cannot have any simple discontinuities
on [a, b].
But f may very well have discontinuities of the second kind.
LH
opitals rule
5.13 Theorem. Suppose f and g are real and differentiable in (a, b), g (x) 6= 0 for all
x (a, b), where 6 a < b 6 +. Suppose
f (x)
A as x a.
(13)
g (x)
If
f (x) 0 and g(x) 0 as x a,
(14)
or if
f (x) and g(x) + as x a,
(15)
then
f (x)
A as x a,
(16)
g(x)
The analogous statement is also true if x b, or if g(x) in (15).
f (x)
0
f (x)
above is called an indeterminate form . Note that if
is an
xa g(x)
0
g(x)

1/g(x)
0
indeterminate form then rewriting it as
we have an indeterminate form

1/f (x)
0
to which lHopitals rule applies.

The limit lim

90

ln x
.
x1
Solution. The given limit is an indeterminate form 00 . Applying lHopitals rule yields

Example. Evaluate lim

x1

1/x
ln x
= lim
= 1.
x1 1
x1 x 1
lim

ln sin x
.
x0 ln tan x
. Applying lHopitals rule yields
Solution. The given limit is an indeterminate form

Example. Evaluate lim+

lim+

x0

ln sin x
cos x/ sin x
= lim+
= lim cos2 x = 1.
ln tan x x0 sec2 x/ tan x x0+


x ex1
.
x1 (x 1)2

Example. Evaluate lim

Solution. The given limit is an indeterminate form 00 . Applying lHopitals rule yields
1 ex1
ex1
1
x ex1
=
lim
=
lim
= .
x1 2(x 1)
x1
x1 (x 1)2
2
2
lim

f (x)
cannot be evaluated, it does not necessarily imply that
xa g (x)

Remark. When lim

f (x)
does not exist. What we can conclude is that lHopitals rule cannot be apg(x)
x + sin x

plied in such a situation. For example, lim

is an indeterminate form so
x
x

1 + cos x
x + sin x
= lim
which cannot
that an application of lHopitals rule yields lim
x
x
x
1
be determined. However, since sin x is bounded, we have


sin x
x + sin x
= 1.
= lim 1 +
lim
x
x
x
x
lim

xa

Example. Evaluate lim (x

x 2

) tan x.
2

) tan x is an indeterminate form 0 .

x 2
x/2
2
However, by writing tan x = 1/ cot x, (x 2 ) tan x = x/2
is now an indeterminate form
cot x
0
to which lHopitals rule is applicable:
0
x 2
1

= lim
= lim ( sin2 x) = 1.
lim (x ) tan x = lim
x 2 cot x
x 2 csc2 x
x 2
x 2
2

Solution. Since lim tan x = , lim (x

91

Example. Evaluate lim (sec x tan x).

x/2

Solution. As x /2, sec x and tan x so that lim (sec x tan x) is an

x/2

indeterminate form . But

cos x
1 sin x
= lim
= lim cot x = 0,
x/2 sin x
x/2
x/2
cos x

lim (sec x tan x) = lim

x/2

where the second limit is an indeterminate form 00 .

Example. Evaluate lim (cos x)csc x .

x0

Solution. The given limit is an indeterminate form 1 . Let y = (cos x)csc x . Then
ln y = (csc x) ln cos x which is an indeterminate form 0 as x 0.
ln cos x
sin x/ cos x
lim ln y = lim (csc x) ln cos x = lim
= lim
= 0,
x0
x0
x0 sin x
x0
cos x


from which 0 = lim ln y = ln lim y since ln y is continuous. Thus, lim y = 1.

x0

x0

x0

Example. Evaluate lim+ (1 cos x)1/ ln x .

x0

Solution. The given limit is an indeterminate form 00 . Let y = (1 cos x)1/ ln x . Then

ln(1 cos x)
which is an indeterminate form as x 0+ .
ln y =
ln x

ln(1 cos x)
sin x/(1 cos x)
x sin x
lim+ ln y = lim+
= lim+
= lim+
x0
x0
x0
x0 1 cos x
ln x
1/x
cos x x sin x + cos x
x cos x + sin x
= lim+
= 2.
= lim+
x0
x0
sin x
cos x


Since ln y is continuous, 2 = lim+ ln y = ln lim+ y lim+ y = e2 .

x0

x0

x0

Example. Evaluate lim x1/x .

x

Solution. The given limit is an indeterminate form 0 . Letting y = x1/x and applying
lHopitals rule, we get
ln x
1/x
lim ln y = lim
= lim
= 0.
x
x x
x 1


Since ln y is continuous, 0 = lim ln y = ln lim y , hence lim y = 1.

x

92

Derivatives of higher order

5.14 Definition. If f has a derivative f on an interval, and if f is itself differentiable,
we denote the derivative of f by f and call f the second derivative of f . Continuing
in this manner, we obtain functions
f, f , f , f (3) , . . . , f (n) ,
each of which is the derivative of the preceding one, i.e.,
f (n) (x) = (f (n1) (x)) ,

n = 1, 2, 3, . . . .

f (n) is called the nth derivative, or the derivative of order n, of f . In order for f (n) (x)
to exist at a point x, f (n1) (t) must exist in a neighborhood of x (or in a one-sided
neighborhood, if x is an endpoint of the interval on which f is defined), and f (n1) must
be differentiable at x.
Theorem. Let f (x) and g(x) be n times differentiable functions. Then
dn
dn
kf (x) = k n f (x), where k is a constant.
(a)
dxn
dx
dn
dn
dn
[f
(x)

g(x)]
=
f
(x)

g(x).
(b)
dxn
dxn
dxn
Proof. Exercise.

Example. Let f (x) =

x
and n be an integer greater than 1. Find f (n) (0).
1 x2

Solution. Resolving f (x) into partial fractions, we have



1
1
1

.
f (x) =
2 1x 1+x

It is easy to see (by induction, for instance) that





1
1
dn
n!
(1)n n!
dn
and
,
=
=
dxn 1 x
(1 x)n+1
dxn 1 + x
(1 + x)n+1


1
n!
(1)n
dn
. Setting x = 0 then gives

so that n f (x) =
dx
2 (1 x)n+1 (1 + x)n+1
(
n!
0 if n is even,
f (n) (0) = [1 (1)n ] =
2
n! if n is odd.

Theorem (Leibniz). Let f (x) and g(x) be n times differentiable functions of x. Then
n  
X
n (r)
(n)
f (x)g (nr) (x).
(f (x)g(x)) =
r
r=0

93

Proof. Induction on n. By the product rule, (f g) (x) = f (x)g(x) + f (x)g (x) so that it
is true for n = 1. Assume that it is true for n = k, i.e.,
k  
X
k (r)
(k)
(f (x)g(x)) =
f (x)g (kr) (x).
r
r=0

Taking derivative, we have that

k  
X
k
(k+1)
(f (x)g(x))
=
[f (r+1) (x)g (kr) (x) + f (r) (x)g (k+1r) (x)]
r
r=0
k+1
k  
X k 
X
k (r)
(r)
(k+1r)
=
f (x)g
(x) +
f (x)g (k+1r) (x)]
r

1
r
r=1
r=0




k
X
k
k
(k+1)
= f (x)g
(x) +
(
+
)f (r) (x)g (k+1r) (x) + f (k+1) (x)g(x)
r1
r
r=1

k 
X
k + 1 (r)
(k+1)
f (x)g (k+1r) (x) + f (k+1) (x)g(x)
= f (x)g
(x) +
r
r=1

k+1 
X
k + 1 (r)
=
f (x)g (k+1r) (x),
r
r=0






k
. Thus, it is also true for n = k + 1.
+ kr = k+1
since r1
r
By the principle of mathematical induction it is true for all n > 1.

Example. Let y = xeax , where a R. Find y (20) .

Solution. Since (x)(n) = 0 for n > 1, by Leibnizs theorem, we have

  20
 
20   r
X
20 d x d20r ax
d
20
20 dx d19 ax
(20)
ax
y
=
( r )( 20r e ) =
x 20 e +
( )(
e )
dx
r dx dx
0
1 dx dx19
r=0
= xa20 eax + 20a19 eax = (ax + 20)a19 eax .

Taylors theorem
5.15 Theorem. Suppose f is a real function on [a, b], n is a positive integer, f (n1) is
continuous on [a, b], f (n) (x) exists for every x (a, b). Let , x [a, b], 6= x, and define
P (x) =

n1 (k)
X
f ()
k=0

k!

(x )k .

(23)

Then there exists a point between and x such that

f (x) = P (x) +

f (n) ()
(x )n .
n!

(24)

94

For n = 1, this is just the mean value theorem. In general, the theorem shows that f can
be approximated by a polynomial of degree n 1, and that (24) allows us to estimate the
error, if we know bounds on |f (n) (x)|.
Proof. Let M be the number defined by
and put

f (x) = P (x) + M (x )n

(25)

g(t) = f (t) P (t) M (t )n (a 6 t 6 b).

(26)
(n)
We have to show that n!M = f () for some between and x. By (23) and (26),
g (n) (t) = f (n) (t) n!M

(a < t < b).

(27)

Hence the proof will be complete if we can show that g (n) () = 0 for some between
and x. Since P (k) () = f (k) () for k = 0, 1, . . . , n 1, we have
g() = g () = = g (n1) () = 0.

(28)

Our choice of M shows that g(x) = 0, so that g (x1 ) = 0 for some x1 between and x, by
the mean value theorem. Since g () = 0, we conclude similarly that g (x2 ) = 0 for some
x2 between and x1 . After n steps we arrive at the conclusion that g (n) (xn ) = 0 for some
xn between and xn1 , that is, between and . Thus, = xn satisfies g (n) () = 0. 

95

Rn =

f (n) ()
(x )n is called the remainder after n terms.
n!

If lim Rn = 0 for all x I, where I is some interval of real numbers containing , then
n

the Taylor series of f (x) about x = exists.

If = 0, the Taylor series is called a Maclaurin series.

If n is large enough, the remainder Rn 0. By dropping Rn , we obtain a Taylor polynomial:
Tn (x) = f () + f ()(x ) +

f ()(x )2
f (n1) ()(x )n1
+ +
.
2!
(n 1)!

For x close to , Tn (x) is a good approximation of f (x).

The following are the graphs of f (x) = sin x, T1 (x), T3 (x), T5 (x) and T7 (x), where
n1 x2n1
5
3
. The thickened graph is that of y = sin x. It
T2n1 (x) = x x3! + x5! + (1)(2n1)!
is clear that as n increases, the graph of y = Tn (x) is getting closer to that of y = sin x.
Also, as one moves away from x = 0, the approximation deteriorates.
T1 (x)

T5 (x)

-4

-2

T7 (x)

-2

T3 (x)
-4

Example. Consider the exponential function f (x) = ex . Taking successive derivatives,

we have for n = 1, 2, 3, . . .,
f (x) = f (x) = f (x) = = f (n) (x) = ex
so that f (0) = f (0) = f (0) = = f (n) (0) = e0 = 1. Also, f (0) = e0 = 1. Expanding
f about a = 0, we then have
ex = 1 +

x2
xn
ex0 xn+1
x
+
+ +
+
1!
2!
n! (n + 1)!

96
ex0 xn+1
n (n+1)!

for some x0 lying in between x and 0. It remains to show that lim

= 0. There

exists an integer k > 0 such that |x| < k2 . Then for n > k,

 x n+1  


x0 k n+1k
 
 e|x| |x|k
 e 0x
e
x
x
1
 


 (n + 1)!  =  k!(k + 1)(k + 2) (n + 1)  6 k!
2n+1k
0

as n . Thus, ex = 1 + x +

x2
2!

x3
3!

+ follows.

Taylor series
(a) gives series representations of functions,
(b) provides a source of polynomial approximations of functions.
The usefulness of the latter depends on the estimation of the remainder term.
Example. Consider the logarithmic function f (x) = ln(1 + x), where 1 < x < 1.
Taking successive derivatives, we have
1
,
1+x
(1)3 (3!)
,
(1+x)4

f (x) =
f (4) (x) =

f (x) =

1
,
(1+x)2

...

(1)2 (2!)
,
(1+x)3
(1)n1 (n1)!
(1+x)n

f (x) =
f (n) (x) =

so that f (n) (0) = (1)n1 (n 1)!. Also, f (0) = ln 1 = 0. The Taylor expansion of f
about a = 0 is
(1)n1 xn
(1)n xn+1
x2 x3
+
+
+
ln(1 + x) = x
2
3
n
(n + 1)(1 + x0 )n+1
for some x0 lying in between x and 0.
Suppose now that we want to estimate ln( 23 ) = ln(1 + 12 ) with error less than 106 .
In this case 0 < x0 <

The least n for which

1
2

so that 1 + x0 > 1 and hence



 (1)n ( 21 )n+1  ( 21 )n+1


 (n + 1)(1 + x0 )n+1  < n + 1 .

( 21 )n+1
n+1

< 106 is 15. Thus,

( 1 )15
1
1 ( 1 )2 ( 1 )3
ln(1 + ) 2 + 2 + 2
2
2
2
3
15
59848147
= 0.405466 (correct to 6 decimal places).
=
147603456
In fact, ln(1 + 12 ) = 0.405465 correct to 6 decimal places.

97

Some important Taylor series together with their intervals of convergence are as follows:
x2 x3 x4
x
+
+
+
< x < ,
e =1+x+
2!
3!
4!
x3 x5 x7
sin x = x
+

+
< x < ,
3!
5!
7!
x2 x4 x6
cos x = 1
+

+
< x < ,
2!
4!
6!
x2 x3 x4
+

+
1 < x 6 1,
ln(1 + x) = x
2
3
4
x3 x5 x7
tan1 x = x
+

+
1 6 x 6 1,
3
5
7

X
( 1) ( n + 1)xn
1 < x < 1,

(1 + x) = 1 +
0 6= R .
n!
n=1

98

The Riemann-Stieltjes integrals

Definition and existence of the integral
6.1 Definition. Let [a, b] be a given interval. By a partition P of [a, b] we mean a finite
set of points x0 , x1 , . . . , xn , where
a = x0 6 x1 6 6 xn = b.

We write

xi = xi xi1 (i = 1, 2, . . . , n).
Now suppose that f is a bounded real function defined on [a, b]. Corresponding to each
partition P of [a, b] we put
Mi =

sup

f (x),

mi =

xi1 6x6xi

U (P, f ) =

n
X

Mi xi ,

inf

xi1 6x6xi

L(P, f ) =

i=1

and finally

n
X

f (x),
mi xi

i=1

Z ab

f d = inf U (P, f ),

(1)

f d = sup L(P, f ),

(2)

where the inf and the sup are taken over all partitions P of [a, b]. The left members of
(1) and (2) are called the upper and lower Riemann integrals of f over [a, b], respectively.
If the upper and lower integrals are equal, we say that f is Riemann-integrable on [a, b],
we write f R (here, R denotes the set of Riemann-integrable functions), and we denote
the common value of (1) and (2) by
Z

or by

f dx,

(3)

f (x) dx.

(4)

b
a

This is the Riemann integral of f over [a, b]. Since f is bounded, m = inf f (x) and
a6x6b

M = sup f (x) exist such that

a6x6b

m 6 f (x) 6 M

(a 6 x 6 b).

Hence, for every P ,

m(b a) 6 L(P, f ) 6 U (P, f ) 6 M (b a),

so that the numbers L(P, f ) and U (P, f ) form a bounded set. This shows that the upper
and lower integrals are defined for every bounded function f . The question of their
equality, and hence the question of the integrability of f , is a more delicate one.

99

R1
Example. Consider the Riemann integral 0 x2 dx. Let Pn = {x0 , x1 , . . . , xn } be the
uniform partition of [0, 1], where xi = ni for i = 0, 1, . . . , n. Then xi = n1 for all i. Since
f (x) = x2 is increasing, we have for i = 1, 2, . . . , n,
Mi =

i
x2 = x2i = ( )2
n
xi1 6x6xi
sup

and

mi =

inf

xi1 6x6xi

x2 = x2i1 = (

i1 2
).
n

The upper and lower Riemann sums are

n
n
X
1 X 2 (n + 1)(2n + 1)
i 2 1
i =
U (P, f ) =
,
( ) ( )= 3
2
n
n
n
6n
i=1
i=1

n
n
X
(n 1)(2n 1)
1 X
i1 2 1
(i 1)2 =
) ( )= 3
,
L(P, f ) =
(
n
n
n i=1
6n2
i=1
P
where we have used the fact that ni=1 i2 = n(n+1)(2n+1)
. Letting n , we have
6
Z 1
1
lim U (P, f ) = lim L(P, f ) = =
x2 dx.
n
n
3
0

6.2 Definition. Let be a monotonically increasing function on [a, b] (since (a) and
(b) are finite, it follows that is bounded on [a, b]). Corresponding to each partition
P = {a = x0 < x1 < < xn = b} of [a, b], we write
i = (xi ) (xi1 ),

i = 1, 2, . . . , n.

It is clear that i > 0. For any real function f which is bounded on [a, b], we put
U (P, f, ) =

n
X

Mi i ,

L(P, f, ) =

i=1

n
X

mi i ,

i=1

called the upper and lower Riemann-Stieltjes sums, respectively, where Mi =

and mi =

inf

xi1 6x6xi

sup

f (x),

xi1 6x6xi

f (x), and we define

Z
Z

f d = inf U (P, f, ),

(5)

f d = sup L(P, f, ),

(6)

a
b

the inf and sup again being taken over all partitions. If
common value is denoted by
Z b
f d,

Rb
a

f d =

Rb
a

f d, then their
(7)

or sometimes by

f (x) d(x),
a

and is called the Riemann-Stieljes integral of f with respect to , over [a, b].

(8)

100

R2
Example. Consider the Riemann-Stieljes integral I = 0 x2 dx, where (x) = x is
the largest integer 6 x. The largest integer function x is right continuous at integers
with left limits. To compute I, we consider the partition Pn = {x0 , x1 , x2 , x3 , x4 }, where
x0 = 0, x1 = 1 n1 , x2 = 1, x3 = 2 n1 and x4 = 2. Then

1
0 = 0 0 = 0,
n
1
2 = 1 1 = 1 0 = 1,
n
1
3 = 2 1 = 1 1 = 0,
n
1
4 = 2 2 = 2 1 = 1.
n
The upper and lower Riemann-Stieljes sums are then
1
1
U (Pn , x2 , x) = (1 )2 (0) + (1)2 (1) + (2 )2 (0) + (2)2 (1) = 5,
n
n
1
1
1
1
L(Pn , x2 , x) = (0)2 (0) + (1 )2 (1) + (1)2 (0) + (2 )2 (1) = (1 )2 + (2 )2 .
n
n
n
n
Letting n ,
1 = 1

L(Pn , x2 , x) sup L(Pn , x2 , x) = 5 = U (Pn , x2 , x).

Consequently, the Riemann-Stieljes integral I = 5. Here, all we need are sufficient number
of points in the partition P to capture the jumps in f and .

If (7) exists, i.e., if (5) and (6) are equal, we say that f is integrable with respect to , in
the Riemann sense, and write f R().
By taking (x) = x, the Riemann integral is seen to be a special case of the RiemannStieltjes integral. Let us mention explicitly, however, that in the general case need not
even be continuous.
Regarding the notation, we prefer (7) to (8), since the letter x, the so-called variable of
integration in (8), is a dummy variable, it does not matter which letter we use to represent
Rb
the integral. For instance, (8) is the same as a f (y) d(y). The integral depends on f, ,
a and b, but not on the variable of integration, which may as well be omitted.
6.3 Definition. We say that the partition P is a refinement of P if P P (that is,
if every point of P is a point of P ). Given two partitions, P1 and P2 , we say that P is
their common refinement if P = P1 P2 .
6.4 Theorem. If P is a refinement of P , then
L(P, f, ) 6 L(P , f, )

(9)

U (P , f, ) 6 U (P, f, ).

(10)

and

101

Proof. To prove (9), suppose first that P contains just one point more than P =
{x0 , x1 , . . . , xn }. Let this extra point be x , and suppose xi1 < x < xi for some i.
Put
w2 = inf f (x).
w1 =
inf f (x),
xi1 6x6x

x 6x6xi

Clearly w1 > mi and w2 > mi , where, as before, mi =

inf

xi1 6x6xi

f (x). Hence

L(P , f, ) L(P, f, ) = w1 [(x ) (xi1 )] + w2 [(xi ) (x )] mi [(xi ) (xi1 )]

= (w1 mi )[(x ) (xi1 )] + (w2 mi )[(xi ) (x )]
> 0.

If P contains k points more than P , we repeat this reasoning k times, and arrive at (9).
The proof of (10) is analogous.


6.5 Theorem.

Rb
a

f d 6

Rb
a

f d.

Proof. Let P be the common refinement of two partitions P1 and P2 . By Theorem 6.4,
L(P1 , f, ) 6 L(P , f, ) 6 U (P , L, ) 6 U (P2 , f, ).
Hence
L(P1 , f, ) 6 U (P2 , f, ).
If P2 is fixed and the sup is taken over all P1 , (11) gives
Z b
f d 6 U (P2 , f, ).

(11)

(12)

The theorem follows by taking the inf over all P2 in (12).

6.6 Theorem. f R() on [a, b] if and only if for every > 0 there exists a partition
P such that
U (P, f, ) L(P, f, ) < .
(13)
Proof. For every partition P of [a, b] we have
Z b
Z b
f d 6
f d 6 U (P, f, ).
L(P, f, ) 6
a

Thus (13) implies

06

b
a

f d

f d < .
a

Hence, if (13) can be satisfied for every > 0, we have

Z b
Z b
f d =
f d,
a

that is, f R().

Suppose f R(), and let > 0 be given. Then there exist partitions P1 and P2
such that
Z b

(14)
U (P2 , f, ) <
f d + ,
2
a

102
b

(15)
f d < L(P1 , f, ) + .
2
a
We choose P to be the common refinement of P1 and P2 . Then Theorem 6.4, together
with (14) and (15), shows that
Z b

U (P, f, ) 6 U (P2 , f, ) <

f d + < L(P1 , f, ) + 6 L(P, f, ) + ,
2
a
so that (13) holds for this partition P .

Z

Theorem 6.6 furnishes a convenient criterion for integrability. Before we apply it, we state
some closely related facts.
6.7 Theorem.
(a) If (13) holds for some P and some , then (13) holds (with the same ) for every
refinement of P .
(b) If (13) holds for P = {x0 , x1 , . . . , xn } and if si , ti are arbitrary points in [xi1 , xi ],
then
n
X
|f (si ) f (ti )|i < .
i=1

(c) If f R() and the hypotheses of (b) hold, then

 n

Z b
X



f (ti )i
f (x) d < .



a
i=1

Proof. (a) Let P be a refinement of P . Theorem 6.4

L(P, f, ) 6 L(P , f, ) 6 U (P , f, ) 6 U (P, f, )

implies
U (P , f, ) L(P , f, ) 6 U (P, f, ) L(P, f, ) < .
(b) Both f (si ), f (ti ) [mi , Mi ] so that |f (si ) f (ti )| 6 Mi mi . Thus
n
X
|f (si ) f (ti )|i 6 U (P, f, ) L(P, f, ) < .
i=1

(c) follows from the following inequalities

Z b
n
X
L(P, f, ) 6
f d 6 U (P, f, ).
f (ti )i 6 U (P, f, ) and L(P, f, ) 6
i=1

103

6.8 Theorem. If f is continuous on [a, b] then f R() on [a, b].

Proof. Let > 0 be given. There exists > 0 such that

[(b) (a)] < .

Since f is uniformly continuous on [a, b], there exists a > 0 such that
|f (x) f (t)| <

(16)

if x, t [a, b] and |x t| < . If P is any partition of [a, b] such that xi < for all i,
then (16) implies that
Mi mi 6 (i = 1, 2, . . . , n)
(17)
and therefore
n
X
U (P, f, ) L(P, f, ) =
(Mi mi )i
i=1
n
X

i=1

By Theorem 6.6, f R().

i = [(b) (a)] < .



6.9 Theorem. If f is monotonic on [a, b], and if is continuous and monotone increasing
on [a, b], then f R().

Proof. Let > 0 be given. For any positive integer n, choose a partition such that
(b) (a)
(i = 1, 2, . . . , n).
i =
n
This is possible since is continuous (Theorem 4.23). We suppose that f is monotonically
increasing. Then
Mi = f (xi ), mi = f (xi1 ) (i = 1, 2, . . . , n),
so that
n
(b) (a) X
U (P, f, ) L(P, f, ) =
[f (xi ) f (xi1 )]
n
i=1
(b) (a)
[f (b) f (a)] <
n
if n is taken large enough. By Theorem 6.6, f R().
=

104

6.10 Theorem. Suppose f is bounded on [a, b], f has only finitely many points of
discontinuity on [a, b], and is continuous at every point at which f is discontinuous.
Then f R().
Proof. Let > 0 be given. Put M = sup |f (x)|, let E be the set of points at which f is
discontinuous. Since E is finite and is continuous at every point of E, we can cover E
by finitely many disjoint intervals [uj , vj ] [a, b] such that the sum of the corresponding
differences (vj ) (uj ) is less than . Furthermore, we can place these intervals in such
a way that every point of E (a, b) lies in the interior of some [uj , vj ].
Remove the segments (uj , vj ) from [a, b]. The remaining set K is compact. Hence f is
uniformly continuous on K, and there exists > 0 such that |f (s) f (t)| < if s, t K
and |s t| < .
Now form a partition P = {x0 , x1 , . . . , xn } of [a, b] as follows: For all j, uj , vj P . No
point of any segment (uj , vj ) occurs in P . If xi1 6= uj for all j, then xi < . Note that
Mi mi < 2M for every i, and that Mi mi < unless xi1 = uj for some j. Hence, as
in the proof of Theorem 6.8,
U (P, f, ) L(P, f, ) 6 [(b) (a)] + 2M .

Since is arbitrary, Theorem 6.6 shows that f R().

Note. If f and have a common point of discontinuity, then f need not be in R().
6.11 Theorem. Suppose f R() on [a, b], m 6 f 6 M , is continuous on [m, M ],
and h(x) = (f (x)) on [a, b]. Then h R() on [a, b].
Proof. Choose > 0. Since is uniformly continuous on [m, M ], there exists > 0 such
that < and |(s) (t)| < if |s t| 6 and s, t [m, M ].
Since f R(), there is a partition P = {x0 , x1 , . . . , xn } of [a, b] such that
U (P, f, ) L(P, f, ) < 2 .

(18)

Let Mi , mi have the same meaning as in Definition 6.1, and let Mi , mi be the analogous
numbers for h. Divide the numbers 1, 2, . . . , n into two classes: i A if Mi mi < ,
i B if Mi mi > .
For i A, our choice of shows that Mi mi 6 .
For i B, Mi mi 6 2K, where K = sup |(t)|, m 6 t 6 M . By (18), we have
X
X

i 6
(Mi mi )i < 2
iB

iB

(19)

105

so that

i < . It follows that

X
X
U (P, h, ) L(P, h, ) =
(Mi mi )i +
(Mi mi )i
iB

iA

iB

6 [(b) (a)] + 2K < [(b) (a) + 2K].

Since is arbitrary, Theorem 6.6 implies that h R().

Remark. This theorem suggests the question: Just what functions are Riemann-integrable?
The answer is given by Theorem 11.33(b).
6.12 Theorem. (Properties of the integral)
(a) If f, f1 , f2 R() on [a, b], then f1 + f2 R() and cf R() for every constant
c, and
Z b
Z b
Z b
Z b
Z b
f d.
cf d = c
f2 d,
f1 d +
(f1 + f2 ) d =
a

(b) If f1 (x) 6 f2 (x) on [a, b], then

Z b

f1 d 6

f2 d.
a

(c) If f R() on [a, b] and if a < c < b, then f R() on [a, c] and on [c, b], and
Z b
Z b
Z c
f d.
f d =
f d +
a

(d) If f R() on [a, b] and if |f (x)| 6 M on [a, b], then

Z b




 6 M [(b) (a)].
f
d


a

(e) If f R(1 ) R(2 ), then f R(1 + 2 ) and

Z b
Z b
Z b
f d2 ;
f d1 +
f d(1 + 2 ) =
a

if f R() and c is a positive constant, then f R(c) and

Z b
Z b
f d.
f d(c) = c
a

Proof. If f = f1 + f2 and P is any partition of [a, b], we have

L(P, f1 , ) + L(P, f2 , ) 6 L(P, f, )

6 U (P, f, ) 6 U (P, f1 , ) + U (P, f2 , ).

(20)

If f1 , f2 R(), let > 0 be given. There are partitions Pj (j = 1, 2) such that

U (Pj , fj , ) L(Pj , fj , ) < .

These inequalities persist if P1 and P2 are replaced by their common refinement P . Then
(20) implies
U (P, f, ) L(P, f, ) < 2,

106

which proves that f R(). With this same P we have

Z
U (P, fj , ) < fj d + (j = 1, 2);
hence (20) implies

f d 6 U (P, f, ) <

f1 d +

f2 d + 2.

Since is arbitrary, letting 0+ , we conclude that

Z
Z
Z
f d 6 f1 d + f2 d2 .

(21)

Replacing f1 and f2 in (21) by f1 and f2 , we have

Z
Z
Z
Z
Z
Z
(f ) d 6 (f1 ) d + (f2 ) d2 f d > f1 d + f2 d.
Combining the latter inequality with (21), the equality
Z
Z
Z
f d = f1 d + f2 d

is proved. The proofs of the other assertions of Theorem 6.12 are so similar that we omit
the details. In part (c) the point is that (by passing to refinements)
we may restrict
R
ourselves to partitions which contain the point c, in approximating f d.

6.13 Theorem. If f, g R() on [a, b], then
(a) f g R();
 R
R

 b
b
(b) |f | R() and  a f d 6 a |f | d.

Proof. (a) If we take (t) = t2 , Theorem 6.11 shows that (f g)2 R(). The identity
4f g = (f + g)2 (f g)2

then shows that f g R().

(b) If we Rtake (t) = |t|, Theorem 6.11 shows similarly that |f | R(). Choose c = 1
so that c f d > 0. Then

Z
Z
Z
Z


 f d = c f d = cf d 6 |f | d,


since cf 6 |f |.

6.14 Definition. The unit step function I is defined by

(
0 if x 6 0
I(x) =
.
1 if x > 0

107

6.15 Theorem. If a < s < b, f is bounded on [a, b], f is continuous at s, and (x) =
I(x s), then
Z b
f d = f (s).
a

Proof. Consider partitions P = {x0 , x1 , x2 , x3 }, where x0 = a, and x1 = s < x2 < x3 = b.

Then
U (P, f, ) = M2 , L(P, f, ) = m2 .
Since f is continuous at s, we see that M2 and m2 converge to f (s) as x2 s.


6.16 Theorem. Suppose cn > 0 for n = 1, 2, 3, . . .,

of distinct points in (a, b) and
(x) =

X
n=1

n cn

converges, {sn } is a sequence

cn I(x sn ).

(22)

Let f be continuous on [a, b]. Then

Z b

X
f (x) d =
cn f (sn ).
a

(23)

n=1

Proof. The comparison test shows that the series

P (22) converges for every x. Its sum (x)
is evidently monotonic, and (a) = 0, (b) =
cn . Let > 0 be given, and choose N so
that

X
cn < .
n=N +1

Put

1 (x) =

N
X
n=1

cn I(x sn ),

2 (x) =

n=N +1

By Theorems 6.12 and 6.15,

Z
Since 2 (b) 2 (a) < ,

f d1 =
a

N
X

cn I(x sn ).

cn f (sn ).

(24)

n=1

Z b




f d2  6 M ,


(25)

where M = sup |f (x)|. Since = 1 + 2 , it follows from (24) and (25) that
Z

N
 b

X


f
d

c
f
(s
)

n
n  < M .
 a

n=1

Letting N , we obtain (23).

108

6.17 Theorem. Assume increases monotonically and R on [a, b]. Let f be a

bounded real function on [a, b]. Then f R() if and only if f R. In that case
Z b
Z b
f (x) (x) dx.
(27)
f d =
a

Proof. Let > 0 be given and apply Theorem 6.6. There is a partition P = {x0 , x1 , . . . , xn }
of [a, b] such that
U (P, ) L(P, ) < .
(28)

The mean value theorem furnishes points ti [xi1 , xi ] such that

i = (ti )xi

for i = 1, 2, . . . , n. If si [xi1 , xi ], then

n
X
i=1

| (si ) (ti )|xi < ,

by (28) and Theorem 6.7(b). Put M = sup |f (x)|. Since

n
X

f (si )i =

i=1

n
X

f (si ) (ti )xi

i=1

it follows from (29) that

 n

n
X

X


f (si )i
f (si ) (si )xi  6 M .



i=1

i=1

In particular,

n
X

f (si )i 6 U (P, f ) + M ,

i=1

for all choices of si [xi1 , xi ], so that

U (P, f, ) 6 U (P, f ) + M .

The same argument leads from (30) to

U (P, f ) 6 U (P, f, ) + M .
Thus
|U (P, f, ) U (P, f )| 6 M .

(31)

Now note that (28) remains true if P is replaced by any refinement. Hence (31) also
remains true. We conclude that

Z b
Z b




f d
f (x) (x) dx 6 M .

a

But is arbitrary. Hence

f d =
a

f (x) (x) dx,

(38)

for any bounded f . The equality of the lower integrals follows from (30) in exactly the
same way. The theorem follows.


109

6.18 Remark. The two preceding theorems illustrate the generality and flexibility which
are inherent in the Stieltjes process of integration. If is a pure step function of the form
(22), the integral reduces to a finite or infinite series. If has an integrable derivative,
the integral reduces to an ordinary Riemann integral. This makes it possible in many
cases to study series and integrals simultaneously, rather than separately.

6.19 Theorem (Change of variable). Suppose : [A, B] [a, b] is a strictly increasing

continuous function. Suppose is monotonically increasing on [a, b] and f R() on
[a, b]. Define and g on [A, B] by
(y) = ((y)),
Then g R() and

g(y) = f ((y)).

g d =
A

(36)

f d.

(37)

Proof. To each partition P = {x0 , x1 , . . . , xn } of [a, b] corresponds a partition Q =

{y0 , y1 , . . . , yn } of [A, B], so that xi = (yi ). All partitions of [A, B] are obtained in
this way. Since the values taken by f on [xi1 , xi ] are exactly the same as those taken by
g on [yi1 , yi ], we see that
U (Q, g, ) = U (P, f, ),

L(Q, g, ) = L(P, f, ).

(38)

Since
f R(), P can be chosen so that both U (P, f, ) and L(P, f, ) are close to
R
f d. Hence (38), combined with Theorem 6.6, shows that g R() and that (37)
holds. This completes the proof.


Let us note the following special case: Take (x) = x. Then = . Assume R on
[A, B]. If Theorem 6.17 is applied to the left side of (37), we obtain
Z B
Z b
f ((y)) (y) dy.
f (x) dx =
a

Integration and differentiation

Integration and differentiation are, in a certain sense, inverse operations.

6.20 Theorem. Let f R on [a, b]. For a 6 x 6 b, put

Z x
F (x) =
f (t) dt.
a

Then F is continuous on [a, b]; furthermore, if f is continuous at a point x0 of [a, b], then
F is differentiable at x0 , and F (x0 ) = f (x0 ).

110

Proof. Since f R, f is bounded. Suppose |f (t)| 6 M for a 6 t 6 b. If a 6 x < y 6 b,

then
Z y



|F (y) F (x)| = 
f (t) dt 6 M (y x),
x

by Theorem 6.12(c)(d). Given > 0, we see that

|F (y) F (x)| < .

|y x| <
M
This proves the (uniform) continuity of F . Now suppose that f is continuous at x0 . Given
> 0, choose > 0 such that
|f (t) f (x0 )| <
if |t x0 | < and a 6 t 6 b. Hence, if

x0 < s 6 x0 6 t < x0 +

and a 6 s < t 6 b,

we have, by Theorem 6.12(d),



 
Z t
 F (t) F (s)

  1

 < .
=
[f
(u)

f
(x
)]
du

f
(x
)
0
0


 t s
ts
s
It follows that F (x0 ) = f (x0 ).

The second assertion of Theorem 6.20 can be interpreted as follows. If f is continuous at

a point x [a, b], then
Z x
d
f (t) dt = f (x).
(*)
dx a
(*) is usually referred to as the first fundamental theorem of calculus.

6.21 The (second) fundamental theorem of calculus. If f R on [a, b] and if there

is a differentiable function F on [a, b] such that F = f , then
Z b
f (x) dx = F (b) F (a).
(**)
a

Proof. Let > 0 be given. Choose a partition P = {x0 , x1 , . . . , xn } of [a, b] so that

U (P, f ) L(P, f ) < . The mean value theorem furnishes points ti [xi1 , xi ] such that
F (xi ) F (xi1 ) = f (ti )xi
for i = 1, 2, . . . , n. Thus
n
X
i=1

f (ti )xi = F (b) F (a).

It now follows from Theorem 6.7(c) that



Z b


F (b) F (a)
f (x) dx < .

a

Since this holds for every > 0, the proof is complete.

111

(**) is also known as the Newton-Leibniz formula. This is exactly the way in which we
calculate definite integral: first find an antiderivative of f , then substitute the limits of
integration.
6.22 Theorem (Integration by parts). Suppose f and g are differentiable functions
on [a, b], f R, and g R. Then
Z b
Z b

f (x)g(x) dx.
f (x)g (x) dx = f (b)g(b) f (a)g(a)
a

Proof. Put h(x) = f (x)g(x) and apply Theorem 6.21 to h and its derivative. Note that
h R, by Theorem 6.13.

The rule of thumb for applying integration by parts can be stated as follows:
1. RThe part selected as g must be readilyRintegrable.
2. f g must not be more complex than f g .

Example. Find

x3 ex dx.
2

Solution. Let f = x2 and g = xex = ( 12 ex ) . Then

Z
Z
1 2
1
1 2
2
3 x2
2 1 x2
x e dx = x ( e ) (2x) ex dx = x2 ex ex + C.
2
2
2
2
2

If f and g are chosen in the other way, i.e., f = ex and g = x3 , then

Z
Z
Z
1 4 x2 1
2
x2 1 4
3 x2
x2 1 4
x5 ex dx,
x e dx = e ( x ) (2xe )( x ) dx = x e
4
4
4
2
R 3 x2
which is more complicated than the original integral x e dx. Thus, the latter choice
of f and g is not recommended.


112

Sequences and series of functions

Reference: W. Rudin, Principles of Mathematical Analysis, 3rd edition, McGraw-Hill,
c1976, Chapter 7.
7.1 Definition. Suppose that {fn } be a sequence of functions defined on a set E such
that the sequence of numbers {fn (x)} converges for every x E. Define a function f by
f (x) = lim fn (x) (x E).

(1)

The sequence {fn } of functions converges to f pointwise on E if (1) holds. Similarly, if

P
n fn (x) converges for every x E and if f is defined as
f (x) =

X
n=1

fn (x) (x E),

the function f is called the sum of the series

(2)

fn .

Here, the main problem is to determine whether important properties of fn are preserved
under the limit operations (1) and (2).
For example, if the functions fn are continuous, is the limit function f continuous? If the
limit function f is continuous at a point x, then
lim lim fn (t) = lim f (t) = f (x) = lim fn (x) = lim lim fn (t),

tx n

tx

n tx

i.e., the limit processes can be interchanged. The same question can be asked if fn are
differentiable or integrable.
7.2 Example. For m, n = 1, 2, 3, . . ., let
m
.
m+n
Then for fixed n, lim sm,n = 1 so that lim lim sm,n = 1. On the other hand, for every
m
n m
fixed m, lim sm,n = 0 so that lim lim sm,n = 0. In this example, the order in which the
n
m n
limits are evaluated is important.

sm,n =

7.3 Example. For x R, m = 0, 1, 2, . . ., let

fn (x) =
and consider
f (x) =

x2
(1 + x2 )n

fn (x) =

n=0

It is easy to see that

f (x) =

X
n=0

0
1 + x2

x2
.
(1 + x2 )n

if x = 0
if x =
6 0

so that a convergent series of continuous functions may have a discontinuous sum.

113

7.4 Example. For m = 1, 2, 3, . . ., put

fm (x) = lim (cos m!x)2n .
n

When m!x is an integer, fm (x) = 1. For all other values of x, fm (x) = 0. Now let
f (x) = lim fm (x).
m

For x R \ Q, fm (x) = 0 for every m so that f (x) = 0. For x Q, say x = pq , where

p, q Z and q 6= 0, m!x Z if m > q so that f (x) = 1. Hence,
(
0 if x R \ Q
lim lim (cos m!x)2n =
,
m n
1 if x Q
which is everywhere discontinuous!

7.5 Example. For x R, n = 1, 2, 3, . . ., let

sin nx
fn (x) =
n

and f (x) = lim fn (x) = 0. Then f (x) = 0 and fn (x) =

n

converge to f . For instance,

fn (0) =

as n , whereas f (0) = 0.

n cos nx so that {fn } does not

n +

7.6 Example. For x [0, 1] and n = 1, 2, 3, . . ., let

fn (x) = n2 x(1 x2 )n .

For 0 < x 6 1, we have

(10)

lim fn (x) = 0

by Theorem 3.20(d) . Since fn (0) = 0, we see that

lim fn (x) = 0 (0 6 x 6 1).

It is routine to see that

Thus, in spite of (11),

1
0

x(1 x2 )n dx =

1
.
2n + 2

n2
+
2n + 2
0
as n . If in (10), n2 is replaced by n, (11) still holds, but we now have
Z 1
1
n
= ,
lim
fn (x) dx = lim
n 0
n 2n + 2
2
whereas

Z 
Z

(11)

fn (x) dx =

lim fn (x) dx = 0.

Thus, the limit of the integral need not be equal to the integral of the limit, even if both
are finite.


114

n
= 0.
Theorem 3.20(d) states that if p > 0 and is real, then lim
n (1 + p)n

Uniform convergence
7.7 Definition. A sequence of functions {fn }, n = 1, 2, 3, . . ., converges uniformly on E
to a function f if for every > 0, there is an integer N such that n > N implies
|fn (x) f (x)| <

(12)

for all x E.
It is clear that every uniformly convergent sequence is pointwise convergent. If {fn }
converges pointwise on E, then there exists a function f such that for every > 0 and
for every x E, there exists an integer N , depending on and x, such that (12) holds if
n > N ; if {fn } converges uniformly on E, then it is possible, for every > 0, to find one
integer N that will do for all x E.
The series
by

fn (x) converges uniformly on E if the sequence {sn } of partial sums defined

n
X

fi (x) = sn (x)

i=1

converges uniformly on E.

7.8 Theorem. (Cauchy criterion for uniform convergence) The sequence of functions
{fn }, defined on E, converges uniformly on E if and only if for every > 0 there exists
an integer N such that m, n > N , x E, implies
|fn (x) fm (x)| < .

(13)

Proof. Let f be the limit function.

There exists an integer N such that n > N , x E implies

|fn (x) f (x)| <

2
so that
|fn (x) fm (x)| 6 |fn (x) f (x)| + |f (x) fm (x)| <

+ =
2 2

if m, n > N and x E.

The sequence {fn (x)} converges, for every x E, to the limit function f (x).

Let > 0 be given and choose N such that (13) holds. Fix n and let m in (13).
Since fm (x) f (x) as m , this gives
|fn (x) f (x)| 6
for every n > N and every x E, which completes the proof.

(14)


115

7.9 Theorem. Suppose lim fn (x) = f (x), where x E. Put Mn = sup |fn (x) f (x)|.
n

xE

Then fn f uniformly on E if and only if Mn 0 as n .

7.10 Theorem. (Weierstrass M -test) Suppose {fn } is a sequence of functions defined
on E, and suppose
|fn (x)| 6 Mn (x E, n = 1, 2, 3, . . .).
P
P
Then n fn converges uniformly on E if n Mn converges.
P
Proof. If n Mn converges, then for arbitrary > 0,
 m

m
X
 X


fi (x) 6
Mi 6 (x E),



i=n

i=n

provided m < n are large enough. Uniform convergence now follows from Theorem
7.8.

Uniform convergence and continuity

7.11 Theorem. Suppose fn f uniformly on a set E in a metric space. Let x be a

limit point of E, and suppose
lim fn (t) = An

tx

Then {An } converges, and

(n = 1, 2, 3, . . .).

(15)

lim f (t) = lim An ,

(16)

lim lim fn (t) = lim lim fn (t).

(17)

tx

i.e.,
tx n

n tx

Proof. Let > 0 be given. By the uniform convergence of {fn }, there exists N such that
n, m > N and t E imply
|fn (t) fm (t)| 6 .
(18)
Letting t x in (18), we obtain
|An Am | 6

for n, m > N ,

i.e., {An } is a Cauchy sequence and therefore converges, say to A. Choose n large enough
so that

(20)
|f (t) fn (t)| 6
3
for all t E (this is possible by the uniform convergence) and that

|An A| 6 .
(21)
3
Then, for this n, we choose a neighborhood V of x such that

(22)
|fn (t) An | 6
3
if t V E and t 6= x.

116

Thus, for t V E and t 6= x,

|f (t) A| 6 |f (t) fn (t)| + |fn (t) An | + |An A|

6 + + = .
3 3 3
This is equivalent to (16).

7.12 Theorem. If {fn } is a sequence of continuous functions on E, and if fn f

uniformly on E, then f is continuous on E.
7.13 Theorem. Suppose K is compact, and
(a) {fn } is a sequence of continuous functions on K,
(b) {fn } converges pointwise to a continuous function f on K,
(c) fn (x) > fn+1 (x) for all x K, n = 1, 2, 3, . . ..
Then fn f uniformly on K.
Proof. Put gn = fn f . Then gn is continuous, gn 0 pointwise, and gn > gn+1 . It
remains to prove that gn 0 uniformly on K.
Let > 0 be given. Let Kn = {x K : gn (x) > }. Since gn is continuous and [, ) is
closed,
Kn = gn1 ([, )) K
is closed, hence compact. Since gn > gn+1 , we have T
Kn Kn+1 . Fix x TK. Since
gn (x) 0, x Kn if n is sufficiently large. Thus, x Kn . In other words, Kn = .
Hence, KN = for some N . It follows that 0 6 gn (x) < for all x K and for all
n > N , and the theorem is proved.

The compactness is really needed. Consider the sequence of functions {fn }, where
1
fn (x) =
(0 < x < 1; n = 1, 2, 3, . . .).
nx + 1
Then fn (x) 0 monotonically in (0, 1): for fixed x (0, 1),
1
= 0;
lim fn (x) = lim
n
n nx + 1
the convergence is not uniform: for fixed x (0, 1) and for each n,
1
n
fn (x) =
< 0 Mn = sup fn (x) = lim+
= 1,
2
x0 nx + 1
(nx + 1)
x(0,1)
but Mn 6 0 as n .

7.14 Definition. Let X be a metric space. Denote by C (X) the set of all complexvalued, continuous, bounded functions with domain X. Associate with each f C (X)
its supremum norm
kf k = sup |f (x)|.
xX

Since f is bounded, kf k < . It is obvious that kf k = 0 f = 0. If h = f + g, then

for all x X,
|h(x)| 6 |f (x)| + |g(x)| 6 kf k + kgk.

117

Taking sup of the left side then yields

kf + gk = khk 6 kf k + kgk.

Define the distance between f, g C (X) to be

d (f, g) = kf gk.

The pair (C (X), d ) is a metric space.

Theorem 7.9 can be rephrased as: A sequence {fn } converges to f in (C (X), d ) if and
only if fn f uniformly on X.
Accordingly, closed subsets of C (X) are sometimes called uniformly closed, the closure of
a set A C (X) is called its uniform closure, and so on.
7.15 Theorem. (C (X), d ) is a complete metric space.
Proof. Let {fn } be a Cauchy sequence in C (X).
For any given > 0, there exists N such that kfn fm k < if n, m > N .
It follows that there exists a function f with domain X such that fn f uniformly.
Since f is the uniform limit of a sequence of continuous functions, f is continuous.
Moreover, f is bounded because there exists an n such that for all x,
|f (x) fn (x)| < 1 |f (x)| < 1 + |fn (x)| 6 1 + Mn ,

where Mn is an upper bound of fn . Thus, f C (X), and since fn f uniformly on X,

kfn f k 0 as n .


118

Uniform convergence and integration

7.16 Theorem. Let be monotonically increasing on [a, b]. Suppose fn R() on [a, b],
for n = 1, 2, 3, . . ., and suppose fn f uniformly on [a, b]. Then f R() on [a, b], and
Z b
Z b
f d = lim
fn d.
(23)
n

Proof. It suffices to prove this for real f . Put

n = sup |fn (x) f (x)|.

(24)

a6x6b

Then
fn n 6 f 6 fn + n ,
so that the upper and lower integrals of f satisfy
Z b
Z b
Z b
Z b
f d 6
f d 6
(fn + n ) d.
(fn n ) d 6
a

Hence,

06

b
a

f d

(25)

b
a

f d 6 2n [(b) (a)].

Since n 0 as n , the upper and lower integrals of f are equal, i.e., f R().
Another application of (25) now yields
Z b

Z b



(26)
f d
fn d 6 n [(b) (a)].

a

Since n 0 as n , (23) follows.

Corollary. If fn R() on [a, b] and if

X
f (x) =
fn (x) (a 6 x 6 b),
n=1

the series converging uniformly on [a, b], then

Z b
Z b
X
f d =
fn d.
a

n=1

In other words, the series may be integrated term by term.

119

Uniform convergence and differentiation

7.17 Theorem. Suppose {fn } is a sequence of functions, differentiable on [a, b] and
such that {fn (x0 )} for some x0 [a, b]. If {fn } converges uniformly on [a, b], then {fn }
converges uniformly on [a, b] to a function f and
f (x) = lim fn (x) (a 6 x 6 b).
n

(27)

Proof. Let > 0 be given. Choose N such that m, n > N implies

(28)
|fn (x0 ) fm (x0 )| <
2
and

|fn (t) fm
(t)| <
(a 6 t 6 b).
(29)
2(b a)
Applying the mean value theorem to fn fm , (29) shows that
|x t|

6
(30)
|fn (x) fm (x) fn (t) + fm (t)| 6
2(b a)
2
for any x, t [a, b] if m, n > N . By (28) and (30), we have for x [a, b] and m, n > N ,
|fn (x) fm (x)| 6 |fn (x) fm (x) fn (x0 ) + fm (x0 )| + |fn (x0 ) fm (x0 )|

< +
2 2
=
so that {fn } converges uniformly on [a, b]. Let

f (x) = lim fn (x) (a 6 x 6 b).

n

Fix now a point x [a, b] and define

fn (t) fn (x)
n (t) =
,
tx
for a 6 t 6 b, t 6= x. Then

(t) =

f (t) f (x)
tx

lim n (t) = fn (x) (n = 1, 2, 3, . . .).

tx

(31)

(32)

From (30), we have

|fm (t) fm (x) fn (t) + fn (x)|

6
(m, n > N )
|t x|
2(b a)
so that {n } converges uniformly for t 6= x. Since {fn } converges to f , we conclude from
(31) that
lim n (t) = (t)
|m (t) n (t)| =

uniformly for t [a, b] \ {x}. Applying now Theorem 7.11 to {n }, (32) and (33) show
that
lim (t) = lim fn (x),
tx

which is (27).

120

7.18 Theorem. There exists a real continuous function on R which is nowhere differentiable.
Proof. Define
(x) = |x|
(1 6 x 6 1)
and extend the definition of (x) to all real x by requiring that
(x + 2) = (x).

(34)
(35)

Then, for all s, t,

|(s) (t)| 6 |s t|.
In particular, is continuous on R. Define

X
3
f (x) =
( )n (4n x).
4
n=0

(36)

(37)

Since 0 6 6 1, Theorem 7.10 shows that the series (37) converges uniformly on R. By
Theorem 7.12, f is continuous on R.
Now fix an x R and a positive integer m. Put
1
,
(38)
m =
2 4m
where the sign is so chosen that no integer lies between 4m x and 4m (x + m ). This can be
done, since 4m |m | = 21 . Define
(4n (x + m )) (4n )
.
(39)
m
When n > m, then 4n m is an even integer so that n = 0. When 0 6 n 6 m, (36) implies
that |n | 6 4n . Since m = 4m , we conclude that


  m
m1

X
 f (x + m ) f (x)  X
3m + 1
3
n

 =  ( )n n  > 3m
3
=
.

 

m
4
2
n =

n=0

n=0

As m , m 0. It follows that f is not differentiable at x.

121

For n = 0, 1, 2 . . . , n, define fn (x) = (4n x). The graphs of y = f0 (x) and y = f1 (x) are
as follows:
y
1

y = f0 (x)

1
y
1
4

14

y = f1 (x)

1
4

As n increases,
fn (x) increases. Since
P 3 nthe number of saw teeth in the graph of y =
k
f (x) = n=0 ( 4 ) fn (x) and the set of locations of saw teeth { 4n : k Z, n N {0}} is
dense in R, the nowhere differentiability of f follows.

122

Equicontinuous families of functions

In Theorem 3.6 we saw that every bounded sequence of complex numbers contains a
convergent subsequence, and the question arises whether something similar is true for
sequences of functions. To make the question more precise, we shall define two kinds of
boundedness.
7.19 Definition. Let {fn } be a sequence of functions defined on a set E. {fn } is pointwise
bounded on E if the sequence {fn (x)} is bounded for every x E, that is, if there exists
a finite-valued function defined on E such that
|fn (x)| < (x) (x E, n = 1, 2, 3, . . .);

{fn } is uniformly bounded on E if there exists a number M such that

|fn (x)| < M

(x E, n = 1, 2, 3, . . .).

If {fn } is pointwise bounded on E and E1 is a countable subset of E, it is always possible

to find a subsequence {fnk } such that {fnk (x)} converges for every x E1 by the diagonal
process which is used in the proof of Theorem 7.23.
However, even if {fn } is a uniformly bounded sequence of continuous functions on a
compact set E, there need not exist a subsequence which converges pointwise on E.
7.20 Example. Let
fn (x) = sin nx (0 6 x 6 2, n = 1, 2, 3, . . .).
Suppose there exists a sequence {nk } such that {sin nk x} converges, for every x [0, 2].
In that case we must have
lim (sin nk x sin nk+1 x) = 0 (0 6 x 6 2);

hence
lim (sin nk x sin nk+1 x)2 = 0 (0 6 x 6 2).

By Lebesgues dominated convergence theorem, (40) implies

Z 2
lim
(sin nk x sin nk+1 x)2 = 0.
k

(40)

(41)

But a simple calculation shows that

Z 2
(sin nk x sin nk+1 x)2 = 2,
0

which contradicts (41).

Another question is whether every convergent sequence contains a uniformly convergent

subsequence. Our next example will show that this need not be so, even if the sequence
is uniformly bounded on a compact set. (Example 7.6 shows that a sequence of bounded
functions may converge without being uniformly bounded; but it is trivial to see that
uniform convergence of a sequence of bounded functions implies uniform boundedness.)

123

7.21 Example. Let

fn (x) =

x2
x2 + (1 nx)2

(0 6 x 6 1, n = 1, 2, 3, . . .).

Then |fn (x)| 6 1, so that {fn } is uniformly bounded on [0, 1]. Also
lim fn (x) = 0 (0 6 x 6 1),

but

1
fn ( ) = 1 (n = 1, 2, 3, . . .),
n
so that no subsequence can converge uniformly on [0, 1].

The concept which is needed in this connection is that of equicontinuity; it is given in the
following definition.
7.22 Definition. A family F of complex functions f defined on a set E in a metric
space X is said to be equicontinuous on E if for every > 0 there exists a > 0 such that
|f (x) f (y)| <
whenever d(x, y) < , x, E, and f F . Here d denotes the metric of X.
It is clear that every member of an equicontinuous family is uniformly continuous.
The sequence of Example 7.21 is not equicontinuous.
Theorems 7.24 and 7.25 will show that there is a very close relation between equicontinuity,
on the one hand, and uniform convergence of sequences of continuous functions, on the
other. But first we describe a selection process which has nothing to do with continuity.
7.23 Theorem. If {fn } is a pointwise bounded sequence of complex functions on a
countable set E, then {fn } has a subsequence {fnk } such that {fnk (x)} converges for
every x E.
Proof. Let {xi }, i = 1, 2, 3, . . ., be the points of E, arranged in a sequence. Since
{fn (x1 )} is bounded, there exists a subsequence, which we shall denote by {f1,k }, such
that {f1,k (x1 )} converges as k .
Let us now consider sequences S1 , S2 , S3 , . . ., which we represent by the array
S1 :
S2 :
S3 :

f1,1
f2,1
f3,1

f1,2
f2,2
f3,2

f1,3
f2,3
f3,3

f1,4
f2,4
f3,4

and which have the following properties:

(a) Sn is a subsequence of Sn1 for n = 2, 3, 4, . . ..
(b) {fn,k (xn )} converges, as k (the boundedness of {fn (xn )} makes it possible
to choose Sn in this way);

124

(c) The order in which the functions appear is the same in each subsequence; i.e., if
one function precedes another in S1 , they are in the same relation in every Sn ,
until one or the other is deleted. Hence, when going from one row in the above
array to the next below, functions may move to the left but never to the right.
We now go down the diagonal of the array; i.e., we consider the sequence
S : f1,1 , f2,2 , f3,3 , f4,4 , .

By (c), the sequence S (except possibly its first n 1 terms) is a subsequence of Sn ,

for n = 1, 2, 3, . . .. Hence (b) implies that {fn,n (xi )} converges, as n , for every
xi E.

7.24 Theorem. If K is a compact metric space, if fn C (K) for n = 1, 2, 3, . . ., and if
{fn } converges uniformly on K, then {fn } is equicontinuous on K.
Proof. Let > 0 be given. Since {fn } converges uniformly, there is an integer N such
that
kfn fN k < (n > N ).
(42)
Since continuous functions are uniformly continuous on compact sets, there is a > 0
such that
|fi (x) fi (y)| <
(43)
if 1 6 i 6 N and d(x, y) < . So, if n > N and d(x, y) < , it follows that
|fn (x) fn (y)| 6 |fn (x) fN (x)| + |fN (x) fN (y)| + |fN (y) fn (y)| < 3.

In conjunction with (43), this proves the theorem.

7.25 Theorem. (Arzel`a-Ascoli) If K is compact, if fn C (K) for n = 1, 2, 3, . . ., and if

{fn } is pointwise bounded and equicontinuous on K, then
(a) {fn } is uniformly bounded on K,
(b) {fn } contains a uniformly convergent subsequence.
Proof. (a) Let > 0 be given and choose > 0, in accordance with Definition 7.22, so
that
|fn (x) fn (y)| <
(44)
for all n, provided that d(x, y) < .
Since K is compact, there are finitely many points p1 , . . . , pr K such that K
S
r
i=1 B (pi ) and every x K corresponds at least one pi with d(x, pi ) < .
Since {fn } is pointwise bounded, there exist Mi < such that |fn (pi )| < Mi for all n. If
M = max Mi , then |fn (x)| < M + for every x K. This proves (a).
16i6r

(b) Let E be a countable dense subset of K. Theorem 7.23 shows that {fn } has a
subsequence {fni } such that {fni (x)} converges for every x E. Put fni = gi to simplify
the notation. We shall prove that {gi } converges uniformly on K.
Let > 0, and pick > 0 as in the beginning of this proof. Let
V (x) = {y K : d(x, y) < }.

125

Since E is dense in K, and K is compact, there are finitely many points x1 , . . . , xm E

such that
K V (x1 ) V (xm ).
(45)
Since {gi (x)} converges for every x E, there is an integer N such that
|gi (xs ) gj (xs )| <

(46)

whenever i, j > N , 1 6 s 6 m. If x K, (45) shows that x V (xs ) for some s, so that

|gi (x) gi (xs )| <

for every i. If i, j > N , it follows from (46) that

|gi (x) gj (x)| 6 |gi (x) gi (xs )| + |gi (xs ) gj (xs )| + |gj (xs ) gj (x)| < 3.

This completes the proof.