Analysis (Lecture Notes by Ali Nesin)

Analysis Notes
(only a draft, and the rst one!)

Ali Nesin
Mathematics Department
Istanbul Bilgi University
Kustepe Sisli
Istanbul Turkey
[email protected]
June 22, 2004
2
Contents
1 Preliminaries 9
1.1 Binary Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2 Binary Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Real Numbers 13
2.1 Axioms for Addition . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Axioms for Multiplication . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Distributivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 Axioms for the Order Relation . . . . . . . . . . . . . . . . . . . 19
2.5 Totally Ordered Sets . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.6 Completeness Axiom . . . . . . . . . . . . . . . . . . . . . . . . . 23
3 Other Number Sets 27
3.1 Natural Numbers and Induction . . . . . . . . . . . . . . . . . . 27
3.1.1 Exponentiation . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.2 Factorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.3 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Integers and Rational Numbers . . . . . . . . . . . . . . . . . . . 34
3.2.1 Exponentiation . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3 Uniqueness of the Real Number System . . . . . . . . . . . . . . 36
3.4 Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4 Real Vector Spaces 43
5 Metric Spaces 49
5.1 Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2 Denition and Further Examples . . . . . . . . . . . . . . . . . . 50
5.3 Normed Real Vector Spaces and Banach Spaces and Algebras . . 52
5.4 Open Subsets of a Metric Space . . . . . . . . . . . . . . . . . . . 54
6 Sequences and Limits 57
6.1 Denition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.2 Examples of Convergence in R and C . . . . . . . . . . . . . . . . 60
6.2.1 The Sequence (1/n)
n
. . . . . . . . . . . . . . . . . . . . . 61
3
4 CONTENTS
6.2.2 The Sequence (
n
)
n
. . . . . . . . . . . . . . . . . . . . . 61
n
/n!)
n
. . . . . . . . . . . . . . . . . . . 62
6.3 Convergence and the Order . . . . . . . . . . . . . . . . . . . . . 63
6.4 Convergence and the Four Operations . . . . . . . . . . . . . . . 64
6.5 More On Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.6 Cauchy Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.7 Convergence of Real Cauchy Sequences . . . . . . . . . . . . . . . 69
6.8 Convergence of Some Sequences . . . . . . . . . . . . . . . . . . . 73
6.8.1 The Sequence ((1 + 1/n)
n
)
n
. . . . . . . . . . . . . . . . . 73
6.8.2 The Sequences (2
1/n
)
n
and (n
1/n
)
n
. . . . . . . . . . . . . 74
6.8.3 The Sequence (nx
n
)
n
. . . . . . . . . . . . . . . . . . . . 74
6.9 Divergence to Innity . . . . . . . . . . . . . . . . . . . . . . . . 76
6.10 Limit Superior and Inferior . . . . . . . . . . . . . . . . . . . . . 78
6.11 Complete Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . 79
6.12 Completion of a Metric Space . . . . . . . . . . . . . . . . . . . . 80
6.13 Supplementary Problems . . . . . . . . . . . . . . . . . . . . . . . 82
7 Series 83
7.1 Denition and Examples . . . . . . . . . . . . . . . . . . . . . . . 83
7.2 Easy Consequences of the Denition . . . . . . . . . . . . . . . . 86
7.3 Absolute Convergence . . . . . . . . . . . . . . . . . . . . . . . . 87
7.4 Alternating Series . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7.5 Criteria for Convergence . . . . . . . . . . . . . . . . . . . . . . . 89
7.6 Supplementary Problems . . . . . . . . . . . . . . . . . . . . . . . 92
7.6.1 Midterm of Math 152 . . . . . . . . . . . . . . . . . . . . 93
7.6.2 Solutions of the Midterm of Math 152 . . . . . . . . . . . 94
8 Supplementary Topics 97
8.1 Liouville Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . 97
9 Convergence of Functions 101
9.1 Pointwise Convergence of a Sequence of Functions . . . . . . . . 101
9.2 Uniform Convergence of a Sequence of Functions . . . . . . . . . 103
9.3 Uniform Convergence of a Series of Functions . . . . . . . . . . . 107
9.4 Uniform Convergence and Metric . . . . . . . . . . . . . . . . . . 109
9.5 Limits of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 109
9.6 Convergence of a Family of Functions . . . . . . . . . . . . . . . 110
9.7 Supplementary Topics . . . . . . . . . . . . . . . . . . . . . . . . 111
10 Topological Spaces 113
10.2 Closed Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
10.3 Interior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
10.4 Closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
10.5 Base of a Topology . . . . . . . . . . . . . . . . . . . . . . . . . . 116
10.6 Compact Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
CONTENTS 5
10.7 Convergence and Limit Points . . . . . . . . . . . . . . . . . . . . 119
10.8 Connected Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
11 Continuity 121
11.1 Continuity on Metric Spaces . . . . . . . . . . . . . . . . . . . . . 121
11.2 Continuity on Topological Spaces . . . . . . . . . . . . . . . . . . 123
11.3 Continuous Functions and R . . . . . . . . . . . . . . . . . . . . . 127
11.4 Uniform Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . 128
11.5 Uniform Convergence and Continuity . . . . . . . . . . . . . . . . 129
11.6 Supplementary Topics . . . . . . . . . . . . . . . . . . . . . . . . 129
11.6.1 A Continuous Curve Covering [0, 1]
2
. . . . . . . . . . . . 129
12 Dierentiable Functions 131
12.2 Dierentiation of Complex Functions . . . . . . . . . . . . . . . . 132
12.3 Basic Properties of Dierentiable Functions . . . . . . . . . . . . 132
12.4 Rules of Dierentiation . . . . . . . . . . . . . . . . . . . . . . . 133
12.5 Relationship Between a Function and Its Derivative . . . . . . . 133
12.6 Uniform Convergence and Dierentiation . . . . . . . . . . . . . 134
12.7 Second and Further Derivatives . . . . . . . . . . . . . . . . . . . 136
13 Analytic Functions 137
13.1 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
13.2 Taylor Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
13.2.1 Calculating Taylor Polynomials . . . . . . . . . . . . . . . 140
13.3 Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 142
13.4 Transcendental Functions . . . . . . . . . . . . . . . . . . . . . . 143
13.4.1 Exponentiation and Trigonometric Functions . . . . . . . 143
13.4.2 Inverse Trigonometric Functions . . . . . . . . . . . . . . 146
13.4.3 Logarithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
13.4.4 Hyperbolic Functions . . . . . . . . . . . . . . . . . . . . 146
13.5 Supplement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
13.5.1 Trigonometric Functions . . . . . . . . . . . . . . . . . . . 146
13.5.2 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
13.6 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
14 Graph Drawing 149
14.1 Drawing in Cartesian Coordinates . . . . . . . . . . . . . . . . . 149
14.1.1 Asymptotes . . . . . . . . . . . . . . . . . . . . . . . . . . 149
14.2 Parametric Equations . . . . . . . . . . . . . . . . . . . . . . . . 149
14.3 Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . 149
14.4 Geometric Loci . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
14.5 Supplement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
14.5.1 Lipschitz Condition . . . . . . . . . . . . . . . . . . . . . 151
14.5.2 A Metric On R
n
. . . . . . . . . . . . . . . . . . . . . . . 151
6 CONTENTS
15 Riemann Integral 155
15.2 Fundamental Theorem of Calculus . . . . . . . . . . . . . . . . . 155
15.3 How To Integrate? . . . . . . . . . . . . . . . . . . . . . . . . . . 155
15.3.1 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . 155
15.3.2 Trigonometric Functions . . . . . . . . . . . . . . . . . . . 155
15.4 Integration of Complex Function . . . . . . . . . . . . . . . . . . 156
15.4.1 Functions Dened by Integration . . . . . . . . . . . . . . 156
15.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
15.5.1 Application to Series . . . . . . . . . . . . . . . . . . . . . 156
15.5.2 Stirling Formula . . . . . . . . . . . . . . . . . . . . . . . 159
15.5.3 Eulers Function . . . . . . . . . . . . . . . . . . . . . . 161
16 Supplements 165
16.1 Stone-Weierstrass Theorem . . . . . . . . . . . . . . . . . . . . . 165
17 Fourier Series 167
17.1 Hilbert Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
17.2 Fourier Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
18 Topological Spaces (continued) 169
18.1 Product Topology . . . . . . . . . . . . . . . . . . . . . . . . . . 169
18.2 Homeomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
18.3 Sequences in Topological Spaces . . . . . . . . . . . . . . . . . . 170
18.4 Sequential Compactness . . . . . . . . . . . . . . . . . . . . . . . 171
18.5 Supplements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
18.5.1 T
0
-Identication . . . . . . . . . . . . . . . . . . . . . . . 171
19 Exams 173
19.1 Midterm Math 121 (November 2002) . . . . . . . . . . . . . . . . 173
19.2 Final Math 121 (January 2003) . . . . . . . . . . . . . . . . . . . 177
19.3 Resit of Math 121, February 2003 . . . . . . . . . . . . . . . . . . 179
19.4 Correction of the Resit of Math 121, February 2003 . . . . . . . . 181
19.5 Second Resit of Math 121, March 2003 . . . . . . . . . . . . . . . 184
19.6 Midterm of Math 152, April 2004 . . . . . . . . . . . . . . . . . . 186
19.7 Final of Math 152, June 2004 . . . . . . . . . . . . . . . . . . . . 186
CONTENTS 7
Foreword
This is a revolving and continuously
changing textbook. The present version
is far from complete. It may contain
several errors, omissions, oversights etc.
Do not circulate and consult it only with
great suspicion. The new version can be
found at www.alinesin.org.
Chapters 1 through 19 form the rst semester of a four semester course.
8 CONTENTS
Chapter 1
Preliminaries
1.1 Binary Operation
Let X be a set. A binary operation on X is just a function from X X into
X. The binary operations are often denoted by such symbols as +, , , ,
etc. The result of applying the binary relation to the elements x and y of X is
denoted as x +y, x y, x y, x y, x y etc.
Examples.
i. Let X be a set and let c X be any xed element. The rule x y = c
denes a binary operation on X. This binary operation satises xy = yx
for all x, y X (commutativity) and x(y z) = (xy)z (associativity).
ii. Let X be a set. The rule x y = x denes a binary operation on X.
Unless [X[ 1, this binary operation is not commutative. But it is always
associative.
iii. Let U be a set. Let X := (U) be the set of subsets of U. The rule
A B = A B denes a binary operation on X. This binary operation is
commutative and associative. Note that AU = U A = A for all A X.
Such an element is called the identity element of the binary operation.
The rule A B = A B denes another binary operation on X, which
also commutative and associative. is the identity element of this binary
operation. Examples i and ii do not have identity elements unless [X[ = 1.
iv. Let U be a set. Let X := (U) be the set of subsets of U. The rule
AB = AB denes a binary operation on X, which is neither associative
nor commutative in general. It does not have an identity element either,
although it has a right identity element, namely .
v. Let U be a set. Let X := (U) be the set of subsets of U. The rule
A B = (A B) (B A) denes a binary operation on X, which is
9
10 CHAPTER 1. PRELIMINARIES
commutative and associative (harder to check) and which has an identity
element. Every element in this example has an inverse element in the
sense that, if e denotes the identity element of X for this operation, then
for every x X there is a y X (namely y = x) such that xy = yx = e.
Exercises.
i. Let A be a set. Let X be the set of functions from A into A. For f, g X,
dene the function f g X by the rule
(f g)(a) = f(g(a))
for all a A. Show that this is a binary operation on X which is as-
sociative, noncommutative if [A[ > 1 and which has an identity element.
The identity element (the identity function) is denoted by Id
A
and it is
dened by the rule Id
A
(a) = a for all a A. Show that if [A[ > 1 then
not all elements of X have inverses.
ii. Let A be a set. Let Sym(A) be the set of bijections from A into A. For
f, g Sym(A), dene the function f g Sym(A) by the rule
(f g)(a) = f(g(a))
for all a A (as above). Show that this is a binary operation on X
which is associative, noncommutative if [A[ > 2 and which has an identity
element. Show that every element of Sym(A) has an inverse.
1.2 Binary Relations
Let X be a set. A binary relation on X is just a subset of X X.
Let R be a binary relation on X. Thus R X X. If (x, y) R, we will
write xRy. If (x, y) , R, we will write x ,Ry.
Binary relations are often denoted by such symbols as R, S, T, <, >, , ,
, _, , _, , , , , etc.
Examples.
i. R = X X is a binary relation on the set X. We have xRy for all
x, y X.
ii. R = is a binary relation on X. For this relation, x ,Ry for all x, y X.
iii. Let R := (X X) := (x, x) : x X. Then R is a binary relation on
X. We have xRy if and only if x = y.
iv. The set R := (x, y) X X : x y is a binary relation on X. Thus,
for all x, y X, (x, y) R if and only if x y.
1.2. BINARY RELATIONS 11
v. Let A and Y be two set and B A. Let X be the set of functions from
A into Y . For f, g X, set f g if and only if f(b) = g(b) for all b B.
This is a binary relation on X. It has the following properties:
Reexivity. For all f X, f f.
Symmetry. For all f, g X, if f g then f f.
Transitivity. For all f, g, h X, if f g and g h, then g h.
A relation satisfying the three properties above is called an equivalence
relation. The relation in Example iii is also an equivalence relation.
Exercises.
i. Let A and Y be two set. Let be a nonempty set of subsets of A satisfying
the following condition: For all B
1
, B
2
, there is a B
3
such that
B
3
B
1
B
2
. Let X be the set of functions from A into Y . For f, g X,
set f g if and only if there is a B such that f(b) = g(b) for all
b B. Show that this is an equivalence relation on X.
12 CHAPTER 1. PRELIMINARIES
Chapter 2
Real Numbers
We will dene the set of real numbers axiomatically. We will supply some
number of axioms (which are by denition statements that we accept without
proofs) and we will state that the set of real numbers is a set that satises
these axioms. We will neither ask nor answer the (important) question of the
existence of the set of real numbers. This question is part of Math 111. However
we will prove that the set of real numbers is unique in a sense to be made precise
(Theorem 3.3.1).
We will be interested just in two (binary) operations on R, called addition
and multiplication. Apart from these two operations, we will also be interested
in a (binary) relation <.
Our denition will take some time, till page 23.
Denition 2.0.1 A set R together with two binary operations + and , two
distinct constants 0 R and 1 R and a binary relation < is called a set of
real numbers if the axioms A1, A2, A3, A4, M1, M2, M3, M4, D, O1, O2,
O3, AO, MO and C (that will be stated in this chapter) hold.
The binary operation + is called addition, the binary operation is called
multiplication, the element 0 is called zero, the element 1 is called one, the
binary relation < is called the order relation. For two elements x, y R, x+y
will be called the sum of x and y; instead of x y we will often prefer to write
xy. xy will be called the product of x and y. If x < y, we will say that x is
less than y. We will use the expressions greater than, not less than etc.
freely.
We also dene the binary relations , > and as follows:
x y x < y or x = y
x > y y < x
x y x > y or x = y
The fact that 0 ,= 1 (which is explicitly stated in the denition) is important
and will be needed later. One cannot prove this fact from the rest of the axioms.
13
14 CHAPTER 2. REAL NUMBERS
Indeed, the set 0 satises all the axioms (take 1 to be 0) that we will state.
In fact, if 0 = 1, then there can be only one element in R, namely 0, because,
for any x R,
x
M2
= 1x = 0x
A2
= 0.
2.1 Axioms for Addition
We start with the axioms that involve only the addition.
A1. Additive Associativity. For any x, y, z R, x+(y +z) = (x+y) +z.
The rst axiom tells us that when adding, the parentheses are unnecessary.
For example, instead of (x + y) + z, we can just write x + y + z. Similarly,
instead of (x+y) +(z +t) or of ((x+y) +z) +t, we can just write x+y +z +t.
Although this fact (that the parentheses are useless) needs to be proven, we will
not prove it. The interested reader may look at Bourbaki.
A2. Additive Identity Element. For any x R, x + 0 = 0 +x = x.
A3. Additive Inverse Element. For any x R there is a y R such that
x +y = y +x = 0.
A set together with a binary operation, say +, and an element denoted 0
that satises the above axioms is called a group. Thus (R, +, 0) is a group.
Below, investigating the structure (R, +, 0), we will in fact investigate only the
properties of a group.
Note that the element y of A3 depends on x. Note also that A3 does not
tell us that x + y = y + x for all x, y R, it only tells us that it is so only for
the specic pair x and y.
We also note that 0 is the only element that satises A2; indeed if 0
1
also
satises A2, then 0 = 0 + 0
1
= 0
1
.
We now prove our rst result:
Lemma 2.1.1 Given x R, the element y as in A3 is unique.
Proof: Let x R. Let y be as in A3. Let y
1
satisfy the equation x + y
1
= 0.
We will show that y = y
1
, proving more than the statement of the lemma. We
start:
y
A2
= y + 0 = y + (x +y
1
)
A1
= (y +x) +y
1
A3
= 0 +y
1
A2
= y
1
.
Thus y = y
1
.
Since, given x, the element y that satises A3 is unique, we can name this
element as a function of x. We will denote it by x and call it the additive
inverse of x or just minus x. Therefore, we have:
x + (x) = (x) +x = 0.
2.1. AXIOMS FOR ADDITION 15
Since, by A2, 0 is its own inverse, we have 0 = 0.
As we have said, the proof of Lemma 2.1.1 proves more, namely the following:
Lemma 2.1.2 If x, y R satisfy x +y = 0, then y = x.
We start to prove some number of well-known results:
Lemma 2.1.3 For all x R, (x) = x.
Proof: It is clear from A3 that if y is the additive inverse of x, then x is the
additive inverse of y. Thus x is the additive inverse of x. Hence x = (x).
Lemma 2.1.4 If x, y R satisfy x +y = 0, then x = y.

Proof: Follows directly from lemmas 2.1.2 and 2.1.3.
Lemma 2.1.5 For all x, y R, (x +y) = (y) + (x).
Proof: We compute directly: (x+y) +((y) +(x)) = x+y +(y) +(x) =
x + 0 + (x) = x + (x) = 0. Here the rst equality is the consequence of A1
(that states that the parentheses are useless).
Thus (x+y)+((y)+(x)) = 0. By Lemma 2.1.2, (y)+(x) = (x+y).
We should note that the lemma above does not state that (x+y) = (x)+
(y). Although this equality holds in R, we cannot prove it at this stage; to
prove it we need Axiom A4, which is yet to be stated.
We now dene the following terms:
x y := x + (y)
x +y := (x) +y
x y := (x) + (y)
Lemma 2.1.6 For all x, y R,
(x y) = y x
(x +y) = y +x
(x y) = y +x
Proof: Left as an exercise.
The next lemma says that we can simplify from the left.
Lemma 2.1.7 (Left Cancellation) For x, y, z R if x + y = x + z then
y = z.
Proof: Add x to the left of both parts of the equality x + y = x + z, and
using associativity, we get y = z.
Similarly we have,
Lemma 2.1.8 (Right Cancellation) For x, y, z R if y + x = z + x then
y = z.
Finally, we state our last axiom that involves only the addition.
A4. Commutativity of the Addition. For any x, y R, x +y = y +x.
A set together with a binary operation, say +, and an element denoted 0
that satises the axioms A1, A2, A3 and A4 is called a commutative or an
abelian group. Thus (R, +, 0) is an abelian group.
2.2 Axioms for Multiplication
Let R
= R 0. Since 1 ,= 0, the element 1 is an element of R
. In this
subsection, we will replace the symbols R, + and 0 of the above subsection, by
R
, and 1 respectively. For example, Axiom A1 will then read

M1. Multiplicative Associativity. For any x, y, z R
, x (y z) =
(x y) z.
As we have said, we will prefer to write x(yz) = (xy)z instead of x(yz) =
(x y) z.
Axioms A2 and A3 take the following form:
M2. Multiplicative Identity Element. For any x R
, x1 = 1x = x.
M3. Multiplicative Inverse Element. For any x R
there is a y R
such that xy = yx = 1.
We accept M1, M2 and M3 as axioms. Thus (R
, , 1) is a group.
All the results of the previous subsection will remain valid if we do the above
replacements. Of course, the use of the axioms A1, A2, A3 in the proofs must
be replaced by M1, M2, M3 respectively. That is what we will do now:
Lemma 2.2.1 Given x R
, the element y as in M3 is unique.

The proof of this lemma can be translated from the proof of Lemma 2.1.1
directly:
Proof: Let x R
. Let y be as in M3. Let y

1
satisfy the equation xy
1
= 1.
We will show that y = y
1
, proving more than the statement of the lemma. We
start:
y
M2
= y1 = y(xy
1
)
M1
= (yx)y
1
M3
= 1y
1
M2
= y
1
.
Thus y = y
1
.
Since, given x R
, the element y R
that satises M3 is unique, we can

name this element as a function of x. We will denote it by x
1
and call it the
multiplicative inverse of x or sometimes x inverse. Therefore, we have:
xx
1
= x
1
x = 1.
Note that x
1
is dened only for x ,= 0. The term 0
1
will never be dened.
Since, by M2, 1 is its own inverse, we have 1
1
= 1.
2.3. DISTRIBUTIVITY 17
As we have noticed, the proof of Lemma 2.2.1 proves more, namely the
following:
Lemma 2.2.2 If x, y R
satisfy xy = 1, then y = x
1
.
Lemma 2.2.3 For all x R
, (x
1
)
1
= x.
Proof: As in Lemma 2.1.3.
Lemma 2.2.4 If x, y R
satisfy xy = 1, then x = y
1
.
Lemma 2.2.5 For all x, y R
, (xy)
1
= y
1
x
1
.
We should note that the lemma above does not state that (xy)
1
= x
1
y
1
.
Although this equality holds in R, we cannot prove it at this stage; to prove it
we need Axiom M4.
The next lemma says that we can simplify from the left and also from the
right.
Lemma 2.2.6 (Cancellation) Let x, y, z R
.
i. If xy = xz then y = z.
ii. If yx = zx then y = z.
The lemma above is also valid if either y or z is zero (without x being zero),
but we cannot prove it yet.
Finally, we state our last axiom that involves only multiplication.
M4. Commutativity of the Multiplication. For any x, y R
, xy = yx.
Thus (R
, , 1) is an abelian group.
Sometimes, one writes x/y or
x
y
instead of xy
1
.
2.3 Distributivity
In the rst subsection, we stated the axioms that involve only the addition,
and in the second subsection, the axioms that involve only the multiplication.
Until now there is no relationship whatsoever between the addition and the
multiplication. For the moment they appear to be two independent operations.
Consequently, at this point we cannot prove any equality that involves both
operation, e.g. the equalities (1)
1
= 1 and (1)x = x cannot be proven
at this stage.
Below, we state an axiom that involves both addition and multiplication.
LD. Left Distributivity. For all x, y, z R, x(y +z) = xy +xz.
Since the multiplication is commutative, we also have the partial right dis-
tributivity valid if x, y, z ,= 0 and y +z ,= 0:
(y +z)x
M4
= x(y +z)
D
= xy +xz
M4
= yx +zx.
Lemma 2.3.1 For all x R, x0 = 0.
Proof: Since x0 +0
A2
= x0
A2
= x(0 +0)
LD
= x0 +x0, by Lemma 2.1.7, 0 = x0.
I do not know whether one can prove the right distributivity in its full
generality or the equality 0x = 0 from the axioms above. It seems (but I may
be wrong) that we need to add the right distributivity to the list of our axioms
as well. This is what we will do now:
D. Distributivity. For all x, y, z R,
x(y +z) = xy +xz and (y +z)x = yx +zx.
Axiom D will be the only axiom relating the addition and the multiplication.
Lemma 2.3.2 For all x R, 0x = 0.
It follows that M4 is valid for all x, y R.
A set R together with two binary operations + and and constants 0 and
1 that satisfy the axioms A1, A2, A3, A4, M1, M2, M3, M4 and D is called a
eld. Thus (R, +, , 0, 1) is a eld.
We investigate some other consequences of distributivity:
Lemma 2.3.3 For all x, y R,
(x)y = (xy)
x(y) = (xy)
(x)(y) = xy
Proof: We compute directly: 0 = 0y
A3
= (x + (x))y
D
= xy + (x)y (the rst
equality is Lemma 2.3.2). Thus (x)y is the additive inverse of xy and so
(x)y = (xy). This is the rst equality. The others are similar and are left as
exercise.
It follows that we can write xy for (x)y or x(y).
Corollary 2.3.4 For all x R, (1)x = x.
Corollary 2.3.5 (1)
1
= 1.
2.4. AXIOMS FOR THE ORDER RELATION 19
2.4 Axioms for the Order Relation
The rst three axioms below involve only the binary relation inequality <.
Very soon we will relate the inequality to the operations + and .
O1. Transitivity. For all x, y, z R, if x < y and y < z then x < z.
O2. Irreexivity. For all x R, it is not true that x < x, i.e. x ,< x.
O3. Total Order. For any x, y R, either x < y or x = y or y < x.
Lemma 2.4.1 For any x, y R, only one of the relations
x < y, x = y, y < x
hold.
Proof: Assume x < y and x = y hold. Then x < x, contradicting O2.
Assume x < y and y < x hold. Then by O1, x < x, contradicting O2.
Assume x = y and y < x hold. Then x < x, contradicting O2.
Now we state the two axioms that relate the inequality with the two opera-
tions + and :
OA. For all x, y, z R, if x < y then x +z < y +z.
OM. For all x, y, z R, if x < y and 0 < z then xz < yz.
A set R together with two binary operations + and , two constants 0 and 1
and a binary relation < that satises the axioms A1, A2, A3, A4, M1, M2, M3,
M4, D, O1, O2, O3, OA and OM is called an ordered eld. Thus (R, +, , 0, 1)
is an ordered eld. Below, we investigate the properties of ordered elds.
Lemma 2.4.2 If x < y then y < x.
Proof: Adding xy to both sides of the inequality x < y, by OA we get the
result.
Lemma 2.4.3 If 0 < x < y then 0 < y
1
< x
1
.
Lemma 2.4.4 If x < 0 and y < 0 then 0 < xy.
Proof: By Lemma 2.4.2, 0 < x and 0 < y. Then, 0
2.3.2
= 0(y)
OM
<
(x)(y)
2.3.3
= xy.
For any x, we dene x
2
to be xx.
Corollary 2.4.5 For any x R, x
2
0.
If x < 0 we say that x is strictly negative, if x > 0 we say that x is strictly
positive. If x 0 we say that x is nonpositive and if x 0 we say that x is
nonpositive.
We let
R
>0
= x R : x > 0
R
0
= x R : x 0 = R> 0 0
R
<0
= x R : x < 0 = R
>0
R
0
= x R : x 0 = R
<0
0 = R
0
To nish the axioms of real numbers, there is one more axiom left. This will
be the subject of one of the later subsections.
Exercises and Examples.
i. The set 0, 1 with the following addition
0 + 0 = 1 + 1 = 0
0 + 1 = 1 + 0 = 1
and multiplication dened by 0 x = x0 = 0 for x = 0, 1 and 1 1 = 1
satises all the axioms about addition and multiplication (A1, A2, A3,
A4, M1, M2, M3, M4, D), but there is no order relation on 0, 1 that
satises the order axioms (O1, O2, O3, OA, OM).
ii. The set Z of integers, with the usual addition, multiplication and the
order relation, satises all the axioms except M3.
iii. The set Q of rational numbers, with the usual addition, multiplication
and the order relation, satises all the axioms. We will see that Q does not
satisfy the Completeness Axiom that we will state in the next subsection.
We assume for the rest that we are in a structure that satises the axioms
stated until now.
iv. Show that if xy = 0 then either x or y is zero.
v. Dene 2 to be 1 + 1. Show that 2x = x +x.
vi. Show that the axioms above imply that R is innite.
vii. Prove that (x + y)
2
= x
2
+ 2xy + y
2
and that x
2
y
2
= (x y)(x + y).
(Recall that x
2
was dened to be xx).
viii. Prove that if 0 < x < y then x
2
< y
2
.
ix. Prove that if x
2
= y
2
then either x = y or x = y.
2.5. TOTALLY ORDERED SETS 21
x. Prove that if x < y and z < t then x +z < y +t.
xi. Prove that if 0 < x < y and 0 < z < t then xz < yt.
xii. Prove that if x < y, then x <
x+y
2
< y.
xiii. Dene [x[ to be max(x, x) (i.e. the largest of the two). For all x, y R,
show the following:
a) [x[ 0.
b) [x[ = 0 if and only if x = 0.
c) [ x[ = [x[.
d) [x +y[ [x[ +[y[.
e) Conclude from (d) that [x[ [y[ [x y[.
f) Show that [[x[ [y[[ [x y[. (Hint: Use (e)).
g) Show that [xy[ = [x[[y[.
The number [x[ is called the absolute value of x.
xiv. Let x, y R. Show that
max(x, y) =
x +y +[x y[
2
, min(x, y) =
x +y [x y[
2
.
xv. Show that [[x[ [y[[ [x y[ for any x, y R.
xvi. Show that Example v on page 9 is a commutative group.
2.5 Totally Ordered Sets
Let X be a set together with a binary relation < that satises the axioms O1,
O2 and O3. We will call such a relation < a totally ordered set, or a linearly
ordered set, or a chain. Thus R is a totally ordered set. Thus (R, <) is a
totally ordered set. We dene the relations x y, x > y, x y, x ,< y etc. as
usual.
In a totally ordered set (X, <) one can dene intervals as follows (below a
and b are elements of X):
(a, b) = x X : a < x < b
(a, b] = x X : a < x b
[a, b) = x X : a x < b
[a, b] = x X : a x b
(a, ) = x X : a < x
[a, ) = x X : a x
(, a) = x X : x < a
(, a] = x X : x a
(, ) = X
An element M of a totally ordered set X is called maximal if no element
of X is greater than M. An element m of a poset X is called minimal if no
element of X is less than m.
Let (X, <) be a totally ordered set. Let A X be a subset of X. An element
x X is called an upper bound for A if x a for all a A.
An element a X is called a least upper bound for A if,
i) a is an upper bound of A, and
ii) If b is another upper bound of A, b ,< a.
Let X be a poset. A subset of X that has an upper bound is said to be
bounded above.
The terms lower bound, greatest lower bound and a set bounded
below are dened similarly.
Lemma 2.5.1 A subset of a totally ordered set which has a least upper bound
(resp. a greatest lower bound) has a unique least upper bound (resp. greatest
lower bound).
Proof: Let X be a totally ordered set. Let A X be a subset which has a
least upper bound, say x. Assume y is also a least upper bound for A. By
denition y ,< x and x ,< y. Then x = y by O3.
Thus if A is a subset of a totally ordered set whose least upper bound exists,
then we can name this least upper bound as a function of X. We will use the
notation lub(A) or sup(A) for the least upper bound of A. We use the notations
glb(A) and inf(A) for the least upper bound.
Examples and Exercises.
i. Let X = (R, <). Let A = (0, 1) R. Then any number 1 is an upper
bound for A. 1 is the only least upper bound of A. Note that 1 , A.
ii. Let X = (R, <). Let A = [0, 1] R. Then any number 1 is an upper
bound for A. 1 is the only least upper bound of A. Note that 1 A.
iii. Let X = (R, <). Let A = (0, ) R. Then A has no upper bound. But
it has a least upper bound.
iv. Let X = (R, <) and A =
x
x+1
: x R and x 1. Show that 1 is the
only least upper bound of A.
v. Let X = (R, <) and A R any subset of R. Dene A = a : a A.
i. Show that if x is an upper bound for A then x is a lower bound for
A.
ii. Show that if x is the least upper bound for A R then x is the
greatest lower bound for A.
2.6. COMPLETENESS AXIOM 23
vi. Let (X, <) be a totally ordered set and A X. Let
B := x X : x is an unpper bound for A.
a) Assume that sup(A) exists. Show that inf(B) exists and inf(B) =
sup(A).
b) Assume that inf(B) exists. Show that sup(A) exists and sup(A) =
inf(B).
vii. Let X be a totally ordered set and b X. Show that b is the only least
upper bound of (, b], and also of (, b).
viii. On R R dene the relation as follows (x, y) (x
1
, y
1
) by either
y < y
1
, or y = y
1
and x < x
1
.
a) Show that this is a total order (called lexicographic ordering).
b) Does every subset of this linear order which has an upper bound has a
least upper bound?
ix. On N N dene the relation as above (lexicographic order).
a) Show that this is a total order.
least upper bound?
x. Find a poset where the intersection of two intervals is not necessarily an
interval.
2.6 Completeness Axiom
We now state the last axiom for R. This axiom is dierent from the others in
the sense that all the other axioms were about a property of one, two or at most
three elements of R. But this one is a statement about subsets of R.
C. Completeness. Any nonempty subset of R which is bounded above has a
least upper bound.
It follows from Lemma 2.5.1, that a subset X of R which has an upper bound
has a unique least upper bound. We denote it by sup(X) or lub(X). The least
upper bound of a set is sometimes called the supremum of the set. Note that
the supremum of a set may or may not be in the set.
This completes our list of axioms. From now on we x a set R together with
two binary operations + and , two distinct constants 0 R and 1 R and a
binary relation < that satises the axioms A1, A2, A3, A4, M1, M2, M3, M4,
D, O1, O2, O3, AO, MO and C stated above. (The existence of such a structure
is proven in Math 112.)
Lemma 2.6.1 Any nonempty subset of R which is bounded below has a greatest
lower bound.
Proof: Left as an exercise. Use Exercise v, 22 or Exercise i, 24 below.
The greatest lower bound of a (nonempty) set is denoted by inf(X) or
glb(X). The greatest lower bound is sometimes called the inmum of the
set. Note that the inmum of a set may or may not be in the set.
Exercises.
i. Show that if X R has a least upper bound then the set X := x :
x X has a greatest lower bound and inf(X) = sup(X).
ii. Suppose X, Y R have least upper bounds. Show that the set X +Y :=
x + y : x X, y Y has a least upper bound and that sup(X + Y ) =
sup(X) + sup(Y ).
iii. Suppose X, Y R
0
have least upper bounds. Show that the set XY :=
xy : x X, y Y has a least upper bound and sup(XY ) = sup(X) sup(Y ).
Does the same equality hold for any two subsets of R?
The completeness axiom makes the dierence between Q and R. The
equation x
2
= 2 has no solution in Q but has a solution in R. This is what we
now prove.
Theorem 2.6.2 Let a R
>0
. Then there is an x R such that x
2
= a.
Proof: Replacing a by 1/a if necessary, we may assume that a 1. Let
A = x R
0
: x
2
a. For x A, we have x
2
a a
2
. It follows that
0 a
2
x
2
= (a x)(a + x), so a x. We proved that a is an upper bound
for A. Let b = lub(A). We will show that b
2
= a. Clearly b 1 > 0.
Assume rst that b
2
< a. Let = min(
ab
2
2b+1
, 1) > 0. Then (b + )
2
=
b
2
+2b +
2
b
2
+2b + = b
2
+(2b +1) b
2
+(ab
2
) = a. Hence b + A.
But this contradicts the fact that b is the least upper bound for A.
Assume now that b
2
> a. Let = min(
b
2
a
2b
, b) > 0. Now (b )
2
=
b
2
2b +
2
> b
2
2b b
2
(b
2
a) = a. Thus (b )
2
> a. Let x A be
such that b x b. (There is such an x because b is not an upper bound
for A). Now we have a < (b )
2
x
2
a (because b 0), a contradiction.
It follows that b
2
= a.
Remark. Since every nonnegative real number has a square root, the order
relation < can be dened from + and as follows: for all x, y R,
x < y if and only if z (z ,= 0 y = x +z
2
).
2.6. COMPLETENESS AXIOM 25
Exercises.
i. Let A R be a subset satisfying the following property: For all a, b A
and x R, if a x b then x A. Show that A is an interval.
ii. Let x
3
mean x x x. Show that for any x R there is a unique y R
such that y
3
= x.
Chapter 3
Other Number Sets
3.1 Natural Numbers and Induction
We say that a subset X of R is inductive if 0 X and if for all x X, x+1 is
also in X. For example, the subsets R
0
, R, R (0, 1) are inductive sets. The
set R
>0
is not an inductive set.
Lemma 3.1.1 An arbitrary intersection of inductive subsets is an inductive
subset. The intersection of all the inductive subsets of R is the smallest inductive
subset of R.
Proof: Trivial.
We let N denote the smallest inductive subset of R. Thus N is the intersection
of all the inductive subsets of R.
The elements of N are called natural numbers.
Theorem 3.1.2 (Induction Principle (1)) Let X be a subset of R. Assume
that 0 X and for any x R, if x X then x + 1 X. Then N X.
Proof: The statement says that X is inductive. Therefore the theorem follows
directly from the denition of N.
Suppose we want to prove a statement of the form for all x N, (x).
For this, it is enough to prove
i) (0),
ii) If (x) then (x + 1).
Indeed, assume we have proved (i) and (ii). let X := x R : (x). By
(i), 0 X. By (ii), if x X then x + 1 X. Thus, by the Induction Principle,
N X. It follows that for all x N, (x).
Lemma 3.1.3 i. lub(N) = 0, i.e. 0 is the least element of N.
ii. If x N 0 then x 1 N.
27
28 CHAPTER 3. OTHER NUMBER SETS
iii. N is closed under addition and multiplication, i.e. if x, y N then
x +y, xy N.
iv. If 0 < y < 1 then y , N.
v. Let x N. If x < y < x + 1, then y , N.
vi. Let x, y N. If x < y then x + 1 y.
vii. Let x, y N. Assume y < x. Then x y N.
viii. If x, y N and y < x + 1, then either y = x or y < x.
Proof: i. Clearly R
0
is an inductive set. But N is dened to be the smallest
inductive subset of R. Thus N R
0
. Since 0 N, it follows that 0 is the least
element of N.
ii. Assume that for x N0, x1 , N. Then the set Nx is an inductive
set (check carefully). Since N is the smallest inductive set, N N x and
x , N.
iii. Let x, y N. We proceed by induction y to show that x + y N, i.e.
letting (y) denote the statement x+y N, we show that (0) holds and that if
(y) holds then (y +1) holds. If y = 0 then x +y = x+0 = x N. Thus (0)
holds. Assume now (y) holds, i.e. that x+y N. Then x+(y+1) = (x+y)+1.
Since x + y N by assumption, we also have x + (y + 1) N. Thus (y + 1)
holds also. Therefore (y) holds for all N.
The proof for the multiplication is left as an exercise.
iv. The set N (0, y] is an inductive set as it can be shown easily. Thus
N N (0, y] and y , N.
v. We proceed by induction on x. The previous part gives us the case x = 0.
Assume the statement holds for x and we proceed to show that the statement
holds for x + 1. Let x + 1 < y < (x + 1) + 1, then x < y 1 < x + 1. By
the inductive hypothesis, y 1 , N. By part (i), either y = 0 or y , N. Since
0 < x + 1 < y, we cannot have y = 0. Thus y , N.
vi. Assume not. Then x < y < x + 1, contradicting part (v).
vii. By induction on x. If x = 0, then the statement holds because there is
no y N such that y < x. Assume the statement holds for x. We will show that
it holds for x+1. Let y N be such that y < x+1. Then either y < x or y = x.
Now (x + 1) y = (x y) + 1. In case y < x, the induction hypothesis gives
x y N and so (x + 1) y N. In case y = x, we have (x + 1) y = 1 N.
viii. By (v), y x.
Lemma 3.1.4 Any nonempty subset of N has a least element.
Proof: Let ,= X N. Assume X does not have a least element. We will
prove that X = , which is the same as proving that no element of N is in X.
We will rst show the following statement (n) for all n: No natural number
m < n is in X.
Since there are no natural numbers < 0 the statement holds for 0.
Assume (n) holds. If (n+1) were false, then n would be in X by Lemma
3.1.3.viii and it would be the smallest element of X, a contradiction.
Thus (n) holds for any n. Now if a natural number n were in X, since (n)
holds, n would be the smallest element of X, a contradiction.
3.1. NATURAL NUMBERS AND INDUCTION 29
There is a slightly more complicated version of the inductive principle that
is very often used in mathematics:
Theorem 3.1.5 (Induction Principle (2)) Let X be a subset of N. Assume
that for any x N,
(y N (y < x y X)) x X.
Then X = N.
Proof: Assume not. Then N X ,= . Let x be the least element of N X.
Thus y N (y < x y X). But then by hypothesis x X, a contradiction.
How does one use Induction Principle (2) in practice? Suppose we have
statement (x) to prove about natural numbers x. Given x N, assuming (y)
holds for all natural numbers y < x, one proves that (x) holds.This is enough
to prove that (x) holds for all x N.
We immediately give some applications of the Induction Principles.
Theorem 3.1.6 (Archimedean Property) Let R
>0
and x R, then
there is an n N such that x < n.
Proof: Assume not, i.e. assume that n x for all n N. Then the set N is
bounded above by x/. Thus N has a least upper bound, say a. Hence a 1
is not an upper bound for N. It follows that there is an element n N that
satises a 1 < n. But this implies a < n+1. Since n+1 N, this contradicts
the fact that a is an upper bound for N.
Lemma 3.1.7 Any nonempty subset of N that has an upper bound contains its
least upper bound.
Proof: Let ,= A N be a nonempty subset of N that has an upper bound.
Let x be the least upper bound of A. Since x 1 is not an upper bound for A,
there is an a A such that x 1 < a x. By parts (iv) and (v) of Lemma
3.1.3.vii a is the largest element of A.
Theorem 3.1.8 (Integral Part) For any x R
0
, there is a unique n N
such that n x < n + 1.
Proof: Let A = a N : a x. Then 0 A and A is bounded above by x.
By Lemma 3.1.7, A contains its least upper bound, say n. Thus n x < n +1.
This proves the existence. Now we prove the uniqueness. Assume m N and
m x < m + 1. If n < m, then by Lemma 3.1.3.vii, x < n + 1 m x, a
contradiction. Similarly m ,< n. Thus n = m.
Theorem 3.1.9 (Division) For any n, m N, m ,= 0 there are unique q, r
N such that n = mq +r and 0 r < m.
Proof: We rst prove the existence. We proceed by induction on n (Induction
Principle 2). If n < m, then take q = 0 and r = n. Assume now n m. By
induction, there are q
1
and r
1
such that n m = mq
1
+ r
1
and 0 r
1
< m.
Now n = m(q
1
+1) +r
1
. Take q = q
1
+1 and r = r
1
. This proves the existence.
We now prove the existence. Assume n = mq + r = mq
1
+ r
1
, 0 r < m
and 0 r
1
< m. Assume q
1
> q. Then m > m r
1
> r r
1
= mq
1
mq =
m(q
1
q) m. This is a contradiction. Similarly q ,> q
1
. Hence q
1
= q. It
follows immediately that r = r
1
.
Exercises.
i. Show that for any n N 0, 1 + 3 +. . . + (2n 1) = n
2
.
ii. Show that for any n N0, 1
2
+2
2
+3
2
+. . . +n
2
= n(n+1)(2n+1)/6.
iii. Show that for any n N,
1
1
1
3
+
1
3
1
5
+. . . +
1
2n 1
1
2n + 1
=
n
2n + 1
.
iv. Show that for any n N 0, 1 + 3 +. . . + (2n 1) = n
2
.
v. Show that for any n N0, 1
2
+2
2
+3
2
+. . . +n
2
= n(n+1)(2n+1)/6.
vi. Show that for any n N,
1
1
1
3
+
1
3
1
5
+. . . +
1
2n 1
1
2n + 1
=
n
2n + 1
.
vii. Show that for any n N 0, 1
2
+ 2
2
+. . . +n
2
=
n(2n+1)(n+2)
6
.
3.1.1 Exponentiation
Let r R. For n N, we dene r
n
, n
th
power of r, as follows by induction
on n:
r
0
= 1 if r ,= 0
r
1
= r
r
n+1
= r
n
r
Note that 0
0
is not dened. We will leave it undened. Note also that the
previous denition of r
2
coincides with the one given above: r
2
= r
1+1
= r
1
r =
rr.
Proposition 3.1.10 For r R and n N not both zero, we have,
i. (rs)
n
= r
n
s
n
.
ii. r
n
r
m
= r
n+m
.
iii. (r
n
)
m
= r
nm
.
Theorem 3.1.11 i. Let r R
0
and n N 0. Then there is a unique
s R such that s
m
= r.
ii. Let r R and let n N be odd. Then there is a unique s R such that
s
m
= r.
Proof: (ii) follows from (i). (i) is proved as in Theorem 2.6.2. Left as an
exercise.
The number s is called the m
th
-root of r.
3.1.2 Factorial
For n N, we dene n! by induction on n: Set 0! = 1, 1! = 1 and (n + 1)! =
n!(n + 1). This just means that n! = 1 2 . . . n.
Exercises.
i. Show that a set with n elements has n! bijections.
ii. Find a formula that gives the number of injections from a set with n
elements into a set with m elements.
iii. Prove that for n N, the set 0, 1, . . . , n 1 has 2
n
subsets. (Hint: You
may proceed by induction on n).
iv. Show that n! > 2
n
for all n large enough.
v. Show that (x 1)
n
x
n
nx
n1
for all x > 1. (Hint: By induction on
n).
vi. Show that if 0 < x < 1 and n > 0 is a natural number, then (1 x)
n
1 nx +
n(n1)
2
x
2
.
vii. Show that for any n N 0, 1
3
+ 2
3
+. . . +n
3
= (1 + 2 +. . . +n)
2
.
viii. Show that for any n N 0, 1
4
+ 2
4
+. . . +n
4
=
n(n+1)(6n
3
+9n
2
+n1)
30
.
n Choose k. For n, k N and k n, dene
_
n
k
_
=
n!
k!(n k)!
.
Exercises.
i. Show that
_
n
k
_
=
_
n
n k
_
.
ii. Show that
_
n
0
_
=
_
n
n
_
= 1.
iii. Show that
_
n
1
_
= n.
iv. Show that
_
n + 1
k + 1
_
=
_
n
k
_
+
_
n
k + 1
_
.
v. Deduce that
_
n
k
_
N. (Hint: By induction on n).
vi. Show that for n N and 0 k n, a set with n elements has
_
n
k
_
subsets with k elements.
vii. Show that for n N and k N with k n, a set with n elements has
_
n
k
_
subsets with k elements.
viii. Show that for x, y R and n N,
(x +y)
n
=
n
k=0
_
n
k
_
x
k
y
nk
.
(Hint: By induction on n).
ix. Show that

n
k=0
_
n
k
_
= 2
n
.
x. Show that

n
k=0
(1)
k
_
n
k
_
= 0.
xi. Compute (x +y +z)
3
in terms of x, y and z.
xii. Compute (x +y +z)
4
in terms of x, y and z.
xiii. Show that for x > 1 and n N, (x 1)
n
x
n
nx
n1
.
xiv. Show that for x < 1, (1 x)
n
1 nx.
xv. Show that for n N0, (1+
1
n
)
n
(1+
1
n+1
)
n+1
. (See Theorem 6.8.1).
3.1.3 Sequences
Let X be a set. A sequence in X is just a function x : N X. We let
x
n
:= x(n) and denote x by listing its values (x
n
)
n
rather than by x. We
can also write
x = (x
0
, x
1
, x
2
, x
3
, . . . , x
n
, . . .).
If X = R, we speak of a real sequence. For example
_
1
n+1
_
n
is a real
sequence. We can write this sequence more explicitly by listing its elements:
(1, 1/2, 1/3, 1/4, 1/5, . . . , 1/(n + 1), . . .).
If we write a sequence such as (1/n)
n
, we will assume implicitly that the
sequence starts with n = 1 (since 1/n is undened for n = 0). Thus, with this
convention, the above sequence (
1
n+1
)
n
may also be denoted by (1/n)
n
. For
example, the sequence
_
1
n(n1)
_
n
starts with n = 2 and its elements can be
listed as
(1/2, 1/6, 1/12, 1/20, 1/30, . . . , 1/n(n 1), . . .).
A sequence (x
n
)
n
is called increasing or nondecreasing, if x
n
x
n+1
for
all n N. A sequence (x
n
)
n
is called strictly increasing, if x
n
< x
n+1
for all
n N. The terms decreasing, strictly decreasing and nonincreasing are
dened similarly.
Let (x
n
)
n
be a sequence. Let (k
n
)
n
be a strictly increasing sequence of
natural numbers. Set y
n
= x
k
n
. Then we say that the sequence (y
n
)
n
is a
subsequence of the sequence (x
n
)
n
. For example, let x
n
=
1
n+1
and k
n
= 2n.
Then y
n
= x
kn
= x
2n
=
1
2n+1
. Thus the sequence (y
n
)
n
is
(1, 1/3, 1/5, 1/7, . . . , 1/(2n + 1), . . .).
If we take k
0
= 1 and k
n
= 2
n
:= 22. . . 2 (n times), then the subsequence
(y
n
)
n
becomes
(1/2, 1/3, 1/5, 1/9, 1/17, . . . , 1/(2
n
+ 1), . . .).
We now prove an important consequence of the Completeness Axiom:
Theorem 3.1.12 (Nested Intervals Property) Let (a
n
)
n
and (b
n
)
n
be two
real sequences. Assume that for each n, a
n
a
n+1
b
n+1
b
n
. Then
nN
[a
n
, b
n
] = [a, b] for some real numbers a and b. In fact a = supa
n
: n N
and b = inf
n
a
n
: n N.
Proof: Since the set a
n
: n N is bounded above by b
0
, it has a least upper
bound, say a. Similarly the set b
n
: n N has a greatest lower bound, say b.
I claim that
nN
[a
n
, b
n
] = [a, b].
If x a, then x a
n
for all n. Likewise, if x b, then x b
n
for all n.
Hence, if x [a, b], the x [a
n
, b
n
] for all n.
Conversely, let x
nN
[a
n
, b
n
]. Then a
n
x b
n
for all n. Thus x is
an upper bound for a
n
: n N and a lower bound for b
n
: n N. Hence
a x b.
Exercises.
i. Can you nd an increasing sequence (a
n
)
n
and a decreasing sequence
(b
n
)
n
N of real numbers with a
n
< b
m
for all n, m N such that
n
[a
n
, b
n
) = ?
ii. Prove Theorem 3.1.12 for R
n
(with closed cubes or balls rather than closed
intervals).
iii. Let G be a subset of R
containing 1, closed under multiplication and

inversion. Let Seq(R) be the set of sequences of R. For a = (a
n
)
n
and
b = (b
n
)
n
in Seq(R), set a b if and only if for some g G, a
n
= gb
n
eventually. Show that is an equivalence relation on Seq(R). Find the
equivalence classes when G = 1, G = 1, 1 and G = R
.
3.2 Integers and Rational Numbers
In N, we can add and multiply any two numbers, but we cannot always subtract
one number from another. We set Z = n m : n, m N. Now in Z we can
add, multiply and subtract any two numbers. The elements of Z are called
integers.
In Z, we can add, multiply and subtract any two numbers, but we cannot
always divide one number to another. We set Q = n/m : n, m Z, m ,=
0. The elements of Q are called rational numbers. Now in Q we can add,
multiply, subtract and divide any two numbers, with the only exception that
we cannot divide a rational number by0.
We will not go into further detail about these number systems. We trust the
reader in proving any elementary statement about numbers, for example the
decomposition of integers into prime factors.
The structure (Q, +, , 0, 1, <) satises all the axioms A1-4, M1-4, O1-3,
OA, OM. But it does not satisfy the Completeness Axiom C as the following
lemma shows (compare with Theorem 2.6.2).
Lemma 3.2.1 There is no q Q such that q
2
= 2.
Proof: Assume not. Let q Q be such that q
2
= 2. Let a, b Z be such that
q = a/b. Simplifying if necessary, we may choose a and b so that they are not
both divisible by 2. From (a/b)
2
= q
2
= 2 we get a
2
= 2b
2
. Thus a
2
is even. It
follows that a is even. Let a
1
Z be such that a = 2a
1
. Now 4a
2
1
= a
2
= 2b
2
and 2a
2
1
= b
2
. Hence b is even as well, a contradiction.
Theorem 3.2.2 Q is dense in R, i.e. for any real numbers r < s, there is a
rational number q such that r q s.
Proof: By Theorem 3.1.6, there is a natural number such that 1 < n(r s).
Now consider the set A := m N : m/n < s. By Theorem 3.1.6 again, A
is a bounded set. By Lemma 3.1.7, A has a maximal element, say m. Thus
m/n < s and
m+1
n
s. We compute: s
m+1
n
=
m
n
+
1
n
< s + (r s) = r.
3.2. INTEGERS AND RATIONAL NUMBERS 35
3.2.1 Exponentiation
Let r R. At page 30, we have dened r
n
for n N (except for 0
0
, which was
left undened). If r ,= 0, we can extend this denition to Z by r
n
= (r
n
)
1
.
Note that the previous denition of r
1
coincides with the one given above.
Proposition 3.2.3 For r R and n Z not both zero, we have,
i. (rs)
n
= r
n
s
n
.
ii. r
n
r
m
= r
n+m
.
iii. (r
n
)
m
= r
nm
.
If r R
0
and q Q, we can also dene r
q
as follows: Let m N 0, by
Theorem 3.1.11, there is a unique s R such that s
m
= r. Set s = r
1/m
. Now
for n Z and m N 0, dene r
n/m
to be (r
1/m
)
n
.
Proposition 3.2.4 For r R
0
and q, q
1
, q
2
Q, we have,
i. (rs)
q
= r
q
s
q
.
ii. r
q
1
r
q
2
= r
q
1
+q
2
.
iii. (r
q1
)
q2
= r
q1q2
.
Exercises
i. Show that if q Q is a square in Q, then 2q is not a square in Q.
ii. Let a < b be real numbers. Show that for each n N, there are rational
numbers a
n
< b
n
such that
n
[a
n
, b
n
] = [a, b]. Hint: See Theorem 3.2.2.
iii. Let a < b be real numbers. Show that for each n N, there are rational
numbers a
n
< b
n
such that
n
(a
n
, b
n
) = [a, b]. Hint: See the exercise
above.
iv. Let a < b be real numbers. Show that for each n N, there are rational
numbers a
n
< b
n
such that
n
[a
n
, b
n
] = (a, b). Show that if a and b are
nonrational numbers (a
n
and b
n
are still rational numbers), then we can
never have
n
[a
n
, b
n
] = [a, b]. Hint: See the exercise above.
v. For each n N, let a
n
and b
n
be such that a
n+1
< a
n
< b
n
< b
n+1
. Show
that
n
(a
n
, b
n
) is an open interval (bounded or not).
vi. Show that for all rational number q > 0, there is a rational number x for
which 0 < x
2
< q.
vii. a) Show that if x and y are two nonnegative rational numbers whose sum
is 1, then a ax +by b.
b) Let q be a rational number such that a q b. Show that there
are two nonnegative rational numbers x and y such that x + y = 1 and
q = ax +by.
c) Show that the numbers x and y of part b are unique.
viii. Show that for any x R, there is a unique n Z such that n x < n+1.
ix. Show that for any n, m Z, m ,= 0 there are unique q, r Z such that
n = mq +r and 0 r < m.
x. Let A
and A
r
be two nonempty subsets of Q such that
a) A
A
r
= Q.
b) A
A
r
= .
c) Any element of A
is less than any element of A

r
.
Show that sup(A
) = inf(A
r
).
3.3 Uniqueness of the Real Number System
We assumed without a proof that there was a structure (R, +, , <, 0, 1) that
satises all our axioms. Assuming there is really such a structure, can there
be other structures satisfying the axioms of real numbers? Of course! Just
rename the elements of R and dene the addition, the multiplication and the
order accordingly to get another structure that satises the same axioms. For
example, the structure (R
, +
, <
, 0
, 1
) dened by
R
= 0 R
0
= (0, 0)
1
= (0, 1)
(0, r) +
(0, s) = (0, r +s)

(0, r)
(0, s) = (0, r s)
(0, r) <
(0, s) r < s
satises the axioms of R.
One might argue that this structure we have just dened is not very dierent
from the old one, that all we did was renaming the element r of R by (0, r).
Indeed... And that is all one can do as we will soon prove.
Note that in the above example, the map f : R R
dened by f(r) =
(0, r) is a bijection that has the following properties:
f(0) = 0
f(1) = 1
f(r +s) = f(r) +
f(s)
f(r s) = f(r)
f(s)
r < s f(r) <
f(s)
for all r, s R.
We will show that if (R
, +
, <
, 0
, 1
) is a structure that satises all the

axioms of real numbers then there is a bijection f : R R
that satises
the above properties. This means that R
is just R with its elements renamed:

r R is named f(r). Note also that all the theorems we have proved for
(R, +, , <, 0, 1) are also valid for (R
, +
, <
, 0
, 1
). In particular R
has a
3.3. UNIQUENESS OF THE REAL NUMBER SYSTEM 37
smallest inductive set N
, contains a dense subset Q
similar to Q in R (Theorem
3.2.2) etc.
Theorem 3.3.1 (Uniqueness of the Real Number System) Let
(R
, +
, <
, 0
, 1
)
be a structure that satises all the axioms of the real numbers. Then there is a
unique nonconstant map f : R R
such that for all r, s R,

(1) f(r +s) = f(r) +
f(s) (f is additive)
(2) f(r s) = f(r)
f(s) (f is multiplicative)
Furthermore such a map must be a bijection and must also satisfy
(3) r < s f(r) <
f(s) (f is order preserving)

(4) f(0) = 0
(5) f(1) = 1
.
Proof: 1. We rst prove that a map that satises (1) and (2) must be a
bijection that satises (3), (4), (5).
1a. f(0) = 0
. A map f that satises (1) must send 0 to 0
, because
0
f(0) = f(0) = f(0 + 0) = f(0) +
f(0) and so by simplifying we get

f(0) = 0
.
1b. f(x) = f(x). (Here the sign on the right hand side stands for
the additive inverse in R
for the binary operation +
). Indeed, 0
= f(0) =
f(x + (x)) = f(x) +
f(x) and so f(x) = f(x).

1c. f is one to one. Assume f(x) = 0
for some x R 0. We will get

a contradiction. We compute: f(1) = f(x
1
x) = f(x
1
)
f(x) = f(x
1
)
= 0
and so, for all r R, f(r) = f(r 1) = f(r)
f(1) = f(r)
= 0
,
contradicting the fact that f is nonconstant. Thus f(x) = 0
implies x = 0.
We can now show that f is one to one. Assume f(x) = f(y). Then 0
=
f(x)
f(y) = f(x) +
(f(y)) = f(x) +
f(y) = f(x +(y)) = f(x y). By

above x y = 0, and x = y. Hence f is one to one.
1d. f(1) = 1
. Since f(1) = f(1 1) = f(1)
f(1), f(1) is either 0
or 1
.
But since f(0) = 0
and f is one to one, f(1) = 1
.
1e. f(N) = N
. We can show by induction on n that f(n) N
. Thus
f(N) N
. It is also clear that f(N) is an inductive subset of N
. Hence
f(N) = N
.
1f. f(Q) = Q
. By 1e and 1b, f(Z) = Z
. Let n/m Q with n, m Z,

m ,= 0. Then f(m)
f(n/m) = f(m n/m) = f(n) f(Z) = Z
. Since
f(m) f(Z) and since f(m) ,= 0 (because m ,= 0, see 1c), from this we get
f(n/m) Q
. Thus f(Q) f(Q
). Since f(Z) = Z
, it is also easy to show that

f(Q) = Q
.
1g. If x < y in R then f(x) <
f(y) in R
. Note rst that, in R, x y if

and only if x +z
2
= y for some z R. The same statement holds in R
.
Assume x, y R are such that x < y. Then there is a z R 0 such that
y = x + z
2
. Therefore f(y) = f(x) +
f(z)
2
(here f(z)
2
is the squaring in R
,
i.e. f(z)
2
= f(z)
f(z)). Since f(z) ,= 0
, we get f(x) <
f(y).
1h. f is onto. Let r R
. Consider the sets

L
r
= q Q
: q <
r = Q
(, r)
and
R
r
= q Q
: r <
q = Q
(r, ).
The sets L
r
and R
r
partition Q
. Since f : Q Q
is an order preserving bi-

jection, f
1
(L
r
) and f
1
(L
r
) partition Q into two nonempty convex subsets (of
Q) and every element of f
1
(L
r
) is strictly less than every element of f
1
(R
r
).
Then sup(f
1
(L
r
)) = inf(f
1
(R
r
) (Exercise x, page 36). Let r be this number.
Thus r = sup(f
1
(L
r
)) = inf(f
1
(R
r
). We claim that f(r) = r
. If f(r) < r
,
then (because Q
is dense in R) there is a q
such that f(r) < q
< r
.
Thus q
r
. Let q Q be such that f(q) = q
. Thus q f
1
(L
r
). So q < r,
hence q
= f(q) < f(r), a contradiction. Thus f(r) r
. Similarly f(r) r
.
Therefore f(r) = r
.
2. Existence. We now prove the existence of the map f.
We dene our map f rst on N. We let f(0) = 0
and assuming f(n) has

been dened for n N, we dene f(n +1) to be f(n) +
. Thus by denition,
f(0) = 0
f(n + 1) = f(n) +
By induction on n, one can prove that for all n N, f(n) N
. Also f(N)
is clearly an inductive subset of R. Hence f(N) = N
.
We claim that f : N R
is one to one. Assuming f(n) = f(m) for

n, m N, we will show that n = m. We proceed by induction on n. If n = 0,
then 0
= f(0) = f(n) = f(m), therefore, by the very denition of f, m cannot

be of the form k + 1 for some k N, i.e. m = 0
. If n = k + 1 for some k N,
then f(k) +
= f(k +1) = f(n) = f(m), therefore m ,= 0 and so m = +1 for

some N. Now f(k) +
= f(m) = f( +1) = f() +
and so f(k) = f().

By induction k = and n = k +1 = +1 = m, proving then f : N R is one
to one.
We will denote the image of n N under f by n
, i.e. we let f(n) = n
.
The proofs that for n, m N, f(n + m) = f(n) +
f(m) and f(n m) =

f(n)
f(m) are easy as well and are left as an exercise. From the additivity
of f, it follows that f must preserve the order on N.
Now consider the subset
Q
:= a/b : a N
and b N
0
of R
. Here a/b stands for a
b
1
and b
1
denotes the multiplicative in-
verse of b R
with respect to
. We extend f : N R
to Q by setting
f(n/m) = f(n)/f(m) Q
. We should rst check that this map is well-

dened, meaning that n/m = p/q for n, m, p, q N should imply f(n)/f(m) =
f(p)/f(q). Indeed, if n/m = p/q for n, m, p, q N, then nq = mp, so
f(n)f(q) = f(nq) = f(mp) = f(m)f(p) and f(n)/f(m) = f(p)/f(q). Thus
f : Q Q
. We still denote by f this extended map.

3.4. COMPLEX NUMBERS 39
f : Q Q
is certainly onto: If a/b Q
with a, b N, then f(n) = a and

f(m) = b for some n, m N and so f(n/m) = f(n)/f(m) = a/b.
It is also easy to check that f : Q R
is one to one, additive, multiplicative

and order preserving.
We now extend f to R. Let r R. Consider the disjoint sets
L
r
= q Q : q r
and
R
r
= q Q : r < q.
Note that the sets L
r
and R
r
partition Q and they do dene r, as r is both
supL
r
and inf R
r
. Consider the sets f(L
r
) and f(R
r
). Since f : Q R
is order preserving, any element of f(L
r
) is strictly less than any element of
f(R
r
). Also, since f(Q) = Q
, the sets f(L

r
) and f(R
r
) partition Q
. Therefore
supf(L
r
) = inf f(R
r
). We let f(r) = sup f(L
r
) = inf f(R
r
).
It is a matter of writing to show that f : R R
satises all the required

properties.
3. Uniqueness. Assume f : R R
and g : R R
are such maps.

Then g
1
f : R R is such a map as well. Therefore it is enough to show
that a nonconstant map f : R R satisfying (1) and (2) (therefore also (3),
(4) and (5)) is Id
R
. One can show quite easily (as above) that f is identity on
Q. Since Q is dense in R, it follows that f = Id
R
.
3.4 Complex Numbers
Let C = RR. On C we dene two operations called addition and multiplication
as follows:
Addition : (x, y) + (z, t) = (x +z, y +t)
Multiplication : (x, y)(z, t) = (xz yt, xt +yt)
It is easy to check that the rst nine axioms about the addition and multiplica-
tion of real numbers do hold in C:
A1, A2, A3, A4 (with O
C
= (0, 0) as the additive identity and (x, y)
as the additive inverse of (x, y)),
M1, M2, M3, M4 (with 1
C
:= (1, 0) as the multiplicative identity and
(
x
x
2
+y
2
,
y
x
2
+y
2
) as the multiplicative inverse of the nonzero element (x, y))
and
D
hold. It can be checked that no order satisfying the axioms O1, O2, O3, OA
and OM can be dened on C (See Exercise x, page 42). Thus C is a eld which
is not an ordered eld.
The set C together with the addition and multiplication is called the eld
of complex numbers, each element of C is called a complex number.
The following property is easy to check
() (x, y) = (x, 0) + (0, y) = (x, 0) + (y, 0)(0, 1).
Let us consider the map : R C given by (r) = (r, 0). Then the
following hold:
is one to one
(r +s) = (r) +(s)
(0) = 0
C
(r) = (r)
(rs) = (r)(s)
(1) = 1
C
(r
1
) = (r)
1
Thus the map transports the structure (R, +, , 0, 1) onto the substructure
((R), +, , 0
C
, 1
C
) of (C, +, , 0
C
, 1
C
); in other words, the only dierence be-
tween the two structures (R, +, , 0, 1) and ((R), +, , 0
C
, 1
C
) is the names of
the objects, what is called r in the rst one is called (r, 0) in the second.
The property () above now can be written as
() (x, y) = (x) +(y)(0, 1)
We also note the following:
(r)(x, y) = (rx, ry).
From now on, we will identify R and its image (R) via the map , i.e.
we will let r = (r) = (r, 0). Although r ,= (r, 0) = (r), we will make the
identication r = (r) for the sake of notational simplicity. In particular, we
identify 0 and 0
C
, as well as 1 with 1
C
. In this way we will see R as a subset of
C. Thus the two properties above are now written as
( ) (x, y) = x +y(0, 1)
r(x, y) = (rx, ry)
Let i = (0, 1). Note that
i
2
= ii = (0, 1)(0, 1) = (1, 0) = 1,
and that for any (x, y) C, we have
( ) (x, y) = x +yi
From now on, we will represent a complex number z as x +iy (or as x +yi)
for x, y R rather than as the pair (x, y). Note that, given z C, the real
numbers x, y for which z = x + iy are unique (this would not have been so
if we assumed x and y were in C rather than in R). The addition and the
multiplication of complex numbers with this notation become:
(x +yi) + (z +ti) = (x +z) + (y +t)i
(x +yi)(z +ti) = (xz yt) + (xt +yz)i
3.4. COMPLEX NUMBERS 41
As is the custom, we will write x yi instead of x + (y)i.
The inverse of a nonzero complex number is given by the formula
(x +yi)
1
=
x
x
2
+y
2

y
x
2
+y
2
i.
Conjugation. We consider the following map : C C given by x +yi =
x yi (here x, y R). For every , C, the following properties are easy to
verify:
+ = +
=
1
=
1
Also,
= if and only if R.
The complex number a is called the conjugate of .
Norm. If = x +yi with x, y R then it is easy to check that
a = x
2
+y
2
R
0
.
Thus we can take its square root. We dene
[[ :=
=
_
x
2
+y
2
.
Hence
[[
2
= x
2
+y
2
.
The nonnegative real number [[ is called the norm indexnorm of a com-
plex number of . Note that if R C, then the norm of is equal to the
absolute value of , so that the two meanings that we have given to the notation
[[ coincide.
For all , C, we have the following properties:
P
1
[[ 0 and [[ = 0 if and only if = 0
P
2
[[ = [[ [[
Lemma 3.4.1 For all , C,
i. [ +[ [[ +[[,
ii. [[[ [[[ [ [.
Proof: i. On the one hand, [ + [
2
= ( + ) + = ( + )( + ) =
a + +a + = [[
2
+ +a +[[
2
.
On the other hand, ([[ +[[)
2
= [[
2
+ 2[[[[ +[[
2
.
Thus we need to prove that + 2[[[[, or that + 2
_
.
Setting = , this means that it is enough to prove that + 2
for all
C. Set = x +yi where x, y R. Then + = 2x and

=
_
x
2
+y
2
.
Thus the statement + 2
for all C is equivalent to the statement

x
_
x
2
+y
2
for all x, y R. We will prove this last inequality. Since
x
2
x
2
+ y
2
, taking the square roots of both sides, we infer [x[
_
x
2
+y
2
.
Thus x [x[
_
x
2
+y
2
.
ii. From part (i) we have [[ = [ +()[ [[ +[[. Thus [[ [[
[ [.
Again from part (i), we have [[ = [+()[ [[ +[[ = [[ +[[.
Thus [[ [[ [ [.
These two inequalities mean exactly that [[[ [[[ [ [.
Exercises.
i. Show that (
3
2
+
1
2
i)
3
= i.
ii. Show that (
2
2
+
2
2
i)
2
= i.
iii. Show that for any complex number there is a polynomial p(X) = aX
2
+
bX + c R[X] such that p() = 0. (Note: a, b and c should be real
numbers).
iv. Show that for every C there is a C such that
2
= .
v. Let , , C. Assume that not both and are zero. Show that the
equation x
2
+ x + = 0 has a solution in C. Show that this equation
has at most two solutions in C. (Hint: Recall the quadratic formula and
its proof).
vi. Show that for every C and every n N 0, there is a C such
that
2
n
= .
vii. Let a
0
, a
1
, . . . , a
n
R. Show that if C is a solution of a
0
+ a
1
x +
. . . +a
n
x
n
then is also a solution of this equation.
viii. Show that there is a bijection between C : [[ > 1 and C : 0 <
[[ < 1.
ix. For , C, dene the distance, d(, ) by d(, ) = [[. Show the
following:
i. d(, ) = 0 if and only if = ,
ii. d(, ) = d(, ),
iii. d(, ) d(, ) +d(, ).
x. Show that no order satisfying the axioms O1, O2, O3, OA and OM can
be dened on C (Hint: See Corollary 2.4.5).
Chapter 4
Real Vector Spaces
Let R
n
denote the cartesian product of n copies of R, i.e.
R
n
= (r
1
, . . . , r
n
) : r
i
R for i = 1, . . . , n.
The elements of R
n
will be called vectors. We will denote a vector by x, v,
a etc. To denote the vectors we will use the following convention as a rule:
x = (x
1
, . . . , x
n
), v = (v
1
, . . . , v
n
), a = (a
1
, . . . , a
n
).
The vector (0, . . . , 0) will be denoted by

0.
Given v = (v
1
, . . . , v
n
), we let v = (v
1
, . . . , v
n
).
Given v = (v
1
, . . . , v
n
) and w = (w
1
, . . . , w
n
), we dene the sum v + w of v
and w as
v + w = (v
1
+w
1
, . . . , v
n
+w
n
).
Given v = (v
1
, . . . , v
n
) and r R, we dene the scalar multiplication of
r R with the vector v as
rv = (rv
1
, . . . , rv
n
).
With these denitions, letting V = R
n
, the following hold:
A1. Additive Associativity. For any x, y, z V , x+(y +z) = (x+y) +z.
A2. Additive Identity Element. There is a vector

0 such that for any
x V , x +
0 =
0 +x = x.
A3. Additive Inverse Element. For any x V there is a y V (namely
x) such that x +y = y +x =
0.
A4. Commutativity of the Addition. For any x, y V , x +y = y +x.
B1. Associativity of the Scalar Multiplication. For any r, s R and
x V , r(sx) = (rs)x.
43
44 CHAPTER 4. REAL VECTOR SPACES
B2. Distributivity of the Scalar Multiplication 1. For any r, s R and
x V , (r +s)x = rx +sx.
B3. Distributivity of the Scalar Multiplication 2. For any r R and
x, y V , r(x +y) = rx +ry.
B4. Identity. For any x V , 1x = x.
Note that the properties A1, A2, A3, A4 are the same as in section 2.1. Thus
(R
n
, +,
0) is a commutative group. A set V on which a binary operation + and

a scalar multiplication R V V that sends a pair (r, v) of R V to
an element of V (denoted by rv) satisfying the properties A1-A4 and B1-B4 is
called a vector space over R or a real vector space. More precisely, a real
vector space is a triple (V, +, R V V ) satisfying the properties above.
If in the above axioms R is replaced by a eld F, the resulting structure is
called a vector space over F. Note that any vector space over R is also a
vector space over the eld Q of rational numbers. But in this monograph, we
will only need real vector spaces.
As with the real numbers, one can show that the element
0 of a vector space
satisfying A2 is unique. One can also show that, given x V , the element y that
satises A3 is unique. We set y = x. Of course all the consequences of Axioms
A1-A4 investigated in Section 2.1 hold. For example, we have
(x) = (x).
Examples.
i. Let V = (x, y, z, t) R
4
: x 2y + 3t = 0. Then V is a vector space
with the usual componentwise addition and scalar multiplication.
ii. R itself is a real vector space.
iii. The set C of complex numbers is a vector space over R.
iv. The singleton set a is a vector space over R if we dene a + a = a and
ra = a for any r R. We will denote a by 0 of course.
v. Let R[x] be the set of real polynomials in x, i.e.
R[x] = a
0
+a
1
x +. . . +a
n
x
n
: a
i
R.
(We will avoid the mathematical denition of the polynomials. Keep in
mind only the fact two polynomials a
0
+a
1
x +. . . +a
n
x
n
and b
0
+b
1
x +
. . . +b
m
x
m
are equal if and only if n = m and a
i
= b
i
for all i = 0, . . . , n.
The precise mathematical denition of polynomials is given in Math 211).
We dene the addition of two polynomials
p(x) = a
0
+a
1
x +. . . +a
n
x
n
and
q(x) = b
0
+b
1
x +. . . +b
m
x
m
45
as
p(x) +q(x) = (a
0
+b
0
) + (a
1
+b
1
)x +. . . + (a
k
+b
k
)x
k
where k = max(n, m) and a
i
= 0 in case i > n and b
j
= 0 in case m > j.
For example
(1 2x
2
+x
3
) + (3 +x + 2x
2
+ 3x
5
+x
6
) = 4 +x +x
3
+ 3x
5
+x
6
.
And we dene the scalar multiplication of a real number r with a polyno-
mial a
0
+a
1
x +. . . +a
n
x
n
as
r(a
0
+a
1
x +. . . +a
n
x
n
) = ra
0
+ra
1
x +. . . +ra
n
x
n
.
With these denitions, it is easy to check that R[x] satises A1-A4 and
B1-B4, thus is a real vector space.
2. Consider the set R(x) := p/q : p R[x], q R[x] 0 with
the convention that p/q = p
1
/q
1
if and only if pq
1
= p
1
q. We dene the
addition and scalar multiplication as expected:
p
q
+
p
1
q
1
=
pq
1
+p
1
q
qq
1
r
p
q
=
rp
q
for p, q, p
1
, q
1
R[x] and r R.
With these denitions, it is easy to check that R(x) satises A1-A4 and
B1-B4, thus is a real vector space. (You may want to check that R(x) is
a eld).
vi. Let X be any set and V any real vector space. Let Func(X, V ) be the set of
all functions from the set X into V . For two functions f, g Func(X, V )
and a real number r R, we dene the functions f +g, rf Func(X, V )
as follows
(f +g)(x) = f(x) +g(x),
(rf)(x) = r f(x),
for all x X. With these denitions the set Func(X, V ) becomes a real
vector space.
The element
0 of Func(X, V ) corresponds to the function that sends every

element of X to

0 V .
Given f Func(X, V ), the element f Func(X, V ) that satises A3 is
dened by (f)(x) = f(x).
vii. Let V be a vector space. Consider the set Seq(V ) of sequences from V .
Thus by denition an element v Seq(V ) is of the form (v
n
)
nN
where v
n
V for all n N. For two sequences v = (v
n
)
n
and w = (w
n
)
n
and a real number r, set v + w = (v
n
+ w
n
)
n
and rv = (rv
n
)
n
. Then
Seq(V ) becomes a vector space. The zero element
0 of Seq(V ) is the zero

sequence that consists of zero vectors and (v
n
)
n
= (v
n
)
n
.
viii. Let V be a vector space. Consider the set Seq
f
(V ) of sequences from V
which are zero after a while. Thus by denition an element v Seq(V )
is of the form (v
n
)
nN
where v
n
V for all n N and for which v
n
=

0
for n large enough, i.e. for n N for certain N (that depends on v).
For two such sequences v = (v
n
)
n
and w = (w
n
)
n
and a real number
r, set v + w = (v
n
+ w
n
)
n
and rv = (rv
n
)
n
. Then Seq
f
(V ) becomes a
vector space. The zero element

0 is the zero sequence of zero vectors and
(v
n
)
n
= (v
n
)
n
.
Exercises. Suggestion: First do Exercise i.
i. Let V be a real vector space. Let W be a subset of V . Suppose that
W is closed under addition and scalar multiplication, i.e. suppose that
w + w
W for all w, w
W and rw W for all r R and w W.

Show that W is a vector space. Such a subset is called a subspace of V .
ii. Are the following sets vector space under the usual addition and scalar
multiplication?
(x, y, z) R
3
: x +y +z = 0
(x, y, z) R
3
: x +y +z = 1
(x, y, z) R
3
: xyz = 0
(x, y, z) R
3
: x
2
+y
2
+z
2
= 0
(x, y, z) R
3
: x
3
+y
3
+z
3
= 0
(x, y, z) R
3
: xyz 0
(x, y, z) R
3
: x Q
iii. Which of the following are not vector spaces over R (with the componen-
twise addition and scalar multiplication) and why?
V
1
= (x, y, z) R
3
: xy 0
V
2
= (x, y, z) R
3
: 3x 2y +z = 0
V
3
= (x, y, z) R
3
: xyz Q
V
4
= (x, y) R
3
: x +y 0
V
5
= (x, y) R
2
: x
2
+y
2
= 0
V
6
= (x, y) C
2
: x
2
+y
2
= 0
iv. Are the following sets vector space under the usual addition of functions
and scalar multiplication of a function with a real number?
f Func(X, R) : f(x) = 0 where x is a xed element of X
f Func(R, R) : f(x) = 0 for any x (0, 1)
f Func(R, R) : f(0) 0
v. Let V and W be two real vector spaces. Consider the set Hom(V, W)
of functions f : V W such that f(v
1
+ v
2
) = f(v
1
) + f(v
2
) and
f(rv) = rf(v) for all v, v
1
, v
2
V, r R Show that Hom(V, W) is vector
space under the usual addition of functions and scalar multiplication (see
Example vi, page 45).
47
vi. Show that R(x) is a eld. Find all orders on R(x) that makes it an
ordered eld.
Chapter 5
Metric Spaces
5.1 Examples.
Real Numbers and Absolute Value. Let us consider the function d :
R R R
0
dened by d(x, y) = [x y[. This function has the follow-
ing properties: For all x, y, z R,
i. d(x, y) = 0 if and only if x = y,
ii. d(x, y) = d(y, x),
iii. d(x, y) d(x, z) +d(z, y).
We will see several examples of sets X together with a function d : XX
R
0
satisfying the three properties above.
Euclidean Spaces. For x = (x
1
, . . . , x
n
) R
n
and y = (y
1
, . . . , y
n
) R
n
dene
d(x, y) =
_
[x
1
y
1
[
2
+. . . +[x
n
y
n
[
2
.
We will show that d satises the properties i, ii, iii stated above. The rst two
are immediate. It will take us some time to prove the third equality.
For two vectors x = (x
1
, . . . , x
n
) and y = (y
1
, . . . , y
n
), dene their scalar
product as
xy =
n
i=1
x
i
y
i
.
We have for all x, y, z R
n
and , R,
C1. (x +y)z = xz +yz
C2. xy = yx
C3. x(y +z) = xy +xz
(C1 and C2 are direct consequences of the denition. C3 follows from C1 and
C2.) Note that xx 0, so we can take its square root. Let |x| = (xx)
1/2
. It is
49
50 CHAPTER 5. METRIC SPACES
easy to show from the denition that, for all x R
n
and R, we have
C4. |x| 0 ,and |x| = 0 if and only if x = 0
C5. |x| = [[|x|
Now we show that for all x R
n
we have
C6. [xy[ |x| |y|
Indeed, let z = xy where = xy and = |x|
2
. Use |z|
2
0 to prove C6.
Details: 0 (x y)(x y) =
2
xx 2xy +
2
yy =
2
|x|
2
2xy +
2
|y|
2
= (xy)
2
|x|
2
2(xy)
2
|x|
2
+|x|
4
|y|
2
= (xy)
2
|x|
2
+|x|
4
|y|
2
. Hence
(xy)
2
|x|
2
|x|
4
|y|
2
. Simplifying we get (xy)
2
|x|
2
|y|
2
. Taking the square
roots we get xy |x| |y|. This proves C6.
Now we prove
C7. |x +y| |x| +|y|
For this we use C6 and the fact that
|x +y|
2
= |x|
2
+ 2xy +|y|
2
.
Details: |x+y|
2
= (x+y)(x+y) = |x|
2
+2xy +|y|
2
|x|
2
+2[xy[ +|y|
2
A6
|x|
2
+2[|x| |y| +|y|
2
= (|x| +|y|)
2
. Thus |x +y|
2
(|x| +|y|)
2
. Taking
the square roots we nally obtain C7.
We now remark that d(x, y) = |xy|
1/2
and we prove the third property (iii)
above: d(x, z) = |xz|
1/2
= (|(xy)+(yz)|)
1/2
C7
= (|xy|+|yz|)
1/2
|x y|
1/2
+|y z|
1/2
= d(x, y) +d(y, z).
Thus we have proved the following result.
Theorem 5.1.1 For x = (x
1
, . . . , x
n
) R
n
and y = (y
1
, . . . , y
n
) R
n
dene
d(x, y) =
_
[x
1
y
1
[
2
+. . . +[x
n
y
n
[
2
.
Then the couple (R
n
, d) satises the properties i, ii, iii above.
5.2 Denition and Further Examples
A set X together with a function d : X X R
0
is said to be a metric
space if for all x, y, z X,
MS1. d(x, y) = 0 if and only if x = y.
MS2. d(x, y) = d(y, x).
MS3. (Triangular Inequality.) d(x, y) d(x, z) +d(z, y).
The function d is called a metric or a distance function on X.
We will give several examples of metric spaces.
5.2. DEFINITION AND FURTHER EXAMPLES 51
Euclidean Metric. By Theorem 5.1.1, (R
n
, d) where
d(x, y) =
_
[x
1
y
1
[
2
+. . . +[x
n
y
n
[
2
is a metric space, called the Euclidean metric space, or sometimes the usual
metric on R
n
.
Metric on a Product. The following theorem will be proven much later
(Lemma 14.5.4), when we will give a sense to the term x
r
for x R
0
and
r R0. In any event, the theorem below makes sense for p N even at this
point.
Theorem 5.2.1 For i = 1, . . . , n, let (X
i
,
i
) be a metric space. Let X =
X
1
. . . X
n
. Let p 1 be any real number. For x, y X let d
p
(x, y) =
(
n
i=1
i
(x
i
, y
i
)
p
)
1/p
. Then (X, d
p
) is a metric space.
If we take X
i
= R and p = 2 in the theorem above with d
i
(x, y) = [x y[,
we get the Euclidean metric on R
n
.
Unless stated otherwise, when we speak about the product of nitely metric
spaces, we will always take p = 2.
Induced Metric. Any subset of a metric space is a metric space with the
same metric, called the induced metric. More precisely we dene the induced
metric as follows: Let (X, d) be a metric space. Let Y X. Then (Y, d
|Y Y
)
is a metric space. We say that the metric of Y is induced from that of X.
Sup Metric. For i = 1, . . . , n, let (X
i
, d
i
) be a metric space and let X =
X
1
. . . X
n
. For x, y X, set
d
(x, y) = maxd
i
(x
i
, y
i
) : i = 1, . . . , n.
Then (X, d
) is a metric space as it can be checked easily.

Discrete Metric. Let X be any set. For x, y X, set
d(x, y) =
_
1 if x ,= y
0 if x = y
Then (X, d) is a metric space. This metric on the set X is called the discrete
metric.
p-adic Metric. Let X = Z and p a positive (prime) integer. For distinct
x, y Z, set d(x, y) = 1/p
n
if p
n
divides x y but p
n1
does not divide x y.
Set d(x, x) = 0. Then (X, d) satises the properties i, ii and iii stated above. In
fact it satises the stronger triangular inequality d(x, y) maxd(x, z), d(z, y).
Such a metric is called ultrametric. Even further d(x, y) = maxd(x, z), d(z, y)
if d(x, z) ,= d(y, z). It follows that any triangle in this space is isosceles.
Metric on Sequences. Let X be any set. Consider the set Seq(X) of se-
quences from X. Thus an element x Seq(X) is of the form (x
n
)
nN
where
x
n
X for all n N. For two distinct sequences x = (x
n
)
n
and y = (y
n
)
n
, set
d(x, y) = 1/2
n
if n is the rst natural number where x
n
,= y
n
. Then (Seq(X), d)
is a metric space. Note that, in this metric, the maximum distance between two
sequences is 1.
We end this subsection with a simple but useful calculation.
Proposition 5.2.2 In a metric space (X, d), for all x, y, z X, we have
[d(x, z) d(z, y)[ d(x, y).
Proof: We have to show that d(x, z) d(z, y) d(x, y) and d(z, y) d(x, z)
d(x, y).
We have d(x, z) d(x, y) +d(y, z) = d(x, y) +d(z, y), so the rst inequality
follows. The second one is similar and is left as an exercise.
Exercises.
i. Show that the discrete metric is ultrametric.
ii. Show that the metric on the set of sequences dened above is an ultra-
metric.
5.3 Normed Real Vector Spaces and Banach Spaces
and Algebras
Most of what one can say about R can be said mot à mot about R
n
and C, and
most of what one can say about R
n
and C can be said more generally (mot à
mot again) for Banach spaces and algebras, concepts that generalize R, R
n
, C
and C
n
and that will be dened in this section. Since the proofs are exactly the
same, we prefer treat the most general case. Readers who will be psychologically
aected by this generality may prefer to read this treatise by considering only
the cases of R and R
n
, and sometimes C.
A normed real vector space is a vector space V together with a map
[ [ : V R such that
NVS1. For all v V , [v[ 0.
NVS2. For all v V , [v[ = 0 if and only if v = 0.
NVS3. For all v V and r R, [rv[ = [r[[v[.
NVS4. For all v, w V , [v +w[ [v[ +[w[.
i. The Euclidean space R
n
with the usual norm is a normed space.
5.3. NORMEDREAL VECTOR SPACES ANDBANACHSPACES ANDALGEBRAS53
ii. Consider the set
R of real sequences (r
n
)
n
whose terms are all zero
except for nitely many of them. It is clear that
R is a real vector
space. Dene [(r
n
)
n
[ =
n=0
r
2
n
. The reader should show that this turns
R into a normed vector space.

iii. Let X be a set and (V, [ [) a normed vector space. Consider the set
B(X, V ) of functions from X into Y which are bounded. Thus, a function
f : X V is in B(X, V ) if and only if there is a real number M such
that for all x X, [f(x)[ < M. Show that B(X, V ) is a vector space.
For f B(X, V ), dene [f[ to be the minimum of the numbers M as
above. Thus
[f[ = sup[f(x)[ : x X.
Show that the vector space B(X, V ) together with [ [ dened as above is
a normed vector space.
Suppose V = R. Show that if f, g B(X, R), then f g B(X, R) and
that [f g[ = [f[ [g[.
iv. Let X be a set and (V, [ [) a normed vector space. Consider the set T(X, V )
of functions from X into Y . Show that T(X, V ) is a vector space.
For f T(X, V ), dene
[f[ = infsup[f(x)[ : x X, 1.
Show that the vector space T(X, V ) together with [ [ dened as above is
a normed vector space.
Suppose V = R. Is it true that if f, g B(X, R), then f g T(X, R)
and that [f g[ = [f[ [g[.
Proposition 5.3.1 Let (V, [ [) be a normed real vector space. The map d(v, w) =
[v w[ denes a metric on V .
Proof: Easy.
A normed real metric space which is complete with respect to the metric
dened above is called a Banach space.
Dene Banach algebra.
Problem. Suppose V is a normed real vector space and W V a subspace.
Can you make V/W into a normed vector space in a natural way?
Exercises.
i. Show that if V is a Banach space, then so is B(X, V ).
ii. Show that if V is a Banach space, then so is T(X, V ).
5.4 Open Subsets of a Metric Space
Let (X, d) be a metric space, a X and r R. Then the open ball with
center a and radius r is dened as
B(a, r) = x X : d(a, x) < r.
The circle with center a and radius r is dened as
B(a, r) = x X : d(a, x) = r.
An arbitrary union of open balls in a metric space is called on open subset.
Proposition 5.4.1 A subset U of a metric space is open if and only if for every
a U there is an r R
>0
such that B(a, r) U.
Proof: () Let a U. Since U is a union of open balls, there is an open
ball in U that contains a. Set a B(b, s) U. Let r = s d(a, b). Then
r > 0. We claim that B(a, r) B(b, s). Indeed, let x B(a, r). Then d(x, b)
d(x, a) + d(a, b) < r + d(a, b) = r + (s r) = s. Thus x B(b, s). This proves
the claim. Now B(a, r) B(b, s) U.
() Suppose that for any a U, there is an r
a
R
>0
such that B(a, r
a
)
U. Since r
a
> 0, a B(a, r
a
). It follows that U =

aU
B(a, r
a
). Hence U,
being a union of open balls, is open.
Lemma 5.4.2 In a metric space, the intersection of two open balls is an open
subset.
Proof: Let B(a, r) and B(b, s) be two open subsets of a metric space (X, d).
To show that B(a, r) B(b, s) is open, we will use Proposition 5.4.1. Let c
B(a, r) B(b, s). Then r d(c, a) > 0 and s d(c, b) > 0. Let
= min(r d(c, a), s d(c, b)).
Then > 0. We claim that B(c, ) B(a, r) B(b, s). Let x B(c, ). Then
d(a, x) d(a, c) + d(c, x) < d(a, c) + d(a, c) + (r d(a, c)) = r. Thus
x B(a, r). Similarly x B(b, s).
Proposition 5.4.3 Let (X, d) be a metric space. Then,
i. and X are open subsets.
ii. An arbitrary union of open subsets is open.
iii. A nite intersection of open subsets is an open subset.
Proof: i. If a X, then the ball with center a and radius 0 is the emptyset.
Hence the emptyset is open (being an open ball). The whole space X is the
union of all the open balls of radius, i.e. X =
[x XB(x, 1). Thus X is open.

ii. This is clear by denition of an open subset.
5.4. OPEN SUBSETS OF A METRIC SPACE 55
iii. Let U and V be two open subsets. Thus U and V are unions of open
balls. Write U =

aA
B(a, r
a
) and V =

bB
B(b, s
b
) where A and B are
subsets of X and r
a
and s
b
are real numbers. Now
U V =
_
aA
B(a, r
a
)
_
bB
B(b, s
b
) =
_
aA, bB
(B(a, r
a
) B(b, s
b
)).
Thus, by part ii, it is enough to show that B(a, r
a
) B(b, s
b
) is open. But this
is the content of Lemma 5.4.2.
Exercises.
i. Describe B(
0 , 1) in the three metric spaces (R

n
, d
1
), (R
n
, d
2
) and (R
n
, d
)
dened above.
ii. Consider R
n
with one of the distances d
p
(p 1) or d
dened above.
Show that any open subset of R
n
is a countable union of open balls.
iii. An isometry between two metric spaces (X, d
1
) and (Y, d
2
) is a bijection
f : X Y such that d
1
(x
1
, x
2
) = d
2
(f(x
1
), f(x
2
)). Find all isometries
of R into R.
iv. Find all isometries of R
2
into R
2
.
v. Show that the metric spaces (R
n
, d
p
) and (R
n
, d
) dened above all have

the same open subsets.
vi. Let X = 1/2
n
: n N R. Consider X with the natural metric (the
induced metric). Show that any subset of X is open.
vii. Let X = 1/2
n
: n N 0 R. Consider X with the natural
metric (the induced metric). Show that open subsets of X are the conite
subsets
1
of X and the ones that do not contain 0.
viii. Show that a closed ball B(a, r) = x R
n
: d(a, x) r is not open in
the Euclidean metric unless r < 0.
ix. Let (X, d) be a metric space. Show that d
1
(x, y) =
d(x,y)
1+d(x,y)
is a metric.
Show that the open subsets in both metrics are the same.
x. Show that R cannot be the union of two nonempty open subsets (for the
usual metric).
xi. Show that R
n
cannot be the union of two nonempty open subsets (for the
usual metric).
1
A subset Y of X is called conite if X \ Y is nite.
Chapter 6
Sequences and Limits
6.1 Denition
Let (X, d) be a metric space, (x
n
)
nN
a sequence in X and x X. We say that
the sequence (x
n
)
nN
converges to x if for any > 0 there is an N N such that
d(x
n
, x) < for all natural numbers n > N. We then write lim
n
x
n
= x. We
call x the limit of the sequence (x
n
)
n
. When the limit of a sequence (x
n
)
n
exists
we say that the sequence (x
n
)
n
is convergent or that it converges, otherwise
we say that the sequence is divergent or that it diverges.
Note that, above, we used the symbol without dening it. Above we
only dened lim
x
x
n
= x as if it were one single word. We will never dene
the symbol , there is no such an object in mathematics.
i. A sequence (x
n
)
nN
where x
n
= x for all n N converges to x. Such a
sequence is called a constant sequence.
ii. A sequence (x
n
)
nN
where x
n
= x for all n greater than a certain n
converges to x. Such a sequence is called an eventually constant se-

quence.
iii. If we delete nitely many elements from a sequence or add nitely many
elements to a sequence, its divergence and convergence remains unaltered,
and its limit (if it exists) does not change.
iv. Let x
n
= n. Then the sequence (x
n
)
n
does not converge in the usual
metric of R.
v. Let x
n
= (1)
n
. Then the sequence (x
n
)
n
does not converge in the usual
metric of R.
57
58 CHAPTER 6. SEQUENCES AND LIMITS
vi. Let X be a metric space. Let a
1
, . . . , a
k
X. Let (x
n
)
n
be a sequence
of X such that for each n, x
n
= a
i
for some i = 1, . . . k. Show that the
sequence (x
n
)
n
converges if and only if it is eventually constant.
vii. Let x
n
= 0 if n is not a power of 2 and x
2
n = 1. Then the sequence (x
n
)
n
does not converge in the usual metric of R.
viii. Consider Z with the 2-adic metric and set x
n
= 2
n
. Then the sequence
(x
n
)
n
converges to 0.
ix. Consider a set X with its discrete metric. Then a sequence (x
n
)
n
converges
to some element in X if and only if the sequence (x
n
)
n
is eventually
constant.
x. Show that if in Theorem 3.1.12, lim
n
(b
n
a
n
) = 0, then
nN
[a
n
, b
n
]
is a singleton set.
Remarks.
i. In the denition of convergence, should be thought as a small (but
positive) real number. The integer N depends on . The smaller is,
the larger N should be.
ii. In the denition of convergence, we could replace d(x
n
, x) < by d(x
n
, x)
, i.e. the sequence (x
n
)
nN
converges to x if and only if for any > 0
there is an N N such that d(x
n
, x) for all natural numbers n > N.
iii. In the denition of convergence, since Q is dense in R, we could take to
range over Q instead of over R.
iv. The convergence of a sequence depends strongly on the metric. It is pos-
sible that a sequences converges to some element for a certain metric, but
that this same sequence diverges for some other metric.
v. The notation lim
x
x
n
= x suggests that the limit of a sequence is
unique when it exists. This is indeed the case and will be proven right
now.
Theorem 6.1.1 In a metric space, the limit of a sequence, if it exists, is unique.
Proof: Let (X, d) be a metric space, (x
n
)
n
a sequence from X and a, b X.
Suppose that a and b are limits of the sequence (x
n
)
n
. We will show that a = b
by showing that d(a, b) < for any > 0.
Let > 0. Since a is a limit of (x
n
)
n
there is an N
such that d(x

n
, a) < /2
for all n > N
. Similarly there is an N
1
such that d(x
n
, b) < /2 for all n > N
1
.
Let N = max(N
, N
1
) + 1. Then d(x
N
, a) < /2 and d(x
N
, b) < /2. Hence
d(a, b) d(a, x
N
) +d(x
N
, b) < /2 +/2 = .
Lemma 6.1.2 In a metric space, lim
n
x
n
= x if and only if lim
n
d(x
n
, x) =
0.
6.1. DEFINITION 59
Proof: This is a triviality.
A subset of a metric space is said to be bounded if it is contained in a
ball. For example Z is not bounded in R (with its usual metric). But the set
1/2
n
: n N is bounded. A sequence (x
n
)
n
is called bounded if the set
x
n
: n N is bounded.
Theorem 6.1.3 A convergent sequence is bounded.
Proof: Let (x
n
)
n
be a sequence converging to some a. Take = 1 in the
denition of the convergence. Thus there exists an N such that d(x
n
, a) < 1
for any n > N. Let r = max(d(x
1
, a), . . . , d(x
N
, a), 1) + 1. Then the sequence
(x
n
)
n
is entirely in B(a, r).
Exercises
i. Show that lim
n
1/n = 0 in the usual metric of R.
ii. For a natural number n we dene (n) = supm N : 2
m
n. Show
that lim
n
(n)/n = 0.
iii. Let x
n
= 0 if n is not a power of 2 and x
2
n = 1/n. Show that the sequence
(x
n
)
n
converges to 0 in the usual metric of R.
iv. Show that a sequence of integers converges in the usual metric of R if and
only if it is eventually constant.
v. Let x
n
=
n
k=0
(1)
nk
_
n
k
_
2
k
. Does the sequence (x
n
)
n
converge?
vi. Let (x
n
)
n
be a real sequence. Assume that there is an r > 0 such that for
all n ,= m, [x
n
x
m
[ > r. Show that the sequence (x
n
)
n
diverges.
vii. Let (X, d) be a metric space, (x
n
)
n
a sequence in X and x X. Show
that lim
n
x
n
= x if and only if lim
n
d(x
n
, x) = 0.
viii. Let (a
n
)
n
be a sequence of nonnegative real numbers. Suppose that the
sequence (a
2
n
)
n
converges to a. Show that the sequence (a
n
)
n
converges
to

a. Does this hold for sequences which are not nonnegative?
ix. Let X be the set of sequences of zeroes and ones. For two distinct elements
x = (x
n
)
n
and y = (y
n
)
n
of X dene d(x, y) = 1/2
n
if n is the least natural
number for which x
n
,= y
n
. Let d(x, x) = 0. Then (X, d) is a metric space.
Let
n
be the element of X whose rst n coordinates are 1, the rest is 0.
Show that lim
n
n
exists. Find the limit.
x. Let X
be the set of sequences of zeroes and ones with only nitely many
ones. Consider X
as a metric space with the metric dened above. Let
n
be the element of X considered above. Note that
n
X
. Show that
lim
n
n
does not exist in X
(but it exists in X as the above exercise

shows).
xi. Let X = Q with the 3-adic distance. Let x
n
= 1 +3 +3
2
+. . . +3
n
. Show
that (x
n
)
n
is convergent in Q.
xii. Let X
= Z with the 3-adic distance. Let x

n
= 1 +3 +3
2
+. . . +3
n
X
.
Show that (x
n
)
n
is not convergent in Z (but it is in Q).
xiii. Let A be a bounded subset of a metric space (X, d). Show that for any
b X there is a ball centered at b that contains A.
xiv. Let A and B bounded subsets of R
n
with its usual Euclidean metric. Let
A+B = a +b : a A, b B
and
AB = ab : a A, b B.
Show that A+B and AB are bounded subsets of R
n
.
xv. Let A be a bounded subset of R (with its usual metric) that does not
contain 0. Let
A
1
= a
1
; a A.
a) Show that A
1
is not necessarily bounded.
b) Assume there is an r > 0 such that B(0, r) A = . Show that A
1
is
bounded.
c) Conversely show that if A
1
is bounded then there is an r > 0 such
that B(0, r) A = .
xvi. Let (x
n
)
n
be a sequence of R (with its usual metric). Show that (x
n
)
n
is
bounded if and only if there is a real number r such that [x
n
[ r for all
n N.
xvii. Show that any real number is the limit of a rational sequence.
6.2 Examples of Convergence in R and C
The most important cases for us are the case of R and C with their usual metric
d(x, y) = [xy[. Translating the denition of convergence of a sequence to this
case, we obtain the following special case: A sequence (x
n
)
n
of real or complex
numbers converges to x R if and only if for any > 0 there is an N N such
that [x
n
x[ < for all natural numbers n > N. We know that the limit x,
when it exists, is unique.
In this subsection, we prove the convergence and divergence of some impor-
tant sequences: (1/n)
n=1
, (
n
)
n
and
n
/n! for any r. In the meantime we prove
a criterion for convergence (Sandwich Lemma).
The following result will be handy.
Theorem 6.2.1 Let
n
C. Then lim
n
n
= 0 if and if lim
n
[
n
[ = 0.
Proof: This is an immediate consequence of the denition.
6.2. EXAMPLES OF CONVERGENCE IN R AND C 61
6.2.1 The Sequence (1/n)
n
Let us start by showing that lim
n
1/n = 0.
Lemma 6.2.2 lim
n
1/n = 0.
Proof: Let > 0 be xed. By Archimedean Property of the real numbers
(Theorem 3.1.6), there is a natural number N such that N > 1. Then 1/N < .
Now for all n > N, [1/n 0[ = 1/n < 1/N < .
Exercises.
i. Prove that lim
n
_
n+3
n
2
+n5
_
= 0.
ii. Prove that lim
n
_
n+3
n
2
+n5
_
n
= 0.
iii. Prove that lim
n
_
n+3
n
2
+n5
_
n1
3n+2
= 0.
iv. Prove that lim
n
_
n+3
n
2
+n5
_
n
2
1
2n3
= 0.
n
)
n
Proposition 6.2.3 If r (1, 1) then the sequence (r
n
)
n
converges to 0. If
r , (1, 1] the sequence (r
n
)
n
diverges.
Proof: We rst claim that if s > 1, then for all natural numbers n, (1+s)
n
1+ns. We proceed by induction on n to prove the claim. The claim clearly holds
for n = 0. We now assume the claim holds for n and prove it for n + 1. Since
1 +s > 0, (1 +s)
n+1
= (1 +s)
n
(1 +s) (1 +ns)(1 +s) = 1 +(1 +n)s +ns
2
1 + (1 +n)s. This proves the claim.

We now return to the proof of the proposition. Assume r (1, 1). Since
for r = 0 the statement is clear, we may assume that r ,= 0. Let > 0 be a
real number. We have to show that there exists a natural number N such that
for all n > N, [r[
n
= [r
n
0[ < . We may therefore assume that r 0. Thus
r > 0. Let s = 1/r 1. Note that r =
1
1+s
. It is also easy to check that s > 0.
Thus by the claim (1 +s)
n
1 +ns. Let N > 0 be a natural number such that
1/s N (Theorem 3.1.6). Now for all n > N,
[r
n
[ = r
n
=
_
1
1 +s
_
n
1
1 +ns
<
1
1 +Ns
<
1
Ns
.
This proves the rst part of the proposition.
Claim. Let r > 1. Then the sequence (r
n
)
n
is unbounded.
Proof of the Claim. By the rst part of the proposition the sequence
(1/r
n
)
n
converges to 0. Hence for all > 0 there is an N such that for all
n > N, 1/r
n
< , i.e. r
n
> 1/. Thus (r
n
)
n
is unbounded. This proves the
claim.
Now by Theorem 6.1.3, (r
n
)
n
is divergent if r > 1. The case r 1 is left
as an exercise.
Corollary 6.2.4 If r > 1 then the sequence (r
n
)
n
is unbounded.
Corollary 6.2.5 Let C. Then the sequence (
n
)
n
converges to 0 if [[ < 1,
converges to 1 if = 1, diverges otherwise.
Proof: If [[ < 1 then the result follows from Theorem 6.2.1 and Proposition
6.2.3. If [[ = 1 then the statement is clear. Assume [[ > 1. Then the sequence
([
n
[)
n
is unbounded by Corollary 6.2.4. By Theorem 6.1.3, (
n
)
n
is divergent.
Exercises.
i. Find lim
n
_
1
n
_
n
.
ii. Find lim
n
_
1
2
+
1
n
_
n
.
n
/n!)
n
Proposition 6.2.6 For any r R, the sequence (r
n
/n!)
n
converges to 0.
Proof: We may assume that r > 0. Let > 0. Let x
n
= r
n
/n!. Let N
be a
natural number such that r/N
< 1 (Theorem 3.1.6). We claim that for each

natural number k,
x
N+k
x
N
_
r
N
+ 1
_
k
.
We prove this by induction on k. The claim is clear for k = 0. Assume it holds
for k. Then
x
N+k+1
=
r
N
+k+1
(N
+k + 1)!
=
r
N
+k
(N
+k)!
r
N
+k + 1
= x
N+k
r
N
+k + 1
x
N
_
r
N
+ 1
_
k
r
N
+k + 1
x
N
_
r
N
+ 1
_
k+1
.
This proves the claim.
Since 0 <
r
N+1
< 1, by Proposition 6.2.3, the sequence ((
r
N+1
)
k
)
k
converges
to 0. Therefore there exists an N
1
such that for k N
1
, (
r
N
+1
)
k
< /x
N
.
Thus for n > N
+N
1
, by letting k = n N
1, we see that x
n
< . This
proves the proposition.
Corollary 6.2.7 For any C, the sequence (
n
/n!)
n
converges to 0.
Proof: Follows from Theorem 6.2.1 and Proposition 6.2.6.
6.3. CONVERGENCE AND THE ORDER 63
6.3 Convergence and the Order
We rst prove a result that will show that a sequence that is squeezed between
two sequences converging to the same number converges.
Lemma 6.3.1 (Sandwich Lemma) Let (x
n
)
n
, (y
n
)
n
and (z
n
)
n
be three real
sequences. Suppose that lim
n
x
n
and lim
n
z
n
exist and are equal. Suppose
also that x
n
y
n
z
n
eventually (i.e. for n > N
for some N
). Then
lim
n
y
n
exists and lim
n
x
n
= lim
n
y
n
= lim
n
z
n
.
Proof: Let a be the common limit of (x
n
)
n
and (z
n
)
n
. Let > 0. Since
lim
n
x
n
= a there is an N
1
such that for n > N
1
, [x
n
a[ < /3. Since
lim
n
z
n
= a there is an N
2
such that for n > N
2
, [z
n
a[ < /3. Let
N = max(N
, N
1
, N
2
). Then for n > N, we have [y
n
a[ [y
n
z
n
[ +[z
n
a[
(z
n
x
n
) +[z
n
a[ (z
n
a) + (a x
n
) +[z
n
a[ < /3 +/3 +/3 = .
Corollary 6.3.2 lim
n
1/n
2
= 0.
Proof: Follows from Lemmas 6.2.2 and 6.3.1.
Second proof: Let > 0. Let N > 1 be such that 1/2 < N. Then for all
n > N, [1/n
2
0[ = 1/n
2
1/2n < 1/2N < .
Corollary 6.3.3 Let (x
n
)
n
be a converging sequences and r R. Assume that
x
n
r. Then lim
n
x
n
r.
Note that we could well have x
n
> r and lim
n
x
n
= r, for example take
x
n
= r + 1/n.
n
)
n
and (y
n
)
n
be two converging sequences. Suppose
that x
n
y
n
for almost all n. Then lim
n
x
n
lim
n
y
n
.
Exercises.
i. Find lim
n
_
1
n
_
n
.
ii. Show that lim
n
_
1
2
+
1
n
_
n
= 0.
iii. Show that lim
n
(
n
2
n n) = 1/2.
iv. Let (x
n
)
n
be a convergent sequence in C (or a Banach space X), r > 0
and assume that [x
n
[ r for all n. Show that [ lim
n
x
n
[ r.
v. Let (x
n
)
n
be a convergent complex sequence. Show that lim
n
[x
n
[ =
[ lim
n
x
n
[. (Solution: Let x be the limit. Since [[a[ [b[[ [a b[ for
all a, b C, [[x
n
[ [x[[ [x
n
x[. Thus it is enough to show that the
limit of the right hand side is 0, which is obvious.)
6.4 Convergence and the Four Operations
In this subsection we study the behavior of the convergence of real sequences
under several operations.
We rst show that a converging sequence of R that does not converge to an
element a is bounded away from a. We will need this result later in this
subsection.
Proposition 6.4.1 Let X be a metric space and a X. Let (x
n
)
n
be a con-
verging sequence of X that does not converge a. Then there is an r > 0 such
that B(a, r) x
n
: n N is nite.
Proof: Let b be the limit of (x
n
)
n
. Let r = d(a, b)/2. Then r > 0. Since b is
the limit of (x
n
)
n
, there exists an N such that d(x
n
, b) < r for all n > N. Thus,
for n > N, d(x
n
, a) [d(x
n
, b) d(a, b)[ = [d(x
n
, b) 2r[ = 2r d(x
n
, b) > r.
(We used Lemma 5.2.2 in the rst inequality). Hence B(a, r) x
n
: n N
x
1
, . . . , x
N
.
Now we analyze the relationship between the four operations on R and the
convergence of sequences.
Theorem 6.4.2 Let (x
n
)
n
and (y
n
)
n
be sequences converging to a and b re-
spectively. Then the sequences (x
n
+y
n
)
n
and (x
n
y
n
)
n
converge to a+b and ab
respectively. In particular, for r R the sequence (rx
n
)
n
converges to ra.
Proof: (+). Let > 0. Since the sequence (x
n
)
n
converges to a, there is an
N
such that [x
n
a[ < /2. Since the sequence (y
n
)
n
converges to b, there
is an N
1
such that [x
n
a[ < /2. Let N = max(N
, N
1
). Then for n > N,
[(x
n
+y
n
) (a +b)[ = [(x
n
a) +(y
n
b)[ [x
n
a[ +[y
n
b[ < /2 +/2 = .
Thus (x
n
+y
n
)
n
converges to a +b.
(). Assume rst b = 0. We have to show that lim[n x
n
y
n
= 0. Let
> 0. By Exercise xvi, page 60, there exists an r > 0 such that [x
n
[ r for all
n. Since (y
n
)
n
converges to 0, there exists an N such that for all n > N, [y
n
[ =
[y
n
0[ <

r
. Now for n > N, we have [x
n
y
n
0[ = [x
n
y
n
[ = [x
n
[[y
n
[ < r
r
= .
Thus the sequence (x
n
y
n
)
n
converges to 0.
Assume next b ,= 0. Let > 0. Since b ,= 0 and (x
n
)
n
converges to a,, there
exists an N
such that [x
n
a[

2|b|
for all n > N
. Since (x
n
)
n
is a convergent
sequence, it is bounded by Theorem 6.1.3. By Exercise xvi, page 60, there is an
M > 0 such that [x
n
[ < M for all n N. Since (y
n
)
n
converges to b, there exists
an N
1
such that for n > N
1
, [y
n
b[ <

2M
. Let N = max(N
, N
1
). For n > N,
we have [x
n
y
n
ab[ = [x
n
(y
n
b) + (x
n
a)b[ [x
n
(y
n
b)[ + [(x
n
a)b[
[x
n
[[y
n
b[ + [x
n
a[[b[ M

2M
+

2|b|
[b[ = . Thus the sequence (x
n
y
n
)
n
converges to ab.
For the last part: Take y
n
= r for all n apply the above part.
i
)
i
and (y
i
)
i
be sequences converging to a and b respec-
tively. Then the sequence (x
i
y
i
)
i
converges to a b.
6.4. CONVERGENCE AND THE FOUR OPERATIONS 65
i
)
i
be a sequence converging to a. Let k N0. Then
the sequence (x
k
i
)
i
converges to a
k
.
i
)
i
and (y
i
)
i
be sequences converging to a and b respec-
tively. Let , C. Then the sequence (x
i
+y
i
)
i
converges to a b.
i
)
i
be a sequence converging to a. Let a
, a
1
, . . . , a
k

C. Then the sequence (a
+a
1
x
i
+. . . +a
k
x
k
i
)
i
converges to a
+a
1
a+. . . +a
k
a
k
.
If a
, a
1
, . . . , a
k
C, then a term of the form a
0
+a
1
x+. . . +a
k
x
k
is called a
(complex) polynomial. The numbers a
, a
1
, . . . , a
k
are called its coecients.
If a
k
,= 0, then the natural number k is called the degree of the polynomial
p(x). A polynomial p(x) = a
0
+a
1
x+. . . +a
k
x
k
gives rise to a function from C
into C whose value at C is p(r) := a
0
+a
1
+. . . +a
k
k
. Thus the corollary
above can be translated as follows:
n
)
n
be a sequence converging to . Let p(x) R[x].
Then the sequence (p(x
n
))
n
converges to p().
In general, if the sequence (x
n
)
n
is convergent and x
n
,= 0 for all n, then the
sequence (x
1
n
)
n
is not necessarily convergent. This is what happens if we take
x
n
= 1/n. However if the limit of (x
n
)
n
is not zero, then the convergence of the
sequence (x
1
n
)
n
is guaranteed as the following theorem shows:
n
)
n
be a sequence of C converging to a nonzero number
. Assume x
n
,= 0 for all n. Then the sequence (x
1
n
)
n
converges to
1
.
Proof: Let > 0. By Proposition 6.4.1, there is an r > 0 such that B(0, r)
x
n
: n N is nite. Let N
be such that [x
n
[ r for all n > N
. Since a and
r are nonzero and since the sequence (x
n
)
n
converges to , there exists an N
1
such that [x
n
[ < for all n > N
1
. Let N = max(N
, N
1
). Now for n > N
we have,
[x
1
n
a
1
[ =
[x
n
a[
[a[[x
n
[

[x
n
a[
[a[r
<
[a[r
[a[r
= .
Thus the sequence (x
1
n
)
n
converges to
1
.
n
)
n
and (y
n
)
n
be sequences converging to and re-
spectively. Assume ,= 0 and that y
n
,= 0 for all n. Then the sequence (x
n
/y
n
)
n
converges to /.
Corollary 6.4.10 Let p(x) and q(x) ,= 0 be two polynomials of degree d and e
respectively. Then the sequence (p(n)/q(n))
n
converges if and only if e f. If
e > f, then the limit is zero. If e = f then the limit is / where and are
the leading coecients of p(x) and q(x) respectively.
Corollary 6.4.11 Assume lim
n
a
n
exists and is nonzero. Then the se-
quence (a
n
/a
n+1
)
n
converges to 1.
The sequence (a
n
/a
n+1
)
n
may not converge if lim
n
a
n
= 0. For example,
choose
a
n
=
_
1/n if n is even
1/n
2
if n is odd
Clearly lim
n
a
n
= 0, but
a
n
a
n+1
=
_
_
_
(n+1)
2
n
if n is even
n+1
n
2
if n is odd
And the subsequence
(n+1)
2
n
diverges to , although the subsequence
n+1
n
2
con-
verges to 0.
n
)
n
be a sequence converging to . Let p(x), q(x)
C[x]. Assume that q(a) ,= 0. Then the sequence (p(x
n
)/q(x
n
)
n
converges to
p()/q().
Exercises.
i. Find the following limits and prove your result using only the denition
of convergence.
a. lim
n
2n5
5n+2
b. lim
n
2n5
n+2
c. lim
n
2n5
n
2
+2
ii. Let 1 < r 1. Show that the sequence (r
n
/n)
n
converges to 0.
iii. Show that
a. lim
n
n+3
n
2
+3n+2
= 0.
b. lim
n
_
n+3
n
2
+n5
_
n
= 0.
c. lim
n
_
n+3
n
2
+n5
_
n1
3n+2
= 0.
d. lim
n
_
n
2
1
n
3
n5
_
n
2
1
2n3
= 0.
iv. Show that if a, b > 0 then lim
n
(a
n
+b
n
)
1
n
= max(a, b).
v. Let (a
n
)
n
and (b
n
)
n
be two real sequences that converge to two dierent
numbers in the usual metric. Show that a
n
: n N b
n
: n N is
nite.
vi. Show that the set of convergent sequences of R (in the usual metric) is a
real vector space. (For the addition of sequences and scalar multiplication
of a sequence by a real number, see Example vii, page 45).
6.5. MORE ON SEQUENCES 67
vii. Show that the set of sequences of R (in the usual metric) that converges
0 is a real vector space.
viii. Show that the set of bounded sequences of R (in the usual metric) is a
real vector space.
6.5 More On Sequences
Lemma 6.5.1 A subsequence of a convergent sequence is also convergent; fur-
thermore the limits are equal.
Proof: Let (x
n
)
n
converge to x and let (y
i
)
i
= (x
n
i
)
i
be a subsequence of
(x
n
)
n
. We will show that lim
i
y
i
= x also. Let > 0. Since lim
n
x
n
= x,
there is an N such that for all i > N, d(x
i
, x) < . Since the sequence (n
i
)
i
is
strictly increasing, n
i
i (one can prove this by induction on i). Thus for all
i > N, n
i
> N also and hence d(y
i
, x) = d(x
n
i
, x) < .
Limit in the Topological Language. We can translate the denition of a
limit of a sequence in a language that involves only open subsets (rather than
the metric). This is what we will do now:
Theorem 6.5.2 Let (X, d) be a metric space, (x
n
)
n
a sequence from X and
x X. Then lim
n
x
n
= x if and only if for any open subset U containing
x, the sequence (x
n
)
n
is eventually in U, i.e. there is an N such that x
n
U
for all n > N.
Proof: () Let U be an open subset containing x. By Proposition 5.4.1, there
is an > 0 such that B(x, ) U. Since lim
n
x
n
= x, there is an N such
that for all n > N, d(x
n
, x) < , i.e. x
n
B(x, ) U.
() Let > 0. Then B(x, ) is open. Thus, by hypothesis, there is an N
such that for all n > N, x
n
B(x, ), i.e. d(x
n
, x) < . Hence lim
n
x
n
= x
by denition of limits.
Exercises.
i. Let (a
n
)
n
be a convergent sequence of real numbers.
a. Does the sequence (a
2n
)
n
converge necessarily?
b. Assume a
n
,= 0. Does the sequence (a
n
/a
n+1
)
n
converge necessarily?
ii. Let x
n,m
=
n
n+m
. Find lim
n
lim
m
x
n,m
and lim
m
lim
n
x
n,m
.
iii. Let (x
n
)
n
and (y
n
)
n
be two sequences. Dene
z
n
=
_
x
n/2
if n is even
yn1
2
if n is odd
Show that (z
n
)
n
converges if and only if (x
n
)
n
and (y
n
)
n
both converge
to the same number.
iv. Let (x
n
)
n
be a real sequence and (y
n
)
n
a complex sequence. Let y C.
Assume that lim
n
x
n
= 0 and [y
n
y[ x
n
for all n. Show that
lim
n
y
n
= y.
v. Let (x
n
)
n
be a convergent sequence of real numbers. Let
y
n
=
x
1
+. . . +x
n
n
.
Show that lim
n
y
n
exists and is equal to lim
n
x
n
.
6.6 Cauchy Sequences
The elements of a convergent sequence, since they approach to a xed element,
should also approach to each other. A sequence that has this property is called
a Cauchy sequence. We formalize this as follows: A sequence (x
n
)
n
is called
a Cauchy sequence if for any > 0 there is an N such that d(x
n
, x
m
) < for
all n, m > N. We formalize this by writing lim
n,m
d(x
n
, x
m
) = 0.
We rst prove what we have said in a mathematical way:
Theorem 6.6.1 A convergent sequence is a Cauchy sequence.
Proof: Let (x
n
)
n
be a sequence converging to x. Let > 0. Since lim
n
x
n
=
x, there is an N such that d(x
n
, x) < /2 for all n > N. Now if n, m > N, then
d(x
n
, x
m
) d(x
n
, x) +d(x, x
m
) < /2 +/2 = .
On the other hand, a Cauchy sequence need not be convergent (see Exercises
x, page 59 and xii, page 60). But we have the following:
Proposition 6.6.2 A Cauchy sequence that has a convergent subsequence is
convergent. Furthermore the limits are the same.
Proof: Let (x
n
)
n
be a Cauchy sequence. Let (y
i
)
i
= (x
n
i
)
i
be a convergent
subsequence of (x
n
)
n
. Say lim
i
y
i
= x. We will show that lim
n
x
n
= x.
Let > 0. Since (x
n
)
n
is a Cauchy sequence, there is an N
such that if
n, m > N
, then d(x
n
, x
m
) < /2. Also, since lim
i
y
i
= x, there is an N
1
such that if i > N
1
then d(y
i
, x) < /2. Let N = max(N
, N
1
). Then for all
i > N, we have n
i
i > N (because the sequence (n
i
)
i
is strictly increasing)
and so d(x
i
, x) d(x
i
, y
i
) +d(y
i
, x) = d(x
i
, x
ni
) +d(y
i
, x) < /2 +/2 = .
Proposition 6.6.3 A Cauchy sequence is bounded.
Proof: Let (x
n
)
n
be Cauchy sequence. Take = 1 in the denition. Thus
there exists an N such that for all n, m > N, d(x
n
, x
m
) < 1. In particular, for
all n > N, d(x
N+1
, x
n
) < 1. Let
r = max(1, d(x
0
, x
N+1
) + 1, . . . , d(x
N
, x
N+1
) + 1).
Then x
n
: n N B(x
N+1
, r).
6.7. CONVERGENCE OF REAL CAUCHY SEQUENCES 69
Exercises.
i. Let (a
n
)
n
and (b
n
)
n
be two Cauchy sequences such that the set a
n
: n
N b
n
: n N is innite. Show that lim
n
a
n
= lim
n
b
n
.
ii. Give an example of each of the following, or argue that such a request is
impossible.
a) A Cauchy sequence that is not monotone.
b) A monotone sequence that is not Cauchy.
c) A Cauchy sequence with a divergent subsequence.
d) An unbounded sequence with a Cauchy subsequence.
iii. Let X
1
, . . . , X
m
be metric spaces. Let X = X
1
. . . X
n
be the product
space. Let (x
n
)
n
be a sequence in X. Set x
n
= (x
n1
, . . . , x
nm
). Then
(x
n
)
n
is Cauchy if and only if x
ni
)
n
is Cauchy for all i = 1, . . . , m.
iv. Let x
1
= 1, x
2
= 2 and x
n
= (x
n1
+x
n2
)/2 for n > 2.
a. Show that 1 x
n
2 for all n. (Hint: By induction on n).
b. Show that [x
n
x
n+1
[ = 1/2
n1
for all n. (Hint: By induction on n).
c. Show that if m > n then x
n
x
m
< 1/2
n2
for all n. (Hint: Use part
c).
d. Show that (x
n
)
n
is a Cauchy sequence.
e. Find its limit. Note that the sequence is formed by taking the arithmetic
mean of the previous two terms. It can be guessed that the limit is at the
two thirds of the way from x1 = 1 to x2 = 2, i.e. it is 5/3. This can
be shown by some elementary linear algebra. But there is an easier way:
Note that part b can be sharpened into x
n
x
n+1
= (1)
n
/2
n1
, thus
x
n+1
1 = x
n+1
x
1
= (x
n+1
x
n
) + ... + (x
2
x
1
) = 1 1/2 + 1/4
... +(1)
n
1/2
n1
. It follows that the limit x, that we know it exists from
the previous question, is equal to the innite sum x = 1 +1 1/2 +1/4
. . . +(1)
n
1/2
n1
+. . . Now just remarque that 2x = 2 + 2-1 + 1/2-1/4
+ . . . = 3 + (1/2-1/4 + 1/8 ...) = 3 + (2-x) and so x = 5/3.
v. We say that a sequence (x
n
)
n
is contractive if there is a constant c,
0 < c < 1, such that [x
n+2
x
n+1
c[x
n+1
x
n
[ for all n. Show that
every contractive sequence is convergent. Hint: [x
n+1
x
n
[ c
n
[x
1
x
0
[.
Thus we can estimate [x
n
x
m
[ as is done above and show that such a
sequence is Cauchy.
6.7 Convergence of Real Cauchy Sequences
The purpose of this subsection is to prove the following important theorem:
Theorem 6.7.1 Every real Cauchy sequence is convergent.
Recall that a sequence (x
n
)
n
of real numbers is called nondecreasing (resp.
nonincreasing) if x
n
x
m
(resp. x
n
x
m
) for all n m. A nonincreasing or
nondecreasing sequence is called a monotone sequence.
To prove this theorem, we rst prove that every monotone and bounded
sequence converges.
Theorem 6.7.2 (Monotone Convergence Theorem) A monotone and bounded
sequence of R converges.
Proof: Let (x
n
)
n
be an increasing sequence of real numbers which is bounded
above. Then the set x
n
: n N has a least upper bound . We claim that
lim
x
x
n
= . Let > 0. Since is not an upper bound, there is an N
such that < x
N
. Since the sequence (x
n
)
n
is increasing, < x
n
for
all n N. Since is an upper bound, < x
n
for all n N. Thus
[ x
n
[ for all n N.
The second part follows easily from the rst.
Theorem 6.7.3 Every real sequence has either an increasing or a nonincreas-
ing subsequence.
Proof: We will choose an increasing sequence (n
i
)
i
such that given any i, either
x
nj
> x
ni
for all j > i or either x
nj
x
ni
for all j > i.
Let n
= 0.
Consider the sets
G
0
= n > 0 : x
n
> x
0
and
L
0
= n > 0 : x
n
x
0
.
Either G
0
or L
0
is innite. If G
0
is innite, consider only the subsequence that
consists of x
0
and the x
n
s for which n G
0
, deleting the rest. If G
0
is nite,
consider only the subsequence that consists of x
0
and the x
n
s for which n L
0
,
deleting the rest.
Renaming the elements, assume that (y
n
)
n
is this new sequence. In this new
sequence y
0
has the following property: Either y
n
> y
0
for all n > 0 or y
n
y
0
for all n > 0.
Now consider y
1
and the sets
G
1
= n > 1 : y
n
> y
1
and
L
1
= n > 1 : y
n
y
1
.
Either G
1
or L
1
is innite. If G
1
consists of y
0
, y
1
and the y
n
s for which n G
1
1
is
nite, consider only the subsequence that consists of y
0
, y
1
and the y
n
s for
which n L
1
, deleting the rest.
Renaming the elements, assume that (y
n
)
n
is this new sequence. In this new
sequence y
0
n
> y
0
for all n > 0 or y
n
y
0
6.7. CONVERGENCE OF REAL CAUCHY SEQUENCES 71
for all n > 0 and y
1
n
> y
1
for all n > 1
or y
n
y
1
for all n > 1.
Next we consider y
2
and the sets
G
2
= n > 2 : y
n
> y
2
and
L
2
= n > 2 : y
n
y
2
.
Either G
2
or L
2
is innite. If G
2
consists of y
0
, y
1
, y
2
and the y
n
s for which n G
2
2
is
nite, consider only the subsequence that consists of y
0
, y
1
, y
2
and the y
n
s for
which n L
2
, deleting the rest.
Continuing this way, we nd a subsequence, say (z
n
)
n
, of the original se-
quence (x
n
)
n
such that for all n, either for all m > n, x
m
> x
n
or for all m > n,
x
m
x
n
.
We call n red if the rst case occurs, i.e. if for all m > n, x
m
> x
n
. We call
n green if the second case occurs, i.e. if for all m > n, x
m
x
n
.
Thus we gave a color, either red or green, to each natural number. If there
are innitely many red numbers, then the sequence (z
n
)
n is red
is increasing.
If there are innitely many green numbers, then the sequence (z
n
)
n is green
is
nonincreasing.
Proof of Theorem 6.7.1. Let (x
n
)
n
be a real Cauchy sequence. By Proposi-
tion 6.6.3, (x
n
)
n
is bounded. By Theorem 6.7.3, (x
n
)
n
has a monotone subse-
quence, say (y
n
)
n
. The subsequence (y
n
)
n
is still bounded. By Theorem 6.7.2,
(y
n
)
n
converges. By Proposition 6.6.2, (x
n
)
n
converges.
Theorem 6.7.1 can also be proven as a consequence of the following important
result:
Theorem 6.7.4 (Bolzano-Weierstrass Theorem) Every bounded sequence
in R has a convergent subsequence.
Proof: Let (x
m
)
m
be a bounded sequence in R. Choose a, b R such that
x
m
[a, b] for all m. We will choose a Cauchy subsequence (x
mi
). Let m
0
= 0.
Separate the interval [a, b] into two parts [a,
a+b
2
] and [
a+b
2
, b]. One of these
two smaller intervals contains x
m
for innitely many m. Let m
1
to be the
smallest index not equal to m
0
which is in this smaller interval that contains
x
m
for innitely many m. Now do the same with this smaller interval. Thus
the sequence y
i
= x
mi
has the property that for all i > j, d(y
i
, y
j
) (b a)/2
j
.
By the Theorem of Nested Intervals 3.1.12, the intersection of these intervals
of length (b a)/2
j
is nonempty. If y is in the intersection, then lim
i
y
i
= y.
Note that this theorem is false in general metric space: In a discrete metric
space a sequence of distinct elements does not have a Cauchy subsequence at
all.
Second Proof of Theorem 6.7.1. Let (x
n
)
n
be a Cauchy sequence in R
n
. By
Proposition 6.6.3, (x
n
)
n
is bounded. By Theorem 6.7.4, (x
n
)
n
has a convergent
subsequence. By Proposition 6.6.2, (x
n
)
n
converges.
Exercises.
i. Let x
n
= 1/1
2
+ 1/2
2
+ 1/3
2
+. . . + 1/n
2
.
a. Show that for all integers n 1, x
n
2 1/n < 2.
b. Conclude that the sequence (x
n
)
n
converges.
c. Show that for an integer n large enough, n
2
2
n
.
d. Conclude that the sequence (x
n
)
n
is bounded above by 1+1/4+1/9+
1/8 (= 107/72).
We will see later that this sequence converges to
2
/6.
ii. Show that the sequence

2,
_
2
2,
_
2
_
2
2 converges. Find its limit.

iii. Discuss the convergence of the sequence
a,
_
a
a,
_
a
_
a
a for aR
0
.
iv. Show that if k 2 is an integer, the sequence (x
n
)
n
where x
n
= 1/1
k
+
1/2
k
+ 1/3
k
+. . . + 1/n
k
is Cauchy and hence converges.
v. Let x
0
= x and x
n+1
=
1
4x
n
.
a) Assume that the sequence converges. Show that the limit is either
2 +
3 or 2
3.
b) Show that if 2
3 x
n
2 +
3 then 2
3 x
n+1
2 +
3.
c) Assume x [2
2, 2 +
2]. Show that (x

n
)
n
is decreasing. Conclude
that lim
n
x
n
= 2
2.
d) Discuss the convergence and the divergence of (x
n
)
n
for other values
of x.
vi. Discuss the convergence of the sequence (x
n
)
n
dened by x
0
= x and
x
n+1
= 4 1/x
n
in terms of x.
vii. Let x
1
= x and x
n+1
=
2x
n
.
a) Write the rst four terms of x
n
when x =
2.
b) Show that if the sequence (x
n
)
n
converges, then, it must converge to
either 0 or 2.
c) Discuss the convergence of the sequence (x
n
)
n
according to the values
of x.
viii. Let x
1
= x and x
n+1
= x
n
/2 + 1/x
n
. Discuss the convergence of the
sequence (x
n
)
n
according to the values of x.
6.8. CONVERGENCE OF SOME SEQUENCES 73
ix. (Existence of Square Roots) Let x
1
= 2 and x
n+1
= (x
n
+2/x
n
)/2. Show
that x
2
n
2 for all n. Use this to prove that the sequence (x
n
)
n
is
decreasing. Show that lim
n
x
n
=

2. Modify the sequence so that it
converges to

c.
x. Prove the Bolzano-Weierstrass Theorem for R
n
. Deduce that every Cauchy
sequence of R
n
converges.
6.8 Convergence of Some Sequences
6.8.1 The Sequence ((1 + 1/n)
n
)
n
Proposition 6.8.1 The sequence ((1 +1/n)
n
)
n
converges to some real number
between 2 and 3.
Proof: We will show that the sequence ((1+1/n)
n
)
n
is bounded and increasing.
Theorem 6.7.2 will then prove the theorem. We proceed in several steps.
Claim 1. For all natural numbers n > 0 and all real numbers x > 1,
(1 +x)
n
1 +nx. In particular if y < 1 then (1 y)
n
1 ny
For n = 0, this is clear. (Note that we need x ,= 1 for this). Assume
the inequality holds for n. Then we have (1 + x)
n+1
= (1 + x)
n
(1 + x)
(1 + nx)(1 + x) = 1 + (n + 1)x + nx
2
1 + (n + 1)x. (The second step is by
the induction hypothesis and for this step we also need the fact that 1 +x > 0,
which we know it holds).
Claim 2. 2 (1 + 1/n)
n
3 for all natural numbers n > 0.
Replacing x by 1/n in the rst claim, we get 2 (1 + 1/n)
n
for all natural
numbers n > 0. Now we show that (1 + 1/n)
n
3, for all natural numbers
n > 0. We compute carefully. For n 1 we have,
(1 + 1/n)
n
=

n
i=1
_
n
i
_
1
n
i
= 1 +
n
i=1
n(n1)...(ni+1)
12...i
1
n
i
= 1 +
n
i=1
n
n
n1
n
. . .
ni+1
n
1
12...i
= 1 +
n
i=1
(1
1
n
) . . . (1
i1
n
)
1
12...i
< 1 +
n
i=1
1
12...i

n
i=1
1
2
i1
= 1 +
1
1
2
n
1
1
2
= 1 + 2(1
1
2
n
)
< 3
Claim 3. n N 0, (1 +
1
n
)
n
(1 +
1
n+1
)
n+1
.
Since (1 +
1
n+1
)
n+1
= (1 +
1
n

1
n
+
1
n+1
)
n+1
= (1 +
1
n

1
n(n+1)
)
n+1
, we
have to show that (1 +
1
n

1
n(n+1)
)
n+1
(1 +
1
n
)
n
. By setting a = 1 + 1/n,
this is equivalent to the statement (a
1
n(n+1)
)
n+1
a
n
, i.e. to the statement
(1
1
an(n+1)
)
n+1
1/a, i.e. to the statement (1
1
(n+1)
2
)
n+1
n
n+1
. By Claim
1, if n 1, (1
1
(n+1)
2
)
n+1
1
1
n+1
=
n
n+1
. This proves Claim 3.
Now apply Theorem 6.7.2.
Exercises.
i. We have seen in this paragraph that the sequence given ((1 + 1/n)
n
)
n
converges to a real number > 2. Let e be this limit. Do the following
sequences converge? If so nd their limit.
a) lim
n
_
1 +
1
n+1
_
n
.
b) lim
n
_
1 +
1
2n
_
n
.
c) lim
n
_
1 +
1
n
_
2n
.
d) lim
n
_
1 +
1
2n
_
3n
.
e) lim
n
_
1 +
1
3n
_
2n
.
6.8.2 The Sequences (2
1/n
)
n
and (n
1/n
)
n
Proposition 6.8.2 lim
n
2
1/n
= 1.
Proof: Clearly 2
1/n
> 1. It is also clear that the sequence (2
1/n
)
n
is decreasing.
Therefore it has a limit, say . Then = lim
n
2
1/n
= lim
n
2
1/n
=
1/2
.
Hence = 1.
n
n
1/n
= 1.
Proof: Clearly n
1/n
1. We will show that the sequence is nonincreasing
for n 3, i.e. we will show that (n + 1)
1/(n+1)
n
1/n
, which is equivalent to
(n + 1)
n
n
n+1
and to (1 + 1/n)
n
n. But we have seen in the proof of
Theorem 6.8.1 that (1 + 1/n)
n
3 for all n. Thus (1 + 1/n)
n
n for n 3.
Therefore lim
n
n
1/n
exists, say . We have, 1 = lim
n
(2n)
1/2n
=
lim
n
2
1/2n
n
1/2n
= 1
. (We used Proposition 6.8.2 in the last equality).

Thus = 1.
6.8.3 The Sequence (nx
n
)
n
n
nx
n
= 0 for [x[ < 1. The sequence diverges other-
wise.
Proof: Since lim
n
n
n+1
= 1, there is an N such that [x[ <
n
n+1
for all
n > N. Now for n > N, (n + 1)[x[
n+1
= (n + 1)[x[ x
n
< nx
n
. Hence the
sequence (nx
n
)
n
is decreasing after a while. Therefore the sequence converges,
say to a. Assume a ,= 0. Then by Corollary 6.4.11, 1 = lim
n
nx
n
(n+1)x
n+1
=
lim
n
n
n+1
1
x
= 1/x, a contradiction. Thus lim
n
nx
n
= 0. The rest is left
as an exercise.
6.8. CONVERGENCE OF SOME SEQUENCES 75
Exercises.
i. Let (x
n
)
n
be a sequence such that x
n
,= 0 for every n N and [x
n+1
/x
n
[
r for some xed r (0, 1). Show that lim
n
x
n
= 0.
ii. Let x
1
= 1, x
2
= 2 and
x
n
=
x
n1
+x
n2
2
for n > 2.
a) Show that 1 x
n
2 for all n.
b) Show that [x
n
x
n+1
[ = 1/2
n1
for all n.
c) Show that if m > n then [x
n
x
m
[ < 1/2
n2
for all n.
d) Show that (x
n
)
n
e) Find its limit. Answer: Note that the sequence is formed by taking
the arithmetic mean of the previous two terms. It can be guessed that
the limit is at the two thirds of the way from x
1
= 1 to x
2
= 2, i.e. it
is 5/3. This can be shown by some elementary linear algebra. But there
is an easier way: Note that part b can be sharpened into x
n
x
n+1
=
(1)
n
/2
n1
, thus x
n+1
1 = x
n+1
x
1
= (x
n+1
x
n
) +. . . +(x
2
x
1
) =
11/2+1/4 . . . +(1)
n1
/2
n1
. It follows that the limit x, that we know
it exists from the previous question, is equal to lim
n
21/2+1/4 . . . +
(1)
n1
/2
n1
. Now we just remarque that 2x = 2 lim
n
2 1/2 +
1/4 . . . +(1)
n1
/2
n1
= lim
n
4 1 + 1/2 . . . +(1)
n1
/2
n2
=
lim
n
4 (1 1/2 . . . + (1)
n
/2
n2
) = lim
n
5 (2 1/2 . . . +
(1)
n
/2
n2
) = 5 x, so 3x = 5 and x = 5/3.
iii. Let x
0
= 1, x
n+1
= 1 + 1/x
n
. Show that (x
n
)
n
is convergent.
iv. For k N and x R, discuss the convergence of the sequence (n
k
x
n
)
n
.
v. Let f
0
= f
1
= 1, f
n+2
= f
n
+ f
n+1
. Show that the sequence (
f
n+1
f
n
)
n
is
convergent. (It converges to
2
1+
5
).
vi. We say that a sequence (x
n
)
n
is contractive if there is a constant c
(0, 1) such that [x
n+2
x
n+1
[ c[x
n+1
x
n
[ for all n. Show that every
contractive sequence is convergent. Hint: Note rst that [x
n+1
x
n
[
c
n
[x
1
x
0
[. Now estimate [x
n
x
m
[ and show that such a sequence is
Cauchy.
vii. Let 0 < a
0
= a < b
0
= b. Dene a
n+1
= (2a
n
b
n
)/(a
n
+ b
n
), b
n+1
=
(a
n
+ b
n
)/2. Show that (a
n
)
n
and (b
n
)
n
are both convergent. Show that
their limits are equal. (Hint: Use mean inequalities.)
viii. Let a
0
= a, b
0
= b, c
0
= c. Dene a
n+1
= (b
n
+c
n
)/2, b
n+1
= (c
n
+a
n
)/2,
c
n+1
= (a
n
+ b
n
)/2. Show that (a
n
)
n
, (b
n
)
n
and (c
n
)
n
are convergent.
Compute their limits.(Hint: First take a = b = 0, c = 1.)
6.9 Divergence to Innity
Let (x
n
)
n
be a real sequence. We say that the sequence (x
n
)
n
diverges to
innity if for all real numbers A there is an N such that for all n > N, x
n
> A.
Here, A should be regarded as a very large number, as large as it can be. In
this case one writes lim
n
x
n
= .
Similarly we say that the sequence (x
n
)
n
diverges to minus innity if
for all real numbers A there is an N such that for all n > N, x
n
< A. Here,
A should be thought as a very small negative number. In this case, we write
lim
n
x
n
=
If an increasing sequence (a
n
)
n
is bounded then lim
n
x
n
R (Theorem
6.7.2), if it is unbounded, then lim
n
x
n
= (Exercise vi, page 77). Similarly
for decreasing sequences.
Note that until now, we have not dened innity, and we will never really
dene it. Above, we only dened the phrase lim
n
x
n
= as if it were one
single entity.
Now we let and to be two new distinct symbols. We will never
give any meaning to these symbols. Their only property is that they are new
symbols. We let R := R , , we extend the operations and the order
relation dened on R to R partially as follows:
< r < for all r R
r += +r = for all r (, ]
r + () = r = +r = for all r [, )
() =
r= r = for all r (0, ]
r= r = for all r [, 0)
r() = ()r = for all r (0, ]
r() = ()r = for all r [, 0)
r/= r/() = 0 for all r R
/r = (1/r) for all r R
= for all r (1, ]

r
= 0 for all r (1, 1)

r
= 0 for all r (1, ]

r
= for all r (0, 1)

r
= for all r (1, 0)

Note that the following terms are not dened:
6.9. DIVERGENCE TO INFINITY 77
() +
+ ()
/
/()
()/
()/
/0
/0
0
()
0
0
(1)

n
)
n
and (y
n
)
n
be two sequences in R such that lim
n
x
n

R and lim
n
y
n
R. Let be any of the ve basic arithmetic operations
(addition, substraction, multiplication, division and power). If lim
n
x
n

lim
n
y
n
is dened, then lim
n
x
n
y
n
= lim
n
x
n
lim
n
y
n
.
Exercises.
i. Show that the sequences (n)
n
and (n
2
)
n
diverge to innity.
ii. Show that the sequence ((1)
n
n)
n
does not converge or diverge neither
to or to .
iii. Show that lim
n
(3/2 + 1/n)
n
= .
iv. Find the following limits and prove your result using only the denition:
a) lim
n
2n
2
5
n+2
.
b) lim
n
2n
2
5
n+2
.
c) lim
n
2n
3
5
n
2
n+2
.
d) lim
n
2n
2
1
n
2
n1
n
.
e) lim
n
_
2 +
1
n
_
n
.
v. Show that if lim
n
x
n
= then lim
n
x
n
= .
vi. Show that an increasing and unbounded sequence converges to .
vii. Let p(x) and q(x) ,= 0 be two polynomials of degree d and e respectively
and with leading coecients a and b respectively. Then the sequence
(p(n)/q(n))
n
diverges to innity if and only if e < f and a/b > 0. And
the sequence (p(n)/q(n))
n
diverges to if and only if e < f and a/b < 0.
viii. Show that if lim
n
x
n
= and lim
n
y
n
= a R then lim
n
x
n
+
y
n
= .
ix. Show that if lim
n
x
n
= and lim
n
y
n
= then lim
n
x
n
+y
n
=
.
x. Show that if lim
n
x
n
= then lim
n
x
1
n
= 0.
xi. Show that there are sequences (x
n
)
n
such that lim
n
x
n
= 0 but lim
n
x
1
n
,=
.
xii. Show that if lim
n
x
n
= and lim
n
y
n
> 0 then lim
n
x
n
y
n
= .
xiii. Show that if lim
n
x
n
= and lim
n
y
n
= then lim
n
x
n
y
n
=
.
xiv. Let x
n
= 1 +1/2 +. . . +1/n. Show that lim
n
x
n
= . (See Theorem
7.1.3).
6.10 Limit Superior and Inferior
Let (a
n
)
n
be a sequence of real numbers. We let
lim
n
a
n
= lim
n
supa
n
, a
n+1
, . . . R.
Since the sequence (supa
n
, a
n+1
, . . .)
n
is decreasing, lim
n
a
n
is either a real
number or is . lim
n
a
n
is called the limit superior of the sequence (a
n
)
n
.
We also dene
lim
n
a
n
= lim
n
infa
n
, a
n+1
, . . . R.
Since the sequence infa
n
, a
n+1
, . . . is increasing, lim
n
a
n
is either a real number
or is . lim
n
a
n
is called the limit inferior of the sequence (a
n
)
n
.
We always have lim
n
a
n
lim
n
a
n
.
Lemma 6.10.1 i. lim
n
a
n
,= if and only if the sequence (a
n
)
n
has an upper
bound.
ii. lim
n
a
n
= if and only if the sequence (a
n
)
n
is unbounded by above.
iii. If lim
n
a
n
= r R then for all > 0 there is an N such that for all
n > N, a
n
< r +.
Proof: (i) and (ii) are clear We prove (iii). Let > 0. Let N be such that for
all n > N, [ supa
n
, a
n+1
, . . .r[ < . Thus < supa
N+1
, a
N+2
, . . .r <
and so supa
N+1
, a
N+2
, . . . < r +. Thus a
n
< r + for all n > N.
We leave to the reader the task of stating and proving the analogue of the
above lemma for the limit inferior.
6.11. COMPLETE METRIC SPACES 79
Theorem 6.10.2 Let (a
n
)
n
be a sequence of real numbers. Then lim
n
a
n
exists if and only if lim
n
a
n
and lim
n
a
n
are real numbers and are equal. In this
case, lim
n
a
n
= lim
n
a
n
= lim
n
a
n
.
Proof: This follows from Lemma 6.10.1.iii and its analogue for the limit infe-
rior.
Exercises.
i. Suppose (u
n
)
n
and (v
n
)
n
are positive sequences. Let U = lim
n
u
n
and
V = lim
n
v
n
. Assume that if one of U and V is , then the other is
nonzero. Then L = lim
n
u
n
v
n
= UV .
6.11 Complete Metric Spaces
A metric space is called complete if its Cauchy sequences converge. Thus Q is
not complete. Theorem 6.7.1 says that R is complete with its usual metric. We
can generalize this to the Euclidean spaces.
Theorem 6.11.1 The Euclidean space R
n
is complete.
To prove this we only need to prove the following.
Theorem 6.11.2 Let x
k
= (x
k,1
, . . . , x
k,n
) R
n
with R
n
considered with its
Euclidean metric. Then
i) (x
k
)
k
is a Cauchy sequence if and only if (x
k,i
)
k
is a Cauchy sequence for
all i = 1, . . . , n.
ii) (x
k
)
k
is convergent if and only if (x
k,i
)
k
is convergent for all i = 1, . . . , n.
And in this case
lim
k
x
k
= ( lim
k
x
k,1
, . . . , lim
k
x
k,n
).
Proof: i. Assume (x
k
)
k
is a Cauchy sequence. Let i = 1, . . . , n. Let > 0. Let
N be such that for all k, > N, d(x
k
, x
) < . Then for all k, > N,

[x
k,i
x
,i
[
_
(x
k,1
x
,1
)
2
+. . . (x
k,n
x
,n
)
2
= d(x
k
, x
) < .
Thus (x
k,i
)
k
Conversely, assume that (x
k,i
)
k
is a Cauchy sequence for all i = 1, . . . , m.
Let > 0. Let N
i
be such that for all k, > N
i
, [x
k,i
x
,i
[ < /n. Let
N = max(N
1
, . . . , N
n
). Now for all k, > N,
d(x
k
, x
) =
_
(x
k,1
x
,1
)
2
+. . . (x
k,1
x
,1
)
2
[x
k,1
x
,1
[ +. . . +[x
k,n
x
,n
[ < /n +. . . +/n = .
ii. The proof is similar and we leave it as an exercise.
One can generalize this theorem easily:
Theorem 6.11.3 For i = 1, . . . , n, let (X
i
, d
i
) be a metric space. Consider
X = X
1
. . . X
n
with the product metric. Let x
k
= (x
k,1
, . . . , x
k,n
) X.
Then
i) (x
k
)
k
is a Cauchy sequence if and only if (x
k,i
)
k
is a Cauchy sequence for
all i = 1, . . . , n.
ii) (x
k
)
k
is convergent if and only if (x
k,i
)
k
is convergent for all i = 1, . . . , n.
And in this case
lim
k
x
k
= ( lim
k
x
k,1
, . . . , lim
k
x
k,n
).
Exercises.
i. Show that lim
n
nx
n
= 0 if 0 x < 1.
n
)
n
and (y
n
)
n
be two converging sequences in R
m
.
Then (x
n
+y
n
)
n
and (x
n
y
n
)
n
are converging sequences and
lim
n
(x
n
+y
n
) = lim
n
x
n
+ lim
n
y
n
and
lim
n
(x
n
y
n
) = lim
n
x
n
lim
n
y
n
.
n
)
n
be a converging sequence in R
m
. Then any subse-
quence of (x
n
)
n
is convergent and it converges to the same limit.
Corollary 6.11.6 Product of nitely many complete metric spaces is complete.
Corollary 6.11.7 The metric space of complex numbers C is complete.
Exercises.
i. Let X be any set. Show that the space Seq(X) dened on page 52 is
complete.
6.12 Completion of a Metric Space
In this subsection, given a metric space X, we will nd the smallest complete
metric space containing X.
Let (X, d) be a metric space. Consider the set ((X) of Cauchy sequences of
X. On ((X) dene the following equivalence relation:
(x
n
)
n
(y
n
)
n
lim
n
d(x
n
, y
n
) = 0.
Lemma 6.12.1 The relation is an equivalence relation on the set ((X).
6.12. COMPLETION OF A METRIC SPACE 81
Proof: Trivial.
Set X = ((X)/ . For (x
n
)
n
((X), let (x
n
)
n
X be its class.
We will turn the set X into a metric space. To do this we need the following.
Lemma 6.12.2 i. Let (x
n
)
n
and (y
n
)
n
be two Cauchy sequences in X. Then
lim
n
d(x
n
, y
n
) exists.
ii. Let (x
n
)
n
, (x
n
)
n
, (y
n
)
n
, (y
n
)
n
be four Cauchy sequences in X. If (x
n
)
n

(x
n
)
n
and (y
n
)
n
(y
n
)
n
, then lim
n
d(x
n
, y
n
) = lim
n
d(x
n
, y
n
).
Proof: We rst note the following:
d(x
n
, y
n
) d(x
n
, x
m
) +d(x
m
, y
m
) +d(y
m
, y
n
),
implying
d(x
n
, y
n
) d(x
m
, y
m
) d(x
n
, x
m
) +d(y
m
, y
n
).
Exchanging n and m, we get
[d(x
n
, y
n
) d(x
m
, y
m
)[ d(x
n
, x
m
) +d(y
m
, y
n
).
i. It is enough to show that the real sequence (d(x
n
, y
n
))
n
is Cauchy. Let >
0. Let N be large enough so that for n, m > N, d(x
n
, x
m
) < /2 and d(y
m
, y
n
) <
/2. By the above inequality for n, m > N, [d(x
n
, y
n
) d(x
m
, y
m
)[ > . .
ii. By assumption lim
n
d(x
n
, x
n
) = lim
n
d(y
n
, y
n
) = 0. So the second
part also follows from the inequality proven in the beginning.
The lemma above says that we are allowed to dene a map d from X X
into R
0
by the rule,
d((x
n
)
n
, (y
n
)
n
) = lim
n
d(x
n
, y
n
).
Lemma 6.12.3 (X, d) is a metric space.
Proof: Suppose d((x
n
)
n
, (y
n
)
n
) = 0, i.e. lim
n
d(x
n
, y
n
) = 0. By denition
(x
n
)
n
= (y
n
)
n
.
Clearly d((x
n
)
n
, (y
n
)
n
) = d((y
n
)
n
, (x
n
)
n
).
It remains to prove the triangular inequality:
d((x
n
)
n
, (y
n
)
n
) = lim
n
d(x
n
, y
n
) lim
n
(d(x
n
, z
n
) +d(x
n
, z
n
))
= lim
n
d(x
n
, z
n
) + lim
n
d(x
n
, z
n
)
= d((x
n
)
n
, (z
n
)
n
) +d((z
n
)
n
, (y
n
)
n
)
.
Lemma 6.12.4 The map i that sends x X to the class (x)

n
of the con-
stant sequence (x)
n
is a continuous embedding of X into X. Furthermore
d(i(x), i(y)) = d(x, y).
Proof: If (x)
n
= (y)
n
, then lim
n
d(x, y) = 0, i.e. d(x, y) = 0, hence x = y,
proving that i is one to one.
For the last equality: d(i(x), i(y)) = lim
n
d(x, y) = d(x, y).
Let us now show that i is continuous. Let (x
k
)
k
be a sequence from X
converging to a. Let us show that the class of constant sequences (x
k
)
n
con-
verge to the class of the constant sequence (a)
n
when k goes to innity. By
Lemma 6.1.2, it is enough to show that lim
n
d((x
k
)
n
, (a)
n
) = 0. We have:
lim
k
d((x
k
)
n
, (a)
n
) = lim
k
d(x
k
, a) = 0. This proves it.
Lemma 6.12.5 (X, d) is a complete metric space.
Proof:
From now on we identify X with its image i(X) and we assume that X is a
subset of X.
Lemma 6.12.6 X is dense in X.
Lemma 6.12.7 Let Y be a metric space. Let f : X Y be a continuous
map. Then there is a unique continuous extension f : X Y of f.
Lemma 6.12.8 Uniqueness of X.
6.13 Supplementary Problems
i. One tosses a coin. Each head adds one point, each tail adds two points. Let
p(n) be the probability of reaching the integer n. Show that lim
n
p(n) =
2/3.
ii. Fibonacci Sequence. Let f
0
= 0, f
1
= 1 and dene f
n+2
= f
n+1
+
f
n
. The sequence (f
n
)
n
is called the Fibonacci sequence. Show that
lim
n
f
n
/f
n+1
=
2
1+
5
.
iii. Let (a
n
)
n
be a bounded real sequence with the property that every con-
verging subsequence converges to the same limit. Show that the sequence
converges.
iv. Let (a
n
)
n
be a bounded real sequence. Let
S = x R : x < a
n
: for innitely many n.
Show that there exists a subsequence converging to sup(S). Deduce the
Bolzano-Weierstrass Theorem from this.
Chapter 7
Series
7.1 Denition and Examples
DO IT IN BANACH SPACES
Let (a
i
)
i
be a sequence of complex numbers. (For most of what follows
we can also assume that (a
i
)
i
is a sequence in a normed vector space). Let
s
n
:=
n
i=0
a
i
. Assume that the sequence (s
n
)
n
converges in C. Then we write
i=0
a
i
= lim
n
n
i=0
a
i
and we say that the series

i=0
a
i
converges to lim
n
n
i=0
a
i
or that its
sum is lim
n
n
i=0
a
i
. Else, we say that the series

i=0
a
i
is divergent.
The numbers s
n
:=

n
i=0
a
i
are called the partial sums of the series
i=0
a
i
.
Although we dened a series for complex numbers, the reader is welcome to
consider only real series, i.e. series where a
i
R for all i.
If

i=0
a
i
is a real series and lim
n
n
i=0
a
i
= , we say that the series
i=0
a
i
diverges to and we write

i=0
a
i
= . Similarly for .
Whenever we write

i=0
a
i
= , we will assume that a
i
R even if this is
not explicitly stated.
Examples.
i. If a
i
= 1 for all i, then

i=0
a
i
= lim
n
n
i=0
a
i
= lim
n
n = .
In general, if (a
i
)
i
is a constant sequence of real numbers, then the series
i=0
a
i
converges if and only if a
i
= 0. If the sequence (a
i
)
i
is a real
constant nonzero sequence, then

i=0
a
i
diverges to depending on
the sign of a
i
.
ii. If a
i
= (1)
i
for all i, then

i=0
a
i
does not exist (is not even ).
83
84 CHAPTER 7. SERIES
Theorem 7.1.1 The set of sequence (a
i
)
i
such that

i
a
i
converges form a
vector space. Furthermore we have,
i
(a
i
+b
i
) =
i
a
i
+
b
i
.
The equality persists if

i
a
i
and

i
b
i
are in R , if the operations
in the equality are allowed.
Proof: Easy.
Remarks.
i. Regrouping the terms of a series may change the convergence of the series.
For example consider the series

i=0
a
i
where a
i
= (1)
i
. If we let
b
i
= a
2i
+a
2i+1
, then

n=0
b
i
= 0 because each b
i
= 0. Thus a divergent
sequence may turn into a convergent sequence.
On the other hand regrouping the terms of a convergent series does not
change its convergence and the sums are equal. For example, the terms
of the alternating series

i=1
(1)
i
/i that we will show is convergent can
be regrouped two by two to get that the series

i=1
1/(2i + 1)(2i + 2)
converges to the same limit.
As we have seen we cannot dissociate a series without altering its sum
(unless the terms are positive).
ii. A rearrangement of the terms of the series may change the convergence
of the series (unless the terms are positive, see Theorem 7.3.2). Consider
the series
i
(1)
i
/i = 1 1/2 + 1/3 1/4 + 1/5 . . . .
We will see later that this series converges to a nonzero number. Now
shue the terms as follows:
(11/21/4)+(1/31/61/8)+. . .+(1/(2n+1)1/2(2n+1)1/2(2n+2))+. . .
Regrouping a positive term with the next term which follows it, we get
1/2 1/4 + 1/6 1/8 +. . . + 1/2(2n + 1) 1/2(2n + 2)) +. . .
and this is half of the initial series.
iii. On the other hand, if

i=0
a
i
converges, then if we regroup the terms
without changing the order, then the sum
(a
1
+a
2
+. . . +a
n1
) + (a
n1+1
+. . . +a
n2
) + (a
n2+1
+. . . +a
n3
) +. . . ,
7.1. DEFINITION AND EXAMPLES 85
remains unaltered. This may be called the associativity of series (see
Exercise v, page 86). But we cannot dissociate: For example
(1 + (1)) + (1 + (1)) +. . . = 0,
but
1 + (1) + 1 + (1) +. . .
is not convergent.
iv. The ordering of the terms of the series is important and in some cases
the sum may change. For example, summing rst the terms with even
subscripts a
2i
and then the terms with odd subscripts a
2i+1
and then
adding these two sums may change the sum. We give an example in page
89.
The series
i=0
r
i
is called geometric series. This will be our rst example
of a convergent series.
Theorem 7.1.2 (Geometric Series) Let C. Then series
i=0
i
is con-
vergent if and only if [[ < 1. In this case
i=0
i
=
1
1
.
Proof: If [[ 1, then lim
n
n
,= 0, so that
i=0
i
cannot be convergent.
Assume [[ < 1. Let s
n
=
n
i=0
i
. Then (1 )s
n
= s
n
s
n
= 1
n+1
and
so s
n
=
1
n+1
1
. Therefore by Proposition 6.2.3,

i=1
i
=
1
1
.
The series

i=1
1/i is called harmonic series.
Theorem 7.1.3 (Harmonic Series)

i=1
1/i = .
Proof: For any natural number k and for the 2
k1
many natural numbers
i [2
k1
, 2
k
), we have 1/i > 1/2
k
, so that

2
k
1
i=2
k1 1/i > 2
k1 1
2
k
= 1/2. It
follows that s
2
n
1
=

2
n
1
i=1
1/i =

n
k=1
2
k
1
i=2
k1
1/i >
n
k=1
1/2 = n/2, so
that the sequence s
n
diverges to innity.
Exercises.
i. Show that the series

i=1
1/i
i
converges. Find an upper bound for the
sum.
ii. Show that the series

i=1
1/2i and

i=0
1/(2i + 1) diverge.
iii. Show that the sum of the reciprocals of natural numbers whose decimal
expansion contains at least a zero 1/10 + . . . + 1/90 + 1/100 + 1/101 +
. . . + 1/109 + 1/110 + 1/120 +. . . diverges.
iv. Let k N be 2. Show that the series

i=1
1/n
i
converges. (Hint: See
Exercise iv, page 72).
v. Suppose
i=0
a
i
converges. Let (n
k
)
k
be a strictly increasing sequence of
natural numbers with n
0
= 0. Set b
k
= a
n
k
+ . . . + a
n
k+1
1
. Show that
k=0
b
k
converges to

i=0
a
i
.
7.2 Easy Consequences of the Denition
Proposition 7.2.1 If

i=0
a
i
converges then lim
i
a
i
= 0.
Proof: Let > 0. Since

i=0
a
i
converges, the partial sums s
n
=

n
i=0
a
i
form a Cauchy sequence. Thus there exists an N such that for all n, m > N,
[s
n
s
m
[ < . Taking m = n+1, we see that for all n > N, [a
n+1
[ = [s
n
s
n+1
[ <
. Thus for all n > N + 1, [a
n
[ < . This shows that lim
n
a
n
= 0.
Corollary 7.2.2 If lim
i
a
i
,= 0 then the series

i=0
a
i
cannot converge.
Since the convergence of a series is nothing else but the convergence of the
sequence of partial sums, we can use the completeness of R:
Proposition 7.2.3 (Cauchys Criterion) A sequence

i=0
a
i
converges if
and only if for all > 0, there exists an N such that for all n > m > N,
[
n
i=m
a
i
[ < .
Proof: The condition just expresses the fact that the sequence of partial sums
(
n
i=0
a
i
)
n
is Cauchy.
Theorem 7.2.4 If a
i
0 and if the partial sums

n
i=0
a
i
are bounded then
i=0
a
i
converges.
Proof: Follows from Theorem 6.7.2.
Corollary 7.2.5 (Comparison Test) Let 0 a
i
b
i
.
a. If
i=0
b
i
converges, then
i=0
a
i
converges as well. Furthermore
i=0
a
i

i=0
b
i
.
b. If

i=0
a
i
= then

i=0
b
i
= .
Corollary 7.2.6 Suppose that u
n
, v
n
> 0 and u
n+1
/u
n
v
n+1
/v
n
eventually.
If
n
v
n
converges then
n
u
n
converges. Hence if
n
u
n
diverges then
n
v
n
diverges.
Proof: Let n
0
be a witness for the eventuality. Let n > n
0
. Multiply the
inequalities from n
0
to n 1 to get u
n
/u
n
0
v
n
/v
n
0
, i.e. u
n
rv
n
where
r =
u
n
0
vn
0
. Clearly

n
v
n
converges if and only if

n
rv
n
converges. Now apply
the comparison test (Corollary 7.2.5).
7.3. ABSOLUTE CONVERGENCE 87
Exercises.
i. Cauchy Condensation Test. Let (x
i
)
i
be a nonincreasing, nonnegative
real sequence. Show that

i=0
x
i

i=0
2
i
x
2
i
converges. [A, page 53]
ii. Let a
n
> 0, b
n
> 0. Show that if lim
i
b
i
/a
i
= 0 and
i=0
a
i
converges
then

i=0
b
i
converges as well.
iii. Let a
i
> 0, b
i
> 0. Show that if lim
i
b
i
/a
i
= and
i=0
a
i
= then
i=0
b
i
= .
7.3 Absolute Convergence
If

i=0
[a
i
[ converges, then we say that the series

i=0
a
i
converges abso-
lutely.
Theorem 7.3.1 (Cauchy) An absolutely convergent series is convergent.
Proof: Let

i=0
a
i
be an absolutely convergent series. We will show that
i=0
a
i
converges by using Cauchys Criterion (Proposition 7.2.3). Let > 0.
Let N such that for all n > m > N,

n
i=m
[a
i
[ < . Then for all n > m > N,
[
n
i=m
a
i
[
n
i=m
[a
i
[ < .
But the converse of this theorem is false. Indeed, we will see that the series
i=1
(1)
i
/i converges (Corollary 7.4.2), but we know that it does not converge
absolutely (Theorem 7.1.3).
We will later show that permuting the terms of a series may change its value.
However, this is not the case with the absolutely convergent series.
Theorem 7.3.2 (Rearrangement of the Terms) Let

i=0
a
i
be an abso-
lutely convergent series. Let f : N N be a bijection. Let b
i
= a
f(i)
. Then
i=0
b
i
is also absolutely convergent and its sum is equal to

i=0
a
i
Proof: Let s
n
and t
n
be the partial sums of the series

i=0
a
i
and

i=0
b
i
respectively. Let a =
i=0
a
i
. We will show that lim
n
t
n
= a as well.
Let > 0. Since lim
m
s
m
= a, there is an N
1
such that [s
m
a[ < /2
for all m > N
1
.
Since
i=0
a
i
converges absolutely, there is an N
2
such that for all m > N
2
,
k=i
[a
i
[ < /2.
Choose an m > max(N
1
, N
2
).
Let N be such that 0, . . . , m f(0), . . . , f(N). Now for n > N, we have,
[t
n
a[ [t
n
s
m
[ +[s
m
a[ < [t
n
s
m
[ +/2 = [(b
0
+. . . +b
n
) (a
0
+. . . +
a
m
)[ +/2 = [(a
f(0)
+. . . +a
f(n)
)(a
0
+. . . +a
m
)[ +/2
k=m+1
[a
k
[ +/2 <
/2 +/2 = .
We can also partition an absolutely convergent series in two (or more) pieces:
Theorem 7.3.3 Let us partition the terms (a
i
)
i
of an absolutely convergent
series

i=0
a
i
in two disjoint and innite subsets (b
i
)
i
and (c
i
)
i
. Then

i=0
b
i
and

i=0
c
i
are absolutely convergent series and

i=0
a
i
=
i=0
b
i
+
i=0
c
i
.
Proof: Let
n
,
n
and
n
be the partial sums of the three series in this order.
Given n, we can nd m
n
and p
n
such that every term of the partial sum
n
appears in one of the partial sums
mn
and
pn
. Thus 0
mn
+
pn

n

i=n+1
a
i
, so that lim
n
(
mn
+
pn

n
) = 0.
Corollary 7.3.4 Let

i
a
i
be absolutely convergent. Let (b
i
)
i
be the positive
terms and (c
i
)
i
the negative terms of (a
i
)
i
. Then

i
a
i
=
i
b
i
+
i
c
i
.
Exercises.
i. Suppose

i
a
i
converge absolutely. Show that [
i
a
i
[

i
[a
i
[. (First
Solution: Let R =

i
[a
i
[. Then

n
i
a
i
B(0, R). Since B(0, R) is
closed (why?), the limit

i
a
i
is still in B(0, R). Second Solution:
Let x
n
=

n
i
a
i
. By Exercise v, page 63, [
i
a
i
[ = [ lim
n
x
n
[ =
lim
n
[x
n
[ lim
n
n
i
[a
i
[ =
i
[a
i
[.)
ii. Show that

i=0
a
i
converges absolutely, then

i=0
a
2
i
converges (abso-
lutely) as well. (See also Exercise vi, page 89).
7.4 Alternating Series
A series of the form

i=0
(1)
i
a
i
where a
i
0 is called alternating series.
Theorem 7.4.1 (Alternating Series Test) Let (a
i
)
i
be a decreasing sequence
such that lim
i
a
i
= 0. Then

i=0
(1)
i
a
i
converges. Furthermore,
2n+1
i=0
(1)
i
a
i

i=0
(1)
i
a
i

2n
i=0
(1)
i
a
i
.
for all n.
Proof: Let s
n
= a
0
a
1
+ . . . + (1)
n
a
n
. We consider the sequences (s
2n
)
n
and (s
2n+1
)
n
of even and odd partial sums.
Since s
2n+1
= (a
0
a
1
) + . . . + (a
2n
a
2n+1
), the sequence (s
2n+1
)
n
is
increasing.
Since s
2n
= a
0
(a
1
a
2
) . . . (a
2n1
a
2n
), the sequence (s
2n
)
n
is
decreasing.
Since s
2n+1
s
2n
= a
2n+1
0, s
2n+1
s
2n
. By the Monotone Conver-
gence Theorem (6.7.2), both the series (s
2n
)
n
and (s
2n+1
)
n
converge, say to a
and b respectively. Then a b = lim
n
s
2n
lim
n
s
2n+1
= lim
n
(s
2n
s
2n+1
) = lim
n
a
2n+1
= 0, so that a = b. Hence the sequence (s
n
)
n
converges
to a as well (see Exercise iii, page 67).
Corollary 7.4.2

i=1
(1)
i
/i converges to a number in (0, 1).
7.5. CRITERIA FOR CONVERGENCE 89
Example. [A, page 36] We now give an example (necessarily non-absolutely
convergent, see Theorem 7.3.2) of a convergent series that, when the terms are
permuted, converges to a dierent number. We consider series

i=1
(1)
i
/i.
Let S be its sum. Multiply the series by 1/2 and compute S +S/2:
3S/2 = S +S/2 =
i=1
(1)
i
/i +
i=1
(1)
i
/2i
and it is easy to see that the partial sums of the right hand side are exactly the
partial sums of the sequence
1 +1/3 1/2 +1/5 +1/7 1/4 +1/9 +. . . = 1 +
i=1
_
1
4i 1

1
2i
+
1
4i + 1
_
,
which is a rearrangement of the alternating series
i=1
(1)
i
/i. Since this sum
is nonzero, we see that a dierent rearrangement of the terms of a series may
change the sum.
Exercises.
i. Show that

i=1
(1)
i
/
i converges.
ii. Prove the Alternating Series Test by showing that the sequence of partial
sums is a Cauchy sequence.
iii. Show that

n
1
(2n+1)(2n+3)
converges. (Hint: Note that
1
2n+1

1
2n+2
=
1
(2n+1)(2n+3)
iv. Prove the Alternating Series Test by using the Nested Intervals Property
(Theorem 3.1.12).
v. Find diverging series
i=0
x
i
and
i=0
y
i
such that
i=0
x
i
y
i
converges.
vi. Find a converging series

i=0
x
n
such that

i=0
x
2
i
is divergent.
7.5 Criteria for Convergence
We rst generalize Theorem 7.1.2:
Proposition 7.5.1 Let (x
i
)
i
be a sequence such that for some r (0, 1), [x
i+1
[
r[x
i
[ eventually. Then

i=0
x
i
converges absolutely.
Proof: We may assume that x
i
R
0
and that the condition [x
i+1
[ r[x
i
[
holds for all i. By induction x
i
r
i
x
0
. Thus 0
n
i=0
x
i
x
0
n
i=0
r
i
. Since
the latter sequence converges,

i=0
x
i
converges as well (Corollary 7.2.5).
Corollary 7.5.2 (Ratio Test, dAlembert) Let (x
i
)
i
be a nonzero sequence
in a Banach space such that
lim
n
[x
i+1
/x
i
[
exists and is < 1. Then

i=0
x
i
converges absolutely.
Proof: We may assume that x
i
> 0 for all i. Let s be the limit of the sequence
(x
i+1
/x
i
)
i
. By hypothesis 0 s < 1. Let =
1s
2
. Then > 0 and there is an
N such that for all i > N, [s
x
i+1
xi
[ < . Let r = s +. Then
x
i+1
xi
< r and the
corollary follows from Proposition 7.5.1.
Examples.
i. The series

i=0
z
n
/n! converges for all z in all Banach spaces.
ii. Consider the series
1 +z +
( 1)
2!
z
2
+. . .
( 1) . . . ( n + 1)
n!
z
n
+. . .
where C. The series converges for [z[ < 1 according to dAlembert
(Corollary 7.5.2).
Corollary 7.5.3 (dAlembert) If lim
i
[a
i+1
/a
i
[ < 1, then

i=0
a
i
con-
verges absolutely. If lim
i
[a
i+1
/a
i
[ > 1, then

i=0
a
i
diverges.
Proof: As the proof of Corollary 7.5.2, using Lemma 6.10.1.
Example. Let
a
i
=
_
1/i
2
i is even
2/(i + 1)
2
i is odd
Then lim
i
[a
i+1
/a
i
[ = 2 > 1 and lim
i
[a
i+1
/a
i
[ = 1/2 < 1. But

i=0
a
i
converges.
Corollary 7.5.4 For any z C, the series
exp(z) =

i=0
z
i
/i!
sin(z) =

i=0
(1)
i
z
2i+1
/(2i + 1)!
cos(z) =

i=0
(1)
i
z
2i
/(2i)!
converge absolutely.
These series then give rise to functions from C into C. They are called
exponentiation, sine and cosine respectively.
Theorem 7.5.5 (Root Test, Cauchy) If lim
i
[a
i
[
1/i
< 1, then

i=0
a
i
converges absolutely. If lim
i
[a
i
[
1/i
> 1, then

i=0
a
i
diverges.
7.5. CRITERIA FOR CONVERGENCE 91
Proof: Let r = lim
i
[a
i
[
1/i
< 1. Let =
1r
2
. Note that r + < 1.
By Lemma 6.10.1 there is an N such that for all i > N, [a
i
[
1/i
< r + , i.e.
[a
i
[ < (r +)
i
. Now our theorem follows from theorems 7.2.5 and 7.1.2.
Corollary 7.5.6 If R = 1/lim
i
[a
i
[
1/i
, then

i=0
a
i
z
i
converges absolutely
if [z[ < R and diverges if [z[ > R.
R = 1/lim
i
[a
i
[
1/i
is called the radius of convergence of the series
i=0
a
i
z
i
.
A series of the form
i=0
a
i
z
i
is called a power series. If R is its radius of
convergence, such a series gives rise to a function from the ball of center 0 and
radius R into C.
Theorem 7.5.7 Let f(z) =
i=0
a
i
z
i
be a power series with radius of conver-
gence R. Then

i=1
ia
i
z
i1
has the same radius of convergence.
Theorem 7.5.8 (Abel, form due to Lejeune-Dirichlet) The (complex or
real) series

i=0
v
n
n
converges if
i. The sum [v
m
+. . . +v
n
[ is bounded for all m n.
ii. The series

n=0
[
n
n+1
[ converges.
iii. lim
n
n
= 0.
Proof: We will use Cauchy criterion of convergence (7.2.3). Let > 0. For
m n, let V
m,n
= v
m
+ . . . + v
n
. By hypothesis, there is an A such that
[V
m,n
[ < A for all m n. Let N
1
be such that for all n m > N
1
, [
m

m+1
[ + . . . + [
n1

n
[ < /2A and let N
2
be such that for all n > N
2
,
[
n
[ > /2A. Let N = maxN
1
, N
2
. For n m > N we compute as follows:
Since V
k,k+1
V
k,k
= v
k+1
, we have, v
m
m
+. . .+v
n
n
= V
m,m
m
+(V
m,m+1
V
m,m
)
m+1
+ . . . + (V
m,n
V
m,n1
)
n
= V
m,m
(
m

m+1
) + V
m,m+1
(
m+1

m+2
) +. . . +V
m,n1
(
n1
n
) +V
m,n
n
. Thus v
m
m
+. . . +v
n
n
A([
m
m+1
[ +[
m+1
m+2
[ +. . . +[
n1
n
[ +[
n
[) < A(/2A) +A(/2A) = .
Corollary 7.5.9 The (complex or real) series

i=0
v
n
n
converges if
i. The sum [v
m
+. . . +v
n
[ is bounded for all m n.
ii. (
n
)
n
is positive, nondecreasing and lim
n
n
= 0.
Proof: We check that the above conditions are met. We only have to be
concerned with condition (ii):

n=0
[
n
n+1
[ =
n=0
(
n
n+1
) =
0
.
We can obtain the alternating series test (Theorem 7.4.1) as a consequence:
Corollary 7.5.10 (Alternating Series Test) Let (v
i
)
i
be a decreasing se-
quence such that lim
i
v
i
= 0. Then

i=0
(1)
i
v
i
converges.
Proof: Take v
n
= (1)
n
in the corollary above.
Exercises.
i. Discuss the convergence and absolute convergence of the alternating series
1 1/2
+ 1/3
1/4
+. . . for various values of Q.

ii. Decimal Expansion. Let r R
0
. Then there are k Z and a
i

0, 1, . . . , 9 such that r =
i=k
a
i
10
i
.
iii. Let n > 0 be a natural number. Let r R
0
. Then there are k Z and
a
i
0, 1, . . . , n 1 such that r =
i=k
a
i
n
i
.
iv. Show that 1/3 =
i=1
2
2i
.
v. Let (a
i
)
i
be a positive decreasing sequence, k an integer greater than 1.
Show that the series
i=1
a
i
and
i=1
k
i
a
k
i either both converge or both
diverge.
vi. Show that if the ratio test says that a sequence converges, so does the root
test.
vii. Show that if a
i
> 0 and lim
i
ia
i
exists and is nonzero, then

i=0
a
i
diverges.
viii. Assume a
i
> 0 and lim
i
i
2
a
i
exists. Show that

i=0
a
i
converges.
7.6 Supplementary Problems
i. Show that the sum of the reciprocals of natural numbers whose decimal
expansion does not contain a 0, i.e. the series 1/1 + . . . + 1/9 + 1/11 +
. . . +1/19 +1/21 +. . . +1/29 +1/31 +. . . +1/99 +1/111 +. . . +1/119 +
1/121 +. . . + 1/129 +. . . is convergent.
ii. Let

i=0
a
i
is given. For each i N, let p
i
= a
i
if a
i
0 and assign
p
i
= 0 if a
i
< 0. In a similar manner, let q
i
= a
i
if a
i
0 and q
i
= 0
otherwise.
a. Show that if

i=0
a
i
diverges, then at least one of

i=0
p
i
or

i=0
q
i
diverges (to and resp.)
b. Show that if
i=0
a
i
converges conditionally (i.e. not absolutely), then
i=0
p
i
= and

i=0
q
i
= . [A, Ex. 2.7.3, page 68]
c. Show that if

i=0
a
i
converges conditionally and r R, then there is
a bijection f : N N such that

i=0
a
f(i)
= r. [A, Section 2.9]
iii. Let x
i
= 1/1
2
+ 1/2
2
+ 1/3
2
+. . . + 1/i
2
.
a. Show that for all integers i 1, x
i
2 1/i < 2.
b. Conclude that the series

i=1
1/i
2
converges.
c. Show that for an integer i large enough, i
2
2
i
.
d. Conclude that

i=1
1/i
2
1 + 1/4 + 1/9 + 1/8 = 107/72.
7.6. SUPPLEMENTARY PROBLEMS 93
iv. Show that the series

i=1
1
i
2
+i
converges. Find its value.
v. Show that the series

i=1
1
i
2
+2i
converges. Find its value.
vi. Which of the following series converge? Can you estimate their sum?
i=1
1
i
2
i
i=3
1
i
2
3i+2
i=0
1
3i2
i=0
i
3i
2
2
i=0
i
3i
3
2
i=0
3
i
+4
i
5
i
i=0
2
i
i!
i=0
(1)
i i
i+1
i=0
(1)
i i
i
2
i+1
i=1
(1)
i
i=0
(1)
i 2
i
i
3
i!
7.6.1 Midterm of Math 152
i. Decide the convergence of the series
n
1
_
[n
2
2[
.
ii. Decide the convergence of the series
n
1
n
2
+ 1
.
iii. Decide the convergence of the series
n
1
_
[n
4
6[
.
iv. Suppose that the series

n
a
n
is convergent. Show that lim
n
a
n
= 0.
v. Suppose that (a
n
)
n
is a positive and decreasing sequence and that the
series

n
a
n
n
na
n
= 0.
vi. Find a positive sequence (a
n
)
n
such that the series

n
a
n
is convergent
but that lim
n
na
n
,= 0.
vii. Suppose that series
n
a
n
is absolutely convergent and that the sequence
(b
n
)
n
is Cauchy. Show that the series

n
a
n
b
n
is absolutely convergent.
viii. Let (a
n
)
n
be a sequence. Suppose that
n=1
[a
n
a
n+1
[ converges. Such
a sequence is called of bounded variation. Show that a sequence of
bounded variation converges.
7.6.2 Solutions of the Midterm of Math 152
n
1
_
[n
2
2[
.
Answer: Since for all n > 1,
1
_
[n
2
2[
=
1
n
2
2

1
n
2
=
1
n
and since

n
1
n
diverges, the series

n
1
|n
2
2|
diverges as well.
n
1
n
2
+ 1
.
Answer: Since for n > 0,
1
n
2
+ 1

1
n
2
+n
2
=
1
n
2
and since

n
1
n
diverges, the series

n
1
n
2
+1
diverges as well.
n
1
_
[n
4
6[
.
Answer: Since for n > 1,
1
_
[n
4
6[
=
1
n
4
6

1
_
n
4
n
4
/2
=
2
n
2
and since

n
1
n
2
converges, the series

n
1
n
4
6
converges as well.

n
a
n
n
a
n
= 0.
Proof: Let s
n
= a
0
+. . . +a
n
and s =
n
a
n
. Thus lim
n
s
n
= s. We
have lim
n
a
n
= lim
n
(s
n
s
n1
) = s s = 0.
v. Suppose that (a
n
)
n
series

n
a
n
n
na
n
= 0.
Proof: We know that lim
n
a
n
= 0. Let s
n
= a
0
+ . . . + a
n
and
s =
n
a
n
. Thus lim
n
s
n
= s.
7.6. SUPPLEMENTARY PROBLEMS 95
Since (a
n
)
n
is decreasing,
na
2n
a
n+1
+. . . +a
2n
= s
2n
s
n
.
Thus lim
n
na
2n
= lim
n
(s
2n
s
n
) = ss = 0. Hence lim
n
2na
2n
=
0.
Also 0 < (2n + 1)a
2n+1
(2n + 1)a
2n
= 2na
2n
+ a
2n
. By the above and
the fact that lim
n
a
2n
= 0, the right hand side converges to 0. Hence
by the squeezing lemma lim
n
(2n + 1)a
2n+1
= 0.
From the above two paragraphs it follows that lim
n
na
n
= 0.
(Remark: As we will see later, the series

n=2
1
nln n
diverges. This ex-
ample shows that the conditions that (a
n
)
n
is positive and decreasing and
that lim
n
na
n
= 0 are not enough for the series

n
a
n
to be conver-
gent.)
n
)
n

n
a
n
is convergent
but that lim
n
na
n
,= 0.
Solution: Take a
n
= 1/n
2
if n is not a square and a
n
= 1/n if n is a
square. Then (na
n
)
n
does not converge as lim
n
(n
2
+1)a
n
2
+1
= 0 and
lim
n
n
2
a
n
2 = 1. On the other hand

n
a
n
=

n non square
a
n
+
n a square
a
n
=
n non square
1/n
2
+
n
1/n
2
< 2
n
1/n
2
< 4.
n
a
n
(b
n
)
n

n
a
n
b
n
Proof: Since (b
n
)
n
is Cauchy the sequence ([b
n
[)
n
is bounded. In fact
that is all we need to conclude. Indeed, let B be an upper bound of
the sequence ([b
n
[)
n
. Then [a
n
b
n
[ B[a
n
[ and since

n
[a
n
[ converges,
n
[a
n
b
n
[ converges as well.
viii. Let (a
n
)
n
n=1
[a
n
a
n+1
[ converges. Such
Proof: Since
n=1
[a
n
a
n+1
[ converges,
n=1
(a
n
a
n+1
) converges as
well. Thus the sequence of partial sums whose terms are
n1
i=1
(a
i
a
i+1
) = a
1
a
n
converges, say to a. Thus (a
n
)
n
converges to a
1
a.
Chapter 8
Supplementary Topics
8.1 Liouville Numbers
A real number r is called algebraic (over Q) if there is a nonzero polynomial
p(x) Q[x] (equivalently p(x) Z[x]) such that p(r) = 0. Otherwise r is called
nonalgebraic or transcendental over Q.
Consider the root R of a polynomial f(x) Z[x]. Set f(x) =
n
i=0
a
i
x
i
where a
n
,= 0.
Lemma 8.1.1 If M = max[a
i
/a
n
[ : i = 0, . . . , n 1 then [[ < 1 +M.
Proof: We may assume ,= 0. Since 0 = f() = a
0
+ a
1
i
+ . . . + a
n
n
and
a
n
n
,= 0, we have
1 =
a
n1
n1
+. . . +a
1
+a
0
a
n
n
=
a
n1
a
n
+. . . +
a
0
a
n
n
.
Now if [[ 1 +M, then the absolute value of the right hand side is less than
M(
1
1 +M
+
1
(1 +M)
2
+. . . +
1
(1 +M)
n
) <
M(
1
1 +M
+
1
(1 +M)
2
+. . . +
1
(1 +M)
n
+. . .) =
M
1 +M
1
1
1
1+M
= 1,
a contradiction.
Assume now 0 < , Q is a root of f(x) =

n
i=0
a
i
x
i
Q[x] with a
n
,= 0.
Let M be as above. Let q N be any natural number large enough so that,
1) 1/q < ,
2) f has no root in ( 1/q, + 1/q),
3) + 1/q < 1 +M, and
4) q > n
n
i=1
[a
i
[(1 +M)
i1
.
97
98 CHAPTER 8. SUPPLEMENTARY TOPICS
Note that if q
0
satises these conditions then any q > q
0
also satises these
conditions.
Note also that there are exactly two natural numbers such that [ p/q[ <
1/q: if p is the smallest such natural number then 1/q < p/q < <
(p + 1)/q < + 1/q < 1 +M. Let p N be such that [ p/q[ < 1/q. We will
show that [ p/q[ > 1/q
n+1
.
Clearly f(p/q) = A/q
n
for some A Z. By (2), A ,= 0. We also have,
A/q
n
= f(p/q) = f() f(p/q) = ( p/q)f
()
for some between p/q and . Note that < maxp/q, < 1 + M. Since
f
(x) =
n
i=1
ia
i
x
i1
,
[f
()[ = [
n
i=1
ia
i
i1
[ n
n
i=1
[a
i
[
i1
n
n
i=1
[a
i
[(1 +M)
i1
< q.
Thus, since A ,= 0,
1/q
n
[ A/q
n
[ = [ p/q[[f
()[ < [ p/q[q.

Hence [ p/q[ > 1/q
n+1
.
Thus if is an irrational root of a polynomial of degree n, for large enough
q, if [ p/q[ < 1/q then [ p/q[ > 1/q
n+1
.
It follows that if is irrational and if there are innitely many natural
numbers q for which [ p/q[ < 1/q
n+1
(< 1/q) for some natural number p
(that may depend on q), then is not the root of a polynomial of degree n.
If this happens for innitely many integers n, then is not algebraic. This is
what happens for numbers of the form
=
i=1
a
i
10
i!
where a
i
= 0, 1, . . . , 9, but the values of a
i
are not all 0 after a while.
Because the decimal expansion is not periodic, a , Q. We check the second
condition.
Let n be any natural number and m > n + 1. Let q = q
m
= 10
m!
and
p/q =
m
i=1
ai
10
i!
. Then
0 < p/q =
i=m+1
a
i
10
i!

i=m+1
9
10
i!
=
9
10
(m+1)!
i=0
1
10
i
=
=
10
10
(m+1)!
<
1
(10
m!
)
m
= 1/q
m
.
Thus cannot be an algebraic number. Such a number is called a Liouville
number.
8.1. LIOUVILLE NUMBERS 99
Theorem 8.1.2 (Liouville) If
=
i=1
a
i
10
i!
where a
i
= 0, 1, . . . , 9, but the values of a
i
are not all 0 after a while, then is
not an algebraic number.
100 CHAPTER 8. SUPPLEMENTARY TOPICS
Chapter 9
Convergence of Functions
9.1 Pointwise Convergence of a Sequence of Func-
tions
Let X be a set and (Y, d) be a metric space. Let (f
n
)
n
be a sequence of functions
from X into Y . Assume that for all x X, the sequence (f
n
(x))
n
of the metric
space Y converges. Let f(x) be this limit:
f(x) := lim
n
f
n
(x).
Then the rule x f(x) denes a function f from X to Y . we say that the
function f is the pointwise limit of the sequence (f
n
)
n
. In this case we write
f = lim
n
f
n
or sometimes
f
p
= lim
n
f
n
to precise that the convergence is pointwise (soon we will dene other kinds of
convergence).
Clearly, if f
n
= g for all n, then lim
n
f
n
p
= g.
Exercises and Examples.
i. Let X = R, Y = R with the usual metric and f
n
: R R be dened by
f
n
(x) = x/n. Then the zero function is the pointwise limit of the sequence
(f
n
)
n
.
ii. Let X = Y = (0, 1) and let f
n
(x) = 1/n for all n N0. Unfortunately,
the pointwise limit of the sequence (f
n
)
n
does not exist because 0 , Y .
This is of course only a minor problem. It is enough to change Y with
[0, 1), or even with R, in which case the pointwise limit is the constant
zero function.
101
102 CHAPTER 9. CONVERGENCE OF FUNCTIONS
iii. Let X = Y = [0, 1] and f
n
(x) = x
n
. Then the function f dened by
f(x) =
_
0 if x ,= 1
1 if x = 1
is the limit of the sequence (f
n
)
n
.
iv. Let X = Y = R and let f
n
: R R be dened by f
n
(x) = x
n
/n!. Then
the zero function is the pointwise limit of the sequence (f
n
)
n
.
v. Let X = Y = [0, 1]. Let
f
n
(x) =
_
2(x
k
2
n
) if
k
2
n
x
k+1
2
n
and k 2N
2(
k+1
2
n
x) if
k
2
n
x
k+1
2
n
and k N 2N
Draw the graph of f
n
. What is its length? Show that the zero function is
the pointwise limit of the sequence (f
n
)
n
. What is the length of the limit
function?
vi. Let X = Y = C (or R). dene
f
n
(z) =

n
i=0
z
i
/i!
g
n
(z) =

n
i=0
(1)
i
z
2i+1
/(2i + 1)!
h
n
(z) =

n
i=0
(1)
i
z
2i
/(2i)!
Then
lim
n
f
n
= exp
lim
n
g
n
= sin
lim
n
f
n
= cos .
Theorem 9.1.1 Let Y = R
k
with the usual metric. Let lim
n
f
n
p
= f and
lim
n
g
n
p
= g. Then
i. lim
n
(f
n
+g
n
)
p
= f +g
ii. For all r R, lim
n
rf
n
p
= rf.
iii. If Y = R, then lim
n
(f
n
g
n
)
p
= fg.
iv. If Y = R and f(x) ,= 0 for all x R, then lim
n
1/f
n
p
= 1/f.
Proof: Trivial
Note that the rst two parts of the theorem above says that the set of
pointwise convergent sequences of functions from X into the Euclidean space
R
n
is a real vector space.
9.2. UNIFORM CONVERGENCE OF A SEQUENCE OF FUNCTIONS 103
9.2 Uniform Convergence of a Sequence of Func-
tions
As we dened the convergence of a series of numbers

i=0
a
i
, we can also
dene the convergence of a series of functions

i=0
f
i
as the pointwise limit of
the sequence of functions

n
i=0
f
i
. Here f and each f
i
are functions from a set
X into R
k
(xed k) or even into any Banach space (a complete normed vector
space). Thus, we say that the series

i=0
f
i
converges (pointwise) to f if for
all x X, the series

i=0
f
i
(x) converges to f(x).
Let us rewrite the denition of pointwise convergence: Let X be a set and
(Y, d) be a metric space. Let (f
n
)
n
be a sequence of functions from X into Y
and f a function from X into Y . Assume that f
p
= lim
n
f
n
. This means
that for all x X, f(x) = lim
n
f
n
(x); in other words, for all x X and all
> 0, there is an N such that d(f
n
(x), f(x)) < whenever n > N. Here, N
depends on , but also on x. That is why, one sometimes writes N
,x
instead of
N.
When N is independent of x, we say that the convergence is uniform.
More precisely, the sequence (f
n
)
n
converges uniformly to f if for all > 0
there is an N (that depends only on ) such that for all x X and n > N,
d(f
n
(x), f(x)) < . When that is the case, one writes f
u
= lim
n
f
n
.
Clearly, if f
n
= g for all n, then lim
n
f
n
u
= g.
It is also clear that the uniform convergence is stronger than the pointwise
convergence, in other words, if f
u
= lim
n
f
n
then f
p
= lim
n
f
n
. But the
converse fails as the following example shows.
Examples.
i. Let X = (1, 1) and Y = R with the usual metric. Consider the sequence
(f
n
)
n
of functions from X into Y dened by
f
n
(x) =
1 x
n
1 x
.
It is clear that the pointwise limit of the sequence (f
n
)
n
is the function f
dened by
f(x) =
1
1 x
.
Therefore the uniform limit of this sequence - if it exists at all - should
also be the same function. Assume lim
n
f
n
u
= f. Then for all > 0
there is an N such that
[x[
n
1 x
= [f
n
(x) f(x)[ <
for all n > N and x X. But whatever n is, the function
|x|
n
1x
is un-
bounded on X, so that there cannot be such an N (independent of all
x).
ii. Let (0, 1). Let X = (1, ) and Y = R with the usual metric.
Consider the sequence (f
n
)
n
of functions from X into Y dened by
f
n
(x) =
1 x
n
1 x
(as above). It is clear that the pointwise limit of the sequence (f
n
)
n
is the
function f dened by
f(x) =
1
1 x
.
Therefore the uniform limit of this sequence if it exists at all should
also be the same function. Indeed it is. Let us prove that. Let > 0. Since
0 < < 1, lim
n
n
= 0, so that there is an N such that
n
< (1 )
for all n > N. (Note that N depends only on ). Now, for all n > N and
for all x X, we have,
[f
n
(x) f(x)[ =
[x[
n
1 x
<
[x[
n
1
<

n
1
< .
Thus lim
n
f
n
u
= f.
iii. Let f
n
: [1, 1] R be dened by f
n
(x) = [x[
1+
1
2n+1
. Show that
lim
n
f
n
(x)
u
= [x[.
The second example above shows that the domain X is important for the
uniform convergence. For this reason, one says that the sequence (f
n
)
n
con-
verges uniformly on X.
Theorem 9.2.1 Let X be any set and Y = R
n
with the usual metric. Let
f
n
, g
n
, f, g (n N) be functions from X into Y . Let lim
n
f
n
u
= f and
lim
n
g
n
u
= g. Then
i. lim
n
(f
n
+g
n
)
u
= f +g
ii. For all r R, lim
n
rf
n
u
= rf.
Proof: Trivial.
It is false that if (f
n
)
n
and (g
n
)
n
are sequences functions that converge
uniformly, then the sequence (f
n
g
n
)
n
converges uniformly.
Suppose the sequence (f
n
)
n
of functions from a set into a metric space
converges uniformly (hence pointwise) to f. Let T
n
= sup
x
d(f
n
(x), f(x))
R
0
infty. Take = 1 in the denition to nd an N such that for all n > N
and x, d(f
n
(x), f(x)) < 1, i.e. for all n > N, T
n
< 1. It follows that if the
sequence (f
n
)
n
of functions from a set into a metric space converges uniformly
(hence pointwise) to f, then the sequence T
n
= sup
x
d(f
n
(x), f(x)) R
0
is eventually bounded.
Lemma 9.2.2 Suppose the sequence (f
n
)
n
of functions from a set into a metric
space converges pointwise to f. Let T
n
= sup
x
d(f
n
(x), f(x)) R
0
infty.
Then lim
n
f
n
= f if and only if lim
n
T
n
= 0.
9.2. UNIFORM CONVERGENCE OF A SEQUENCE OF FUNCTIONS 105
Proof: Trivial.
Consider the set of functions T from a set X into a metric space (Y, d). For
f, g T, let d
(f, g) = minsup
xX
d(f(x), g(x)), 1.
Lemma 9.2.3 (T, d
) is a metric space and a sequence (f

n
)
n
from T con-
verges to a function in f T if and only if lim
n
d
(f
n
, g) = 0.
Proof: Easy.
If we consider the set B of bounded functions from X into the metric space
(Y, d), then in the lemma above we can take d
(f, g) = sup
xX
d(f(x), g(x))
as the next lemma will show.
If (X, d) is a normed vector space (V, [ [), then |f| = sup[f(x)[ : x X is
a norm on the vector space of bounded functions from the set X into V , as it
can easily be checked.
Lemma 9.2.4 Suppose the sequence (f
n
)
n
of functions from a set X into a
metric space (Y, d) converges uniformly to f. Assume that the functions f
n
are
eventually bounded. Then f is bounded.
Proof: Take = 1 in the denition of the uniform convergence. Let N be such
that for all n > N and all x, d(f
n
(x), f(x)) < 1. Let n
> N be such that f

n
is bounded. Let b Y and R R be such that f
n
(x) B(b, R) for all x R.
Then for all x R, d(f(x), b) d(f(x), f
n
(x)) +d(f
n
(x), b) < 1 +R.
Let T be the set functions fromX into the metric space (Y, d). Set d
(f, g) =
sup
xX
d(f(x), g(x)), 1. Then d
is a distance on T and a sequence (f

n
)
n
of
functions converge uniformly if and only if the sequence converges in this metric
space.
The following criterion à la Cauchy shows the uniform convergence of a
sequence without explicitly calculating the limit:
Theorem 9.2.5 (Cauchy Criterion for Uniform Convergence) A sequence
(f
n
)
n
of functions from a set X into a complete metric space (Y, d) converges
uniformly if and only if for any > 0 there is an N
such that for all n, m > N

and for all x X, d(f
n
(x), f
m
(x)) < .
Proof: By hypothesis, for all x X, (f
n
(x))
n
is a Cauchy sequence, hence it
has a limit. Let f(x) denote this limit. We will prove that lim
n
f
n
u
= f. Let
> 0.
By hypothesis there is an N, independent of x, such that for all n, m > N,
d(f
n
(x), f
m
(x)) < /2.
Let x X be given. Since the sequence (f
n
(x))
n
converges to f(x), there is
an N
x
(that may depend on x) such that for all m > N
x
, d(f
m
(x), f(x)) < /2.
Let n > N be any natural number. Let x X be any. We will show
that d(f
n
(x), f(x)) < , proving that (f
n
)
n
converges to f uniformly. Let m
x
>
maxN, N
x
be xed. Then d(f
n
(x), f(x)) d(f
n
(x), f
m
x
(x))+d(f
m
x
(x), f(x)) <
/2 +/2 = .
Theorem 9.2.6 Let (f
n
)
n
be a sequence of functions dened on a subset A of
R that converges uniformly to a function f. If each f
n
is continuous at a point
a A, then f is continuous at a.
Proof: Let > 0. We want to nd a > 0 such that for x A and [x a[ <
then [f(x) f(a)[ < . For all n and x A we have
[f(x) f(a)[ [f(x) f
n
(x)[ +[f
n
(x) f
n
(a)[ +[f
n
(a) f(a)[.
We will make each term of the right hand side less than /3, for large enough n
and x close enough to a.
Since lim
n
f
n
=
u f, there is an N such that for n > N and x A,
[f(x) f
n
(x)[ < /3. Thus we also have [f
n
(a) f(a)[ < /3. (Here we use the
uniform continuity).
Pick such an n > N. Since f
n
is continuous at a, there is a > 0 such that
for x A and [x a[ < , f
n
(x) f
n
(a)[ < /3.
Example. The function f
n
(x) = (1[x[)
n
dened on [1, 1] converges point-
wise (but not uniformly) and its limit is not continuous at a = 0.
This example shows also that pointwise convergence on a compact set does
not imply the uniform convergence.
Exercises.
i. Show that the functions f
n
(x) =
x
nx+1
converge uniformly on [0, 1].
ii. Determine the pointwise or the uniform convergence of the following func-
tions:
i. f
n
(x) =
n
k=0
x
k
on X = (1, 1).
ii. f
n
(x) =
n
k=0
x
k
on X = (1/2, 1/2).
iii. f
n
(x) = x
n
on X = (0, 1).
iv. f
n
(x) = x
n
on X = [, ] where (0, 1).
v. f
n
(x) =
x
n
1x
on X = (0, 1).
vi. f
n
(x) =
x
n
1x
on X = (0, 1 ) for any (0, 1).
iii. Let f
n
(x) =
1
1+x
n
a. Find the set A = x R : (f
n
(x))
n
converges.
For x A, let f(x) = lim
n
f
n
(x).
b. What is f?
c. Is the convergence uniform? Justify your answer.
d. Discuss the uniform convergence of (f
n
)
n
in the (open or closed) inter-
vals contained in A.
9.3. UNIFORM CONVERGENCE OF A SERIES OF FUNCTIONS 107
iv. Let f
n
: R R be dened as follows:
f
n
(x) =
_
x
2
if [x[ n
n
2
if [x[ n
Does (f
n
)
n
converge uniformly?
v. Let f
n
: R R be dened as follows:
f
n
(x) =
_
x
2
if [x[ n
n
2
if [x[ n
Does (f
n
)
n
converge uniformly?
vi. Is part iv of Theorem 9.1.1 true for uniform convergence?
vii. Let (f
n
)
n
be a sequence of functions from X into a metric space Y . Let
A, B X. Assume that the sequence (f
n
)
n
converges uniformly on A
and also on B. Show that (f
n
)
n
converges uniformly on A B.
viii. Study the convergence of f
n
(x) =
nx
1+nx
2
on intervals of R.
ix. Study the convergence of f
n
(x) = sin(nx)/n.
x. Study the convergence of f
n
(x) =
x
1+x
n
.
xi. Assume (f
n
)
n
converges uniformly to a function f on a compact set K
R. Let g : K R be a continuous function satisfying g(x) ,= 0. Show
that (f
n
/g)
n
converges uniformly to f/g on K.
xii. Give an example of functions (f
n
)
n
and (g
n
)
n
that converge uniformly,
but whose product sequence (f
n
g
n
)
n
does not converge uniformly.
9.3 Uniform Convergence of a Series of Func-
tions
The pointwise or the uniform convergence of a series

nN
f
n
(x) of functions
is dened in the expected way as the pointwise or the uniform convergence of
the partial sums

n
i=0
f
i
. We dene it more precisely
Let f
n
(n N) and f be functions dened on a subset A of R. The series
of
i=0
f
n
is said to converge uniformly to f if the sequence of partial sums
(s
n
)
n
where s
n
= f
1
+. . . +f
n
converges uniformly to f.
Theorem 9.3.1 Assume each f
i
is continuous and
i=0
f
i
converges uniformly
to f on A. Then f is continuous on A.
Proof: Follows directly from Theorem 9.2.6.
Theorem 9.3.2 (Cauchys Criterion for Uniform Convergence of Series)
The series

i=0
f
n
converges uniformly on A R if and only if for every > 0
there exists an N N such that for all N < m < n, [
n
i=m+1
f
n
[ < .
Proof: Exercise.
Theorem 9.3.3 (Weierstrass M-Test) Let (f
i
)
i
be a sequence of functions
from a set X into C. Suppose that [f
i
(x)[ M
i
for all x X and for all i N.
If the series

i=0
M
i
converges, then the sequence

i=0
f
i
converges uniformly
on X.
Proof: We will use Theorem 9.3.2. Let > 0. Since the series

i=0
M
i
converges, by Cauchy Criterion for series, there is an N N such that for all
N < m < n,

n
i=m+1
M
i
< . Thus for all N < m < n and all x X,
[
n
i=m+1
f
i
(x)[
n
i=m+1
M
i
< .
Theorem 9.3.4 If a power series

i=0
a
n
x
n
converges absolutely at a point
x
0
then it converges uniformly on the closed ball B(0, [x
0
[).
Proof: Follows directly from Weierstrass M-Test (Theorem 9.3.3).
What happens at the end points? Here is the answer:
Theorem 9.3.5 (Abels Theorem) Let the power series
n=0
a
n
x
n
converge
at the point R > 0 (resp. at the point R < 0) . Then the series converges uni-
formly on the interval [0, R] (resp. on the interval [R, 0]).
To prove this theorem, we need the following lemma:
Lemma 9.3.6 (Abels Lemma) Let (b
n
)
n
satisfy b
1
b
2
. . . 0. Let
i=0
a
i
be a series whose partial sums are bounded, by A say. Then for all
n N, [a
1
b
1
+. . . a
n
b
n
[ Ab
1
.
Proof: To be proven.
Proof of Abels Theorem. To be proven.
Corollary 9.3.7 If a power series converges pointwise on the set A R, then
it converges uniformly on any compact set K A.
Proof: By Theorem 9.3.4 and Abels Theorem (9.3.5).
Corollary 9.3.8 A power series is continuous at every point at which it is
convergent.
Exercises.
i. Show that if
i=0
f
i
converges uniformly, then (f
n
)
n
converges uniformly
to the zero function.
ii. Show that the series
i=1
cos(2
i
x)/2
i
converges uniformly. Therefore the
limit function is continuous.
iii. Show that

n=1
x
n
/n
2
is continuous on [1, 1].
9.4. UNIFORM CONVERGENCE AND METRIC 109
9.4 Uniform Convergence and Metric
Let X be a set and (V, [ [) a normed vector space. Consider the normed vector
space B(X, V ) of function from X into Y which are bounded. (See Exercise
iii, page 53). As any normed vector space, B(X, V ) is a metric space (Lemma
5.3.1) and so we may speak about the convergence of a sequence of B(X, V ).
Lemma 9.4.1 Let (f
n
)
n
be a sequence from B(X, V ) and f B(X, V ). If
lim
n
f
n
= f in the metric of B(X, V ), then lim
n
f
n
u
= f.
Conversely if (f
n
)
n
is a sequence from B(X, V ) that converges uniformly to
a function f : X V , then f B(X, V ) and lim
n
f
n
= f in the metric of
B(X, V ).
Proof: We rst show that if (f
n
)
n
is a sequence from B(X, V ) that converges
uniformly to a function f : X V , then f B(X, V )
Let > 0. Since lim
n
f
n
= f in the metric of B(X, V ), there is an N such
that for all n > N, [f
n
(x)f(x)[ < , i.e., by the very denition of the metric on
B(X, V ), sup[f
n
(x) f(x)[ : x X < . This means that [f
n
(x) f(x)[ <
for all x X, proving the uniform convergence.
Let us now show the converse. TO BE COMPLETED
9.5 Limits of Functions
Let (X, d) be a metric space. Let A X be a subset. A point x X is called a
limit point of A if any open subset of X containing x contains an element of
A dierent from x.
Let (X, d) and (Y, d
) be two metric spaces, A a subset of X and f : A Y

a function. Let a X be a limit point of A and b Y . We say that the limit
of f(x) when x goes to a is b if for all > 0 there is a > 0 such that for all
x X, if x ,= a and d(x, a) < , then d
(f(x), b) < . We then write

lim
xa
f(x) = b.
If X = R or C and A is an unbounded subset of X, we would like to be able
to speak about the limit at innity. The denitions in R or in C are dierent.
Let A be a subset of R without upper bound. Let (Y, d
) be a metric space.
Let f : A Y be a function. Let b Y . We say that the limit of f(x) when
x goes to is b if for all > 0 there is a N > 0 such that for all x A, if
x > N then d
(f(x), b) < . We then write

lim
x
f(x) = b.
The limit when x goes to is dened similarly.
Note that when A = N, we obtain the denition of the limit of a sequence
in the metric space Y .
Now let A be an unbounded subset of C. Let (Y, d
) be a metric space. Let

f : A Y be a function. Let b Y . We say that the limit of f(z) when z
goes to is b if for all > 0 there is a N > 0 such that for all z A, if [z[ > N
then d
(f(z), b) < . We then write

lim
z
f(z) = b.
Exercises.
i. Show that lim
x5
(x
2
3x + 5) = 15 by using the denition of limits.
ii. Let X = R, A = R
>0
, Y = R and f : A Y be dened by f(x) = 1/x.
Show that lim
x0
f(x) does not exist.
iii. Let X = A = Y = R with the usual metric. Let f : A Y be dened
as follows:
f(x) =
_
1 if x < 0
1 if x 0
Show that lim
x0
iv. Let A = (0, 1), X = [0, 1], Y = R with the usual metric. Let f : A Y
be dened as follows:
f(x) =
_
0 if x Q
1 if x , Q
Show that lim
x1
Theorem 9.5.1 The limit lim
xa
f(x) of a function f is unique whenever it
exists.
Theorem 9.5.2 The relationship of the limits with the functional operations.
9.6 Convergence of a Family of Functions
If f(x, y) is a function that depends on two parameters, sometimes we may want
to view the parameter x as an index set. In this case we may write f
x
(y) instead
of f(x, y). Suppose f is such a function going from AY into Z where A X
and X and Z are metric spaces. Suppose that x
0
X is a limit point of A.
Let f : Y Z be a function. Then we say that f is the pointwise limit
of the family (f
x
(y))
xA
when x approaches x
0
and we write lim
xx
0
f
x
p
= f
if for all y Y , lim
xx0
f
x
(y) = f(y). This makes sense because for xed
y Y , x f
x
(y) is map from the metric space X into the metric space Z.
More formally this means the following: For all y Y and > 0, there is a
=
,y
> 0 such that if x B(x
0
, ) x
0
then f
x
(y) B(f(y), ). We
emphasize the fact that depends on y. If we can choose independent of y,
then we say that the convergence is uniform.
9.7. SUPPLEMENTARY TOPICS 111
If x varies over R or C, we would like to be able to make x to . This is
easy to do. Assume rst A R has no upper bound. Dene lim
x
f
x
p
= f
if lim
x
f
x
(y) = f(y) for all y Y , i.e. if for all y Y and > 0 there is
an N such that for all x A, if x > N = N
,y
then d
(f
x
(y), f(y)) < . Limit
when x goes to is dened similarly. If A C is an unbounded subset, then
we dene lim
x
f
x
p
= f if lim
x
f
x
(y) = f(y) for all y Y , i.e. if for all
y Y and > 0 there is an N = N
,y
such that for all x A, if [x[ > N then
d
(f
x
(y), f(y)) < .
The convergence is said to be uniform if N can be chosen independently of
y.
Theorem 9.6.1 The uniform limit of continuous functions is continuous.
NOTE: I should think about dening these convergence denitions
once for all, without specic cases.
9.7 Supplementary Topics
Theorem 9.7.1 (Weierstrass) Let (f
i
)
i
be a sequence of functions from an
unbounded subset X of R into C. Suppose that [f
i
(x)[ M
i
for all x X and
for all i N. Assume that the series

i=0
M
i
converges. Then the sequence
i=0
f
i
(x) converges uniformly on X. Assume also that lim
x
f
i
(x) =
i
for
all i. Then

i=0
i
converges absolutely and lim
x
i=0
f
i
(x) =
i=0
i
.
Proof: The rst part is just Weierstrass M-Test. We prove the last part.
Clearly [
i
[ M
i
, so that the absolute convergence occurs. Let > 0. Let n be
such that
i=n+1
M
i
< /4. Choose x
0
such that for x > x
0
and i = 0, 1, . . . , n,
[f
i
(x)
i
[ < /2(n + 1). We have
[
i=0
f
i
(x)
i=0
i
[ = [
i=0
(f
i
(x)
i
)[
i=0
[f
i
(x)
i
[
=

n
i=0
[f
i
(x)
i
[ +
i=n+1
[f
i
(x)
i
[

n
i=0
[f
i
(x)
i
[ + 2
i=n+1
M
i
<
n
i=0
/2(n + 1) + 2/4 = .
This proves the theorem.
Example.
i. Let n N
>0
and z C. We have
(1 +z/n)
n
=
n
i=0
_
n
i
_
(z/n)
i
.
Set
f
1
(n) = 1 +z,
and for i > 1,
f
i
(n) =
_
_
_
_
n
i
_
(z/n)
i
=
n(n1)...(ni+1)
i!
(z/n)
i
if i n
0 if i > n
Since, for i n,
f
i
(n) =
z
i
i!
_
1
1
n
_
. . .
_
1
i 1
n
_
,
we have
lim
n
f
i
(n) = z
i
/i! =: M
i
and [f
i
(n)[ M
i
and

i
M
i
converges by Example i, page 90. Thus the
conditions are realized and so lim
n
(1 + z/n)
n
exists and is equal to
i=0
z
i
/i!.
Chapter 10
Topological Spaces
A set X together with a set of subsets of X is called a topological space if
satises the three conditions of Proposition 5.4.3:
i. , X .
ii. If U
1
, . . . , U
n
then U
1
. . . U
n
.
iii. If U
i
for all i then

i
U
i
.
The elements of are called the open subsets of the topological space X.
Every metric space induces a topological space as stated in Proposition 5.4.3.
But the converse is false as the following example shows: Let X be any set. Let
and X be the only open subsets, i.e. let = , X. Then X becomes a
topological space. If [X[ 2 then there is no metric on X that gives this
topology. The topology on X is called the coarsest topology or the weakest
topology on X.
A topological space whose topology is induced by a metric is called a metris-
able topology.
Sometimes two dierent metrics on the same set give the same topologies,
i.e. the same open sets. In this case we say that the two metrics are equivalent.
Exercises.
i. Show that the following metrics on R
2
are equivalent.
ii. Show that every metric d on a set X is equivalent to the metric d
dened
by d
(x, y) = maxd(x, y), 1).

iii. Show that for equivalent metrics, Cauchy sequences, convergent sequences
and limits do not change.
Discrete Topology. In the topology induced by the discrete metric, every
subset is open. This is called the discrete or the nest topology on X.
113
114 CHAPTER 10. TOPOLOGICAL SPACES
Induced Topology. Let X be a topological space and Y X. By letting
the open subsets of Y to be the traces of open subsets of X on Y , we dene a
topology on Y . Thus an open subset of Y is of the form U Y where U is an
open subset of X. This topology on Y is called the restricted or the induced
topology.
Note that the open subsets of Y and X are dierent. So, we must be careful
when speaking about the open subsets of Y . For this reason, we may say that
a subset of Y is X-open or Y -open.
But if Y is open in X, then subsets of Y which Y -open are exactly X-open
subsets of Y .
If the topology on X is induced by a metric and Y X, then the restricted
topology on Y is induced by the restricted metric on Y .
Order Topology. TO BE DEFINED.
Topology Generated. Let X be a set. Let (
i
)
i
be a family of topologies on
X, i.e. each
i
is the set of open subsets of a topology on X. Then
i
i
denes
a topology on X. An open subset of
i
i
is a subset of X that is open in all
these topologies.
Let be a set of subsets of a set X. Consider the intersection of all the
topologies that contain . By the previous item this is a topology. It is the
smallest topology that contains the subsets of as open sets. We will denote
this topology by ) and call it the topology generated by .
We can describe the topology above more precisely. Consider the subsets
which are arbitrary unions of nite intersections of elements from . These
subsets form the open subsets of a topology and this topology is obviously the
smallest topology that contains all the elements of .
Examples. Let X be a set. The topology generated by consists of
and X. If A X, the topology generated by A is , A, X. If A, B X,
the topology generated by A, B is , A B, A, B, A B, X. The topology
generated by the singleton sets is the set of all subsets of X.
Extended Real Line
Exercises.
i. Show that in a metric space a nite subset is closed.
ii. Let X be any set. Show that the conite subsets of X dene a topology
on X. Show that this topology is not metrisable if X is innite.
iii. Show that a pseudometric on a set denes a topology in a natural way.
(d : X X R
0
is a pseudometric if d(x, x) = 0, d(x, y) = d(y, x) and
d(x, y) d(x, z) +d(z, y).
10.2. CLOSED SUBSETS 115
iv. The topologies dened on R
n
by the metrics d
p
and d
are all equal, i.e.

any open subset with respect to any of them is also open in the other one.
v. Let X be a set and A, B, C be three subsets of X. Show that the topology
generated by A, B, C has at most ??? open subsets.
vi. Let X = R and the set of open and bounded intervals of R. Show that
the topology generated by is the topology induced by the usual metric.
vii. Let X = R and the set of intervals of the form [a, b) where a, b R. Is
the topology generated by metrisable?
viii. Let X = R and the set of intervals of the form [a, b) where a, b Q. Is
the topology generated by equal to the topology above?
ix. Let X be a set. Let d and d
be two distance functions on X. Show that

the two metrics generate the same topology if and only if any sequence of
X that converges for one of the metrics, converges to the same element
for the other metric.
10.2 Closed Subsets
The complement of an open subset of a topological space is called a closed
subset. The properties of open subsets reect upon closed subsets naturally:
Proposition 10.2.1 Let X be a topological space. Then
i. and X are closed subsets.
ii. A nite union of closed subsets is a closed subset.
iii. An arbitrary intersection of closed subsets is closed.
10.3 Interior
Let X be a topological space. Let A X. Then the union A
of all the open

subsets of X which are in A is also an open subset of X which is in A. Thus
A
is the largest open subset of X which is in A, in other words it contains all

the open subsets (in X) of A. The set A
is called the interior of A.

Lemma 10.3.1 Let A, B X. Then the following hold:
i. A is open if and only if A
= A. In particular A
= A
.
ii. If A B then A
.
iii. A
= (A B)
.
iv. A
(A B)
.
Proof: i. Clearly, if A is open then A is the largest open subset of itself, so
A = A
.
ii. Since A
A B, A
is an open subset of B. Thus A
.
10.4 Closure
Let X be a topological space. Let A X. Then the intersection A of all the
closed subsets of X that contain A is also an closed subset of X that contains
A. Thus A is the smallest closed subset of X which contains A. The set A is
called the closure of A.
Lemma 10.4.1 Let A, B X. Then the following hold:
i. A is closed if and only if A = A. In particular A = A.
ii. If A B then A B.
iii. A B A B.
iv. A B = A B.
Lemma 10.4.2 x A if and only if every open subset containing x intersects
A nontrivially.
Exercises.
i. Find an example where the inclusion of Lemma 10.3.1.iii is strict.
ii. Find an example where the inclusion of Lemma 10.4.1.iii is strict.
iii. Show that (A)
c
= (A
c
)
and that (A
)
c
= A
c
.
iv. Find the interior and the closure of an arbitrary subset in the nest topol-
ogy.
v. Find the interior and the closure of an arbitrary subset in the coarsest
topology.
vi. Find the closure and the interior of an arbitrary interval in the usual
topology of R.
10.5 Base of a Topology
Let X be a topological space. Let B be a set of open subsets of X such that
every open subset of X is a union of open sets from B. Then B is called a base
of the topology. Obviously, the set of open subsets of X form a base of the
topological space X.
Examples.
1. Assume X is a metric space. Then the open balls of X form a base of the
topology.
2. Assume X = R
n
with the Euclidean topology. Then the balls of the form
B(a, q) such that a Q
n
and q Q form a base of the topology. Note that this
base is countable.
3. Let be a set of subsets of a set X. The topology ) generated by
has a base consisting of sets of the form U
1
. . . U
n
where n N and U
i
.
10.6. COMPACT SUBSETS 117
Let X be a topological space and x X. Let B
x
be a set of open subsets
of X that contain x such that every open subset of X that contains x contains
one of the of open sets from B
x
. Then B
x
is called a base of the topology at
the point x.
Assume X is a metric space and x X. Then the open balls of the form
B(x, q) where q Q form a base of the topology at x. Note that this base at x
is countable.
10.6 Compact Subsets
A topological space X is called compact if any open covering of X has a nite
subcovering. A subspace A of X is compact if A is compact with the induced
topology.
If X has nitely many open sets then any subset of X is compact.
A subset A of a topological space is called compact if A is compact as a
topological space, i.e. if any family of open subsets of X that covers A has a
nite subcover that covers A.
Singleton subsets of a topological set are certainly compact. Clearly the
union of nitely many compact subsets of a topological space is compact.
Lemma 10.6.1 Let K be a closed subset of a compact topological space X.
Then K is compact.
Proof: Let (U
i
)
iI
be an open covering of K. Then (U
i
)
iI
K
c
is an open
covering of X. Hence a nite number of (U
i
)
iI
K
c
cover X. Hence a nite
number of (U
i
)
iI
cover K.
Proposition 10.6.2 A compact subset of a metric space is closed and bounded.
Proof: Let K be a compact subset of the metric space (X, d). Let a X be
any subset. Then the concentric open balls (B(a, n))
nN
cover K. Therefore a
nite number of them, hence only one of them, covers K. This shows that K is
bounded.
We now show that K is closed, by showing that its complement K
c
is open.
Let a K
c
. The the concentric open subsets (B(a, 1/n)
c
)
nN
cover K. Hence
only one of them cover K. If K B(a, 1/n)
c
, then B(a, 1/n) K
c
. Thus K
c
is open and K is closed.
Exercises.
i. The converse of the above proposition is false. Find a counterexample.
Theorem 10.6.3 (Heine-Borel Theorem) A closed and bounded subset of
R
n
(with the usual metric) is compact.
Proof: Assume rst that n = 1. Let K be a closed and bounded subset of R.
By Lemma 10.6.1, we may assume that K = [a, b] is a closed interval. Assume
[a, b] is not compact. Let (U
i
)
iI
be an open covering of [a, b] without a nite
subcover. We want to choose sequences (x
n
)
nN
and (y
n
)
nN
so that for all n,
i) x
n
x
n+1
< y
n+1
y
n
,
ii) 0 < y
n
x
n
< (b a)/2
n
,
iii) No nite subcover of (U
i
)
iI
covers [x
n
, y
n
].
Let x
0
= a and y
0
= b. Assume x
0
, x
1
, . . . , x
n
and y
0
, y
1
, . . . , y
n
are chosen
so that (i, ii, iii) hold, except of course (i) for n (since x
n+1
is not dened
yet). We choose x
n+1
and y
n+1
as follows. One of the two intervals [x
n
,
x
n
+y
n
2
]
and [
x
n
+y
n
2
, y
n
] is not covered by a nite subcover of (U
i
)
iI
. If [x
n
,
x
n
+y
n
2
] is
not covered by a nite subcover of (U
i
)
iI
, let x
n+1
= x
n
and y
n+1
=
x
n
+y
n
2
.
Otherwise, [
x
n
+y
n
2
, y
n
] is not covered by a nite subcover of (U
i
)
iI
and in this
case we let x
n+1
=
x
n
+y
n
2
and y
n+1
= y
n
. Clearly (i, ii, iii) hold.
By Theorem 3.1.12 (see also Exercise x, page 58),
nN
[x
n
, y
n
] is a singleton
set, say c. Since c [a, b], there is an i I such that c U
i
. Let > 0 be
such that [c , c + ] U
i
. By (ii) there is an n N such that y
n
x
n
< .
Then [x
n
, y
n
] [c , c + ] U
i
as one can show easily. But this contradicts
(iii), because only one of the U
i
s suces to cover [x
n
, y
n
].
The proof for n > 1 is similar, instead of the intervals [x
n
, y
n
] be nd closed
cubes (C
n
)
n
by dividing the previous cube C
n
to 2
n
equal parts and choosing
C
n+1
the one which is not covered by a nite subcover of (U
i
)
iI
.
Theorem 10.6.4 If (K
n
)
nN
is a descending sequence of nonempty compact
subsets of R
n
, then
n
K
n
,= .
Proof: Assume otherwise. Then
n
K
c
n
= R, so that (K
c
n
)
nN
is an open
covering of K
0
. Thus a nite subsequence of (K
c
n
)
nN
covers K
0
, hence K
0

K
c
n
. Since K
n
K
0
, this implies that K
n
= , a contradiction.
Exercises.
i. Let (K
i
)
i
be a family of nonempty compact subsets of a topological space.
Show that if a nite intersection of the K
i
s is nonempty, then
i
K
i
,= .
ii. X be a complete metric space. Let p be an element not in X. Show that
the metric of X cannot be extended to a metric on X p without p
being isolated.
Conclude that if f : R R
>0
is a function satisfying [f(x) f(y)[
[x y[ f(x) + f(y) for all x, y R then f is bounded away from 0,
i.e. there is an > 0 such that f(x) > for all x R. (Hint: Set
d(x, p) = f(x) for x R).
iii. Let F R
n
be a nonempty closed subset. Let A R
n
. Show that there
is a B F such that d(A, B) = infd(A, P) : P F. (Proof: Let
d = infd(A, P) : P F. Let > 0. Then B(A, d+) F is a closed and
10.7. CONVERGENCE AND LIMIT POINTS 119
bounded subset of R
n
, hence it is compact. The distance function f from
the compact subset B(A, d + ) F into R dened by f(P) = d(A, P) is
continuous, so it attains its inmum d.)
10.7 Convergence and Limit Points
Exercises. For each of the topological spaces (X, ), describe the convergent
sequences and discuss the uniqueness of their limits.
i. = (X). ((X) is the set of all subsets of X, 2 pts.).
Answer: Only the eventually constant sequences converging to that con-
stant.
ii. = , X (2 pts.).
Answer: All sequences converge to all elements.
iii. a X is a xed element and is the set of all subsets of X that do not
contain a, together with X of course. (5 pts.)
Answer: First of all, all sequences converge to a. Second: If a sequence
converges to b ,= a, then the sequence must be eventually the constant b.
iv. a X is a xed element and is the set of all subsets of X that contain
a, together with of course. (5 pts.)
Answer: Only the eventually constant sequences converge to a. A se-
quence converge to b ,= a if and only if the sequence eventually takes only
the two values a and b.
v. is the set of all conite subsets of X, together with the of course. (6
pts.)
Answer: All the sequences without more than one innitely repeating
terms converge to all elements. Eventually constant sequences converge
to the constant. There are no others.
vi. Let be the topology on R generated by [a, b) : a, b R. Compare this
topology with the Euclidean topology. (3 pts.) Is this topology generated
by a metric? (20 pts.)
Answer: Any open subset of the Euclidean topology is open in this topol-
ogy because (a, b) =
n=1
[a + 1/n, b). But of course [0, 1) is not open in
the usual topology.
Assume a metric generates the topology. Note that [0, ) is open as it
is the union of open sets of the form [0, n) for n N. Thus the sequence
(1/n)
n
cannot converge to 0. In fact for any b R, no sequence can
converge to b from the left. Thus for any b R there is an
b
> 0 such
that B(b,
b
) [b, ). Let b
0
be any point of R. Let
0
> 0 be such that
B(b
0
,
0
) [b
0
, ). Since b
0
is not open, there is b
1
B(b
0
,
0
) b
0
.
Let 0 <
1
<
0
/2 be such that B(b
1
,
1
) [b
1
, ) B(b
0
,
0
). Inductively
we can nd (b
n
)
n
and (
n
)
n
such that B(b
n
,
n
) [b
n
, )B(b
n1
,
n1
)
b
n1
and
n
<
0
/2
n
. Then (b
n
)
n
is a strictly increasing convergent
sequence, a contradiction.
10.8 Connected Sets
A topological space is called connected if it is not the union of two nonempty
disjoint subsets. TConclude hus a topological space X is connected if whenever
U and V are disjoint open subsets of X whose union is X, one of the open
subsets must be the emptyset. Otherwise X is called disconnected. Note that
if X = U V with U V = , then U and V are also closed. Note also that a
singleton set is connected.
Lemma 10.8.1 The union of two connected subspace with a nonempty inter-
section is connected. The union of a chain of connected subsets is connected.
Lemma 10.8.2 An open and connected subset of a topological space is con-
tained in a maximal open and connected subset, which is necessarily closed.
Theorem 10.8.3 A subset X of R is connected if and only if X is an interval
(all possible kinds).
Dene totaly disconnected. Show Q is totally disconnected. Show Z and
Z
p
are totally disconnected with the p-adic metric.
Chapter 11
Continuity
11.1 Continuity on Metric Spaces
) be two metric spaces, f : X Y a function and a X.

We say that f is continuous at a if for any > 0 there is a > 0 such that
for all x X if d(x, a) < then d
(f(x), f(a)) < . The last condition says

that if x B(a, ) then f(x) B(f(a), ), i.e. that f(B(a, )) B(f(a), ), i.e.
that B(a, ) f
1
(B(f(a), )). Thus f : X Y is continuous at a X if
and only if a is in the interior of f
1
(B(f(a), )) for any > 0. It follows that
f : X Y is continuous at a X if and only if a is in the interior of f
1
(V )
for any open subset V of Y containing f(a). We note this for future reference:
Lemma 11.1.1 Let (X, d) and (Y, d
) be two metric spaces, f : X Y a

function and a X. Then f : X Y is continuous at a X if and only if a
is in the interior of f
1
(V ) for any open subset V of Y containing f(a).
Note that if a X is an isolated point of X, i.e. if B(a, ) = a for some
> 0 then f is always continuous at a. A point of X which is not an isolated
point is called a limit point of X.

function and a X a limit point of X. Then f is continuous at a if and only
if lim
xa
f(x) = f(a).
Proof: This is immediate.

function and a X. Then f is continuous at a if and only if for any sequence
(x
n
)
n
of X converging to a, (f(x
n
))
n
is a sequence of Y converging to f(y).
Proof: Suppose f is continuous at a. Let (x
n
)
n
be a sequence of X converging
to a. Let > 0. Let > 0 be such that f(B(a, )) B(f(a), ). Let N
be such that for all n > N, d(x
n
, a) < , i.e. x
n
B(a, ). Hence f(x
n
)
121
122 CHAPTER 11. CONTINUITY
f(B(a, )) B(f(a), ), i.e. d
(f(x
n
), f(a)) < . This shows that the series
(f(x
n
))
n
converges to f(a).
Conversely, suppose that for any sequence (x
n
)
n
of X converging to a,
(f(x
n
)
n
is a sequence of Y converging to f(y). Assume that f is not con-
tinuous at . Let > 0 be such that for every > 0, there is an x
B(a, )
for which f(x
) , B(f(a), ). Choose = 1/n for n N

>0
and let y
n
= x
1/n
.
Then (y
n
)
n
converges to a because d(y
n
, a) = 1/n. But d(f(y
n
), f(a)) > and
so the sequence (f(y
n
))
n
does not converge to f(a). This is a contradiction.
) be two metric spaces and f : X Y a function. We

say that f is continuous if f is continuous at every point of X.
) be two metric spaces and f : X Y a

function. The following conditions are equivalent:
i. f is continuous
ii. The preimage of every open subset of Y is an open subset of X.
iii. The preimage of any open ball in Y is an open subset of X.
iv. The preimage of every closed subset of Y is a closed subset of X.
v. For any convergent sequence (x
n
)
n
of X, f(lim
n
x
n
) = lim
n
f(x
n
).
Proof: (i ii) Suppose f is continuous. It is enough to show that the preimage
of an open ball of Y is an open subset of X. Let b Y and r > 0. Let
a f
1
(B(b, r)). Then f(a) B(b, r). Let = r d(f(a), b) > 0. Then
B(f(a), ) B(b, r). Since f is continuous at a, there is a > 0 such that
B(a, ) f
1
(B(f(a), )) f
1
(B(b, r)). It follows that f
1
(B(b, r)) is open.
(ii i) Suppose now that the preimage of every open subset of Y is an open
subset of X. Let a X. Let > 0. Since f
1
(B(f(a), )) is open and a is in
this set, there is a > 0 such that B(a, ) f
1
(B(f(a), )). It follows that f
is continuous at a.
The rest of the equivalences are now immediate.
Note that in the lemmas 11.1.1 and 11.1.4 above the existence of a metric on
X and Y disappeared, only topological concepts are left. This will be the basis
for extending the concept of continuity of a function between metric spaces to
the concept of continuity of a function between topological spaces. We will do
this in the next subsection.
Corollary 11.1.5 Continuity of functions between metric spaces depends on
the topologies generated by the metrics rather than on the metrics themselves.
In other words continuity of functions is preserved under equivalent metrics.
Lemma 11.1.6 Let (X, d) be a metric space. Then the map d : X X R
is continuous.
Proof: It is enough to show that the inverse image of any open bounded interval
(r, s) is open. Let us take the sup distance on XX. Let (a, b) XX be such
that d(a, b) (r, s). Let x B(a, s) B(a, r) and y B(b, s) B(b, r) Then r <
sup(d(a, x), d(b, y)) = d((a, b), (x, y)) = sup(d(a, x), d(b, y)) < s. Since B(a, s)
B(a, r) and B(b, s) B(b, r) are open in X, this proves that d is continuous.
11.2. CONTINUITY ON TOPOLOGICAL SPACES 123
Exercises.
i. Let f : [a, b] [a, b] be a function with the following property: There
exists a real number c (0, 1) such that for any x, y [a, b] one has
[f(x) f(y)[ < c[x y[.
a) Show that f is continuous, and
b) Show that there exists an x [a, b] such that f(x) = x.
ii. Assume that X is a complete metric space which is also connected. Show
that for any a X and any r 0 there is a b X such that d(a, b) =
r. Conclude that if x has more than one point, show that X must be
uncountable.
11.2 Continuity on Topological Spaces
Let X and Y be two topological spaces. A function f : X Y is called
continuous if f
1
(V ) is open for all open subsets V of Y . The function f is
said to be continuous at a if a is in the interior of f
1
(V ) for any open subset V
of Y containing f(a). By lemmas 11.1.1 and 11.1.4, this generalizes continuity
in metric spaces.
Lemma 11.2.1 Let X and Y be two topological spaces. A function f : X Y
is continuous if and only if f is continuous at a for all a X.
Examples.
i. Let X be the discrete topological space (i.e. every subset is open). Then
any function from X into any topological space is continuous.
ii. Let X be any topological space. The identity function Id
X
from X into
X is continuous.
iii. One should note that the identity map Id
X
: X X from any topo-
logical space into itself is continuous if one considers the domain X and
the arrival set X with the same topologies (i.e. the same open subsets).
Otherwise this may be false. For example if X
1
denotes the topological
space on the set X where only X and are open and X
1
denotes the
discrete topological space on X, then the identity map Id
X
: X
1
X
2
is not continuous unless [X[ = 1.
iv. Let X be a set. Let X
1
and X
2
be two topologies on the set X. Then the
map Id
X
: X
1
:X
2
is continuous if and only if any subset of X which
is open for the topological space X
2
is open in the topological space X
1
.
In this case we say that X
1
is a renement of X
2
.
v. Let X and Y be any two topological spaces. Then any constant map from
X into Y is continuous.
Proposition 11.2.2 i. Composition of continuous functions is continuous.
ii. A function is continuous if and only if the inverse image of a closed set
is closed.
iii. Let X and Y be topological spaces. Let B be a base of Y . Then a function
f : X Y is continuous if and only if the inverse image of a set in B is open
in X.
iv. A function f : X Y is continuous if and only if the inverse image of
a closed subset of Y is closed in X.
Let X be a set, X
i
be topological spaces and f
i
: X X
i
be functions.
Then there is a smallest/weakest topology on X that makes all the functions f
i
continuous. This topology is generated by the set
f
1
i
(U) : i I and U X
i
is open.
Exercises.
i. By using the denition of continuity, show that the function f(x) =
x
x1
is continuous in its domain of denition.
ii. Let X be a topological space. Given two continuous numerical functions
f and g on X, show that maxf(x), g(x) is also continuous.
iii. Show that if + : R R R and : R R R are continuous.
iv. Show that if f, g : X R
n
are continuous, then so is f +g.
v. Show that if f, g : X C are continuous, then so are f +g and fg.
vi. Show that any polynomial map from C into C is continuous.
vii. Show that the map x 1/x from (0, 1) into R is continuous.
viii. Show that the map x
(x1)
2
x(x2)
is continuous from (0, 2) into R.
ix. Show that the map x
1
1+x
2
is a homeomorphism from (0, ) onto (0, 1).
x. Show that there is no homeomorphism from (0, 1) onto [0, 1).
xi. Show that the map that sends rational numbers to 0 and irrational num-
bers to 1 is not continuous.
xii. Let f : R R be a continuous function. Let Z = r R : f(r) = 0.
Show that 1/f : R Z R is continuous.
xiii. Let X be a topological space and Y X. Show that the smallest topology
on Y that makes the inclusion i : Y X is the induced topology.
11.2. CONTINUITY ON TOPOLOGICAL SPACES 125
xiv. Let X, Y be topological spaces, f : X Y a continuous function, A X
and f(A) B Y . Show that f
|A
: A B is continuous.
xv. Let X, Y be topological spaces, f : X Y a function and f(X) B
Y . Show that f : X Y is continuous if and only if f : X B is
continuous.
xvi. Consider the spaces R
n
and R
m
with their usual topology. Let X R
n
and Y R
m
. Let f : X Y be a function. Show that f is continuous
if and only if each f
i
=
i
f : X R is continuous.
xvii. Let f : X Y be continuous and B Y . Show that f
1
(B
) =
f
1
(B)
and f
1
(B) = f
1
(B).
xviii. II. Subgroup Topology on Z. Let = nZ + m : n, m Z, n ,=
0 . We know that (Z, ) is a topological space.
a. Let a Z. Is Z a open in ?
Answer: Yes.
n=0,1
nZ (3Z + 2) = Z 1.Translating this set by
a 1, we see that Z a is open.
b. Find an innite non open subset of Z.
Answer: The set of primes is not an open subset. Because otherwise, for
some a ,= 0 and b Z, the elements of aZ + b would all be primes. So b,
ab +b and 2ab +b would be primes, a contradiction.
c. Let a, b Z. Is the map f
a,b
: Z Z dened by f
a,b
(z) = az + b
continuous? (Prove or disprove).
Answer: Translation by b is easily shown to be continuous. Let us con-
sider the map f(z) = az. If a = 0, 1, 1 then clearly f is continuous.
Assume a ,= 0, 1 and that f is continuous. We may assume that a > 1
(why?) Choose a b which is not divisible by a. Then f
1
(bZ) is open,
hence contains a subset of the form cZ + d. Therefore a(cZ + d) bZ.
Therefore ac = b and so a divides b, a contradiction. Hence f is not
continuous unless a = 0, 1.
d. Is the map f
a,b
: Z Z dened by f(z) = z
2
continuous? (Prove or
disprove).
e. Is the topological space (Z, ) compact? (Prove or disprove).
Answer: First Proof: Note rst the complement of open subsets of
the form aZ + b are also open as they are unions of the form aZ + c
for c = 0, 1, . . . , a 1 and c , b mod a. Now consider sets of the form
U
p
= pZ + (p 1)/2 for p an odd prime. Then
p
U
p
= because if
a
p
U
p
then for some x Z 1, 2a + 1 = px + p, so that a is
divisible by all primes p and a = 0. But if a = 0 then (p1)/2 is divisible
by p, a contradiction. On the other hand no nite intersection of the U
p
s
can be emptyset as (aZ + b) (cZ + b) ,= if a and b are prime to each
other (why?) Hence (U
c
p
)
p
is an open cover of Z that does not have a nite
cover. Therefore Z is not compact.
First Proof: Let p be a prime and a = a
0
+a
1
p +a
2
p
2
+. . . be a p-adic
integer which is not in Z. Let
b
n
= a
0
+a
1
p +a
2
p
2
+. . . +a
n1
p
n1
.
Then
n
p
n
Z +b
n
= but no nite intersection is empty. We conclude as
above.
xix. Let f : R R be the squaring map. Suppose that the arrival set is
endowed with the usual Euclidean topology. Find the smallest topology
on the domain that makes f continuous. (5 pts.)
Answer: The smallest such topology is the set
U U : U open in the usual topology of R.
xx. Let (V, [ [) be a normed real metric space. Show that the maps V V V
and RV V dened by (v, w) v +w and (r, v) rv respectively
are continuous.
xxi. Show that an isometry between metric spaces is a homeomorphism be-
tween the topological spaces that they induce.
We end this subsection with easy but important results.
Theorem 11.2.3 i. The image of a compact set under a continuous map is
compact.
ii. The image of a connected set under a continuous map is compact.
Proof: Let f : X Y be a continuous map.
i. Let K X be a compact subset of X. Let (V
i
)
iI
be an open covering
of f(K). Then (f
1
(V
i
))
iI
is an open covering of f
1
(f(K)). Since K
f
1
(f(K)), (f
1
(V
i
))
iI
is also an open covering of K. Hence K is covered by
f
1
(V
i1
), . . . , f
1
(V
i
k
) for some i
1
, . . . , i
k
I, i.e. K f
1
(V
i1
). . .f
1
(V
i
k
).
By applying f to both sides, we get f(K) f(f
1
(V
i
1
) . . . f
1
(V
i
k
)) =
f(f
1
(V
i
1
)) . . . f(f
1
(V
i
k
)) V
i
1
. . . V
i
k
. This shows that f(K) is
compact.
ii. Let f(X) U V where U and V are open in Y and U V f(X) = .
Then X = f
1
(f(X)) f
1
(U) f
1
(V ) and = f
1
(U V f(X)) =
f
1
(U) f
1
(V ) f
1
(f(X)) = f
1
(U) f
1
(V ) X = f
1
(U) f
1
(V ), so
either f
1
(U) = of f
1
(V ) = , i.e. either U f(X) = or V f(X) = .
Corollary 11.2.4 If K is a compact set and f : K Y a continuous bijection
into a topological space Y then f is a homeomorphism from K onto f(K).
Proof: Let F K be closed. Then F is compact. So f(F) is compact in the
compact subset f(K). So f(F) is closed in f(K). Thus f
1
is continuous.
11.3. CONTINUOUS FUNCTIONS AND R 127
11.3 Continuous Functions and R
Continuous functions assume their extreme values on compact subsets:
Theorem 11.3.1 Let X be a compact topological space and f : X R be
continuous. Then there are a, b X such that f(a) f(x) f(b) for all
x X.
Proof: By Theorem 11.2.3, f(X) is a compact subset of R. By Proposition
10.6.2, f(X) is closed and bounded. Clearly, a closed and bounded subset of
R contains its upper bound. Thus f(X) has a maximal element, say f(b).
Similarly for the existence of a.
A continuous real valued function which is positive once, remains positive
for a while, i.e. if a continuous function is positive at a point then it is positive
in a neighborhood of this point:
Lemma 11.3.2 Let f : (a, b) R and c (a, b). Assume that f is continuous
at c. If f(c) > 0 then f > 0 in an open neighborhood of c. If f(c) < 0 then
f < 0 in an open neighborhood of c.
Proof: Let = f(c). Since f is continuous at c, there is a > 0 such that if
x (c , c +) then f(x) (f(c) , f(c) +) = (0, 2).
A continuous real valued function which assumes two values assumes all the
values between them:
Theorem 11.3.3 (Intermediate Value Theorem) If f : [a, b] R is con-
tinuous and d is between f(a) and f(b), then d = f(c) for some c [a, b].
Proof: Replacing f by f d, we may assume that d = 0. Replacing f by f if
necessary, we may assume that f(a) 0 f(b). We may further assume that
f(a) < 0 < f(b). Consider the set A := x [a, b] : f(x) < 0. Since a A,
A ,= . Let c = sup(A) [a, b]. By Lemma 11.3.2 a < c < b. We will show that
f(c) = 0.
If f(c) < 0 then we get a contradiction by Lemma 11.3.2.
If f(c) > 0 then we again get a contradiction by Lemma 11.3.2.
Thus f(c) = 0.
The image of a compact and connected subset under a continuous real valued
function is a closed interval. Prove this under this generality.
Corollary 11.3.4 If f : [a, b] R is continuous, then f([a, b]) = [m, M]
where m = inf(f([a, b])) and M = sup(f([a, b])).
Theorem 11.3.5 Let f : [a, b] [m, M] be one to one, onto and continuous.
Then f
1
: [m, M] [a, b] is continuous also.
Exercises.
i. [Cauchy] Show that for all real numbers r
1
, . . . , r
n
, (
i
r
i
)
2
n
i
r
2
i
.
(Solution: Replacing r
i
by [r
i
[, we may assume that all r
i
0. If
i
r
i
=
0, the result is obvious. Replacing r
i
by r
i
/
i
r
i
, we may assume that
i
r
i
= 1. So we have to show that if

i
r
i
= 1, then

i
r
2
i
1/n. Let
A = (r
1
, . . . , r
n
) R
0
:

i
r
i
= 1 and let f : A R be dened
by f(r
1
, . . . , r
n
) =

i
r
2
i
. Then A is a compact subset of R
n
because
A is a closed subset of the compact set [0, 1]
n
(Heine-Borel Theorem,
Theorem 10.6.3), being the inverse image of 1 under the continuous map
(r
1
, . . . , r
n
)

i
r
i
. Thus f assumes its minimal value. We will show
that f(r
1
, . . . , r
n
) f(1/n, . . . , 1/n) = 1/n. Let R = (r
1
, . . . , r
n
) be the
point where the minimal value is assumed. Assume r
i
,= r
j
for some i and
j. Assume r
i
< r
j
. Let 0 < < r
j
r
i
. Now look at the point S where all
the coordinates are the same as the coordinates of R except that the i-th
coordinate is r
i
+ and the j-th coordinate is r
j
. Then S A and an
easy calculation shows that f(S) < f(R), a contradiction. Thus r
i
= r
j
for all i, j. Since

i
r
i
= 1, this implies that r
i
= 1/n for all n.)
ii. Let f : [a, b] [a, b] be a continuous function. Show that f has a xed
point.
iii. Show that if f is one to one and continuous on [a, b], then f is strictly
monotone on [a, b].
iv. Let f be a continuous numerical function on the closed interval [a, b]. Let
x
1
, . . . , x
n
be arbitrary points in [a, b]. Show that f(x
0
) = (f(x
1
) + ... +
f(x
n
)) for some x
0
[a, b].
v. Show that there does not exist any continuous function f : R R which
assumes every x R twice.
vi. Does there exist a continuous function f : R R which assumes every
real number three times? (Yes!)
vii. Assume f, g : [0, 1] [0, 1] be continuous functions and assume that
f(0) g(0) and f(1) g(1). Show that there exists an x [0, 1] such
that f(x) = g(x).
viii. Assume that f is a continuous real-valued function on [0, 1] and that
f(0) = f(1). If n N is positive show that there is a point x in [0, 1]
such that f(x) = f(x +
1
n
).
11.4 Uniform Continuity
Exercises.
i. Assume (f
n
)
n
converges uniformly to f on a subset A of R and that each
f
n
is uniformly continuous on A. Show that f is uniformly continuous on
A.
11.5. UNIFORM CONVERGENCE AND CONTINUITY 129
11.5 Uniform Convergence and Continuity
Theorem 11.5.1 Uniform limit of a family of continuous functions is contin-
uous.
11.6 Supplementary Topics
11.6.1 A Continuous Curve Covering [0, 1]
2
Proposition 11.6.1 [0, 1] and [0, 1]
2
are not homeomorphic.
Proof: Take away one point from [0, 1].
Corollary 11.6.2 There is no continuous bijection from [0, 1] onto [0, 1]
2
.
Proof: From Corollary 11.2.4 and Proposition 11.6.1.
Theorem 11.6.3 (Peano, 1890) There is a continuous map (curve) from [0, 1]
onto [0, 1]
2
.
Proof: Let I = [0, 1]. Subdivide I into four closed intervals of equal length
[0, 1/4], [1/4, 1/2], [1/2, 3/4] and [3/4, 1]. Subdivide I
2
into four smaller squares
of length 1/2 1/2: [0, 1/2] [0, 1/2], [0, 1/2] [1/2, 1], [1/2, 1] [1/2, 1] and
[1/2, 1] [0, 1/2]. Make a correspondence between the four smaller intervals
and the four smaller squares in the order they appear above. Call this stage
1. At stage 2, subdivide each segment and each square into four equal parts
again. Make the correspondence in such a way that the intersecting intervals
correspond to neighboring squares. Continue in this fashion. Now for any point
x of [0, 1], consider the set of intervals that x belongs to. The point x will belong
to either one or two of the four intervals of stage n, of total length 1/2
n1
.
To these intervals correspond closed squares of stage n, one inside the other and
of total area /2
2n1
. Because of the size of the squares, these corresponding
squares intersect at a unique point, say f(x). The map f so constructed is a
map from I onto I
2
, but is not a bijection.
This map is continuous, because for x, y I close to each other, the images
f(x) and f(y) are also close to each other, as can be easily checked. In fact if
[x y[ < 1/4
n
then d(f(x), f(y)) 2
2/2
n
.
Note that I and I
2
are not homeomorphic, because taking one point away
from I disconnects I, a phenomenon that does not occur in I
2
. Indeed the
map f
1
dened above is not continuous, as the images under f
1/2
of two
points close to the center of I
2
which belong to the centers of opposite squares
[0, 1/2] [0, 1/2] and [1/2, 1] [1/2, 1] are never at a distance < 1/4.
Question. Is there a continuous bijection from I onto I
2
?
Exercises.
i. Show that a subgroup of R is either discrete, in which case it is one-
generated, or is dense.
ii. Let A R. Show that Z +Z is discrete if and only if Q.
Chapter 12
Dierentiable Functions
DO ALSO DIFFERENTIATION IN C, R
n
and in normed vector
spaces more generally, all at once
Let X R, f : X R a function and a X
. We say that f is
dierentiable at a if
lim
h0
f(a +h) f(a)
h
exists. We then write
f
(a) = lim
h0
f(a +h) f(a)
h
.
If X is open, the function f is called dierentiable on X if f is dierentiable
at all the points of X. Such a function denes a function f
: X R called
the derivative of f.
Examples.
i. Let n N and f : R R be dened by f(x) = x
n
. Then
lim
h0
f(x+h)f(x)
h
= lim
h0
(x+h)
n
x
n
h
= lim
h0
n
i=0

n
i
x
i
h
ni
x
n
h
= lim
h0
n1
i=0

n
i
x
i
h
ni
h
= lim
h0
n1
i=0
_
n
i
_
x
i
h
ni1
= lim
h0
_
n
n 1
_
x
n1
h
0
= nx
n1
.
131
132 CHAPTER 12. DIFFERENTIABLE FUNCTIONS
ii. Let f
n
: [0, 1] [0, 1] be dened by f
n
(x) = x
n
. Then the sequence of
dierentiable functions (f
n
)
n
converges to the function f dened by
f(x) =
_
0 for 0 x < 1
1 for x = 1
and f is not dierentiable.
Lemma 12.1.1 If f is dierentiable at a then f is continuous at a. Thus a
dierentiable function is continuous.
Exercises.
i. Let f : R R be the squaring function f(x) = x
2
. Show that f is
dierentiable and f
(x) = 2x.
ii. Let n N 0 and f
n
n
(x) = [x[
1+1/n
.
Show that f
n
is dierentiable on [1, 1].
iii. Show that the function
f(x) =
_
x
2
sin(1/x) if x ,= 0
0 if x = 0
is dierentiable but f
is not continuous.
12.2 Dierentiation of Complex Functions
12.3 Basic Properties of Dierentiable Functions
From now on we let f and a be as above.
Lemma 12.3.1 If f is dierentiable at a and if f(a) is maximal then f
(a) = 0.
Exercises.
i. Let f : U R (or C) be dierentiable at a U, where U is open.
Assume that there is a sequence (x
n
)
n
of U converging to a such that
f(x
n
) = f(x
m
) all n, m. Then f(a) = f(x
n
) and f
(a) = 0. (Hint:
Follows almost directly from the denition of derivatives).
ii. Let f : [a, b] R be dierentiable. Assume that f
(x) ,= 0 for x [a, b].

Then f has nitely many zeroes in [a, b]. (Hint: See Exercise i, page 132).
12.4. RULES OF DIFFERENTIATION 133
12.4 Rules of Dierentiation
12.5 Relationship Between a Function and Its
Derivative
Letf : X R be a function from a topological space into R. We say that a
point x
is a local maximum (resp. local minimum) of f if there is an open

subset U of X containing x
such that for all x U, f(x) f(x
) (resp.
f(x) f(x
)). An absolute maximum (resp. absolute minimum) of f is a

point x
such that x U, f(x) f(x
) (resp. f(x) f(x
)). An extremum
(local or absolute) point is a maximum or a minimum point.
Theorem 12.5.1 Let x
X R and f : X R be a function dierentiable

at x
. If x
is a local extremum then f
(x
) = 0.
Proof: Suppose x
is a local maximum. Then there is a > 0 such that

(x
, x
+) X and
f(x) f(x
) if x (x
, x
)
f(x) f(x
) if x (x
, x
+).
Now
f(x
+h)f(x
)
h
0 for h (0, )
f(x+h)f(x)
h
0 for h (, 0).
Thus f
(x
) = lim
n
=
f(x
+h)f(x
)
h
0. If x
is a local minimum the proof

is similar.
Theorem 12.5.2 (Rolles Theorem) For a < b, let f : [a, b] R be dier-
entiable on (a, b). If f(a) = f(b) then f
(c) = 0 for some c (a, b).

Proof: If f is constant then f
= 0 and there is nothing to prove. Assume

f is not a constant. By Theorem 11.3.1, f has an absolute maximum and an
absolute minimum. Since f is not constant, these are two distinct points. Since
f(a) = f(b), one of these points, say c, must be in the open interval (a, b). By
Theorem 12.5.1, f
(c) = 0.
The next theorem is one of the most important result in this series. We will
have several opportunities to use it.
Theorem 12.5.3 (Mean Value Theorem of Dierential Calculus) For a <
b, let f : [a, b] R be dierentiable on (a, b). Then
f(a) f(b)
a b
= f
(c)
for some c (a, b).
Proof: For t [0, 1], let g(t) = f(ta + (1 t)b) tf(a) + (1 t)f(b). Then
g(0) = g(1) = 0. By Rolles Theorem (12.5.2), g
(t
) = 0 for some t
(0, 1).
But g
(t) = (a b)f
(ta + (1 t)b) f(a) + f(b). Thus 0 = g
(t
) = (a
b)f
(t
a +(1 t
)b) f(a) +f(b). If c = t
a +(1 t
)b, then f
(c) =
f(a)f(b)
ab
.
Theorem 12.5.4 If f is dierentiable on (a, b) and f
(x) 0 for every x

(a, b) then f is monotone decreasing on (a, b).
Proof: Let a < x
1
< x
2
< b. By the Mean Value Theorem of Dierential
Calculus (12.5.3), there is an x
(x
1
, x
2
) such that f
(x
) =
f(x
2
)f(x
1
)
x2x1
.
Since f
(x
) 0 and x
2
x
1
> 0, f(x
2
) f(x
1
) 0.
The fact that f
(c) > 0 at some c does not imply that f is increasing around

c. It is easy to nd a counterexample, nd one.
12.6 Uniform Convergence and Dierentiation
The fact that a sequence of dierentiable function (f
n
)
n
converges uniformly to
a function f, does not imply that f is dierentiable.
Example. Let f
n
n
(x) = [x[
1+
1
2n+1
. Then
each f
n
is dierentiable on [0, 1] (see Exercise ii, page 132). Also lim
n
f
n
(x)
u
=
[x[. (See Exercise iii, page 132). But the function f(x) = [x[ is not dierentiable
at x = 0.
Question. Suppose (f
n
)
n
is a sequence of dierentiable functions that con-
verge uniformly to a dierentiable function f. Is it true that f
p
= lim
n
f
n
?
Theorem 12.6.1 Let the sequence of dierentiable functions (f
n
)
n
converge
pointwise to f on the closed interval [a, b]. Assume that the sequence (f
n
)
n
converges uniformly. Then f is dierentiable and f
= lim
n
f
n
.
Proof: Let lim
n
f
n
u
= g. Let c [a, b]. We want to show that
lim
xc
f(x) f(c)
x c
= g(c).
Thus given an > 0, we want to nd a > 0 such that
f(x) f(c)
x c
g(c)
<
for all x [a, b] that satises 0 < [x c[ < . Let > 0 be given. By the
triangular inequality, for any n, we have,
f(x)f(c)
xc
g(c)
f(x)f(c)
xc

fn(x)fn(c)
xc
fn(x)fn(c)
xc
f
n
(c)
+[f
n
(c) g(c)[.
12.6. UNIFORM CONVERGENCE AND DIFFERENTIATION 135
We will make each term of the right hand side less than /3. It is easy to make
the second term small by choosing small enough. It is also easy to make the
last term small by choosing n large enough. The main problem is with the rst
term, that we deal rst.
Assume c < x. The proof is the same if x < c. Let n and m be any natural
numbers. Applying the Mean Value Theorem to the function f
n
f
m
on the
interval [c, x] we can nd an =
m,n
(c, x) such that
f
m
() f
n
() =
f
m
(x) f
n
(x)
x c

f
m
(c) f
n
(c)
x c
.
On the other hand, by the Cauchy Criterion for Uniform Convergence (Theorem
9.2.5), there exists an N
1
such that for all n, m > N
1
and x [a, b],
[f
m
(x) f
n
(x)[ < /3.
In particular,
[f
m
() f
n
()[ < /3.
(It is worth while to notice that, since depends on n and m, we really need
the uniform convergence of (f
n
)
n
for this part of the argument). Thus
f
m
(x) f
n
(x)
x c

f
m
(c) f
n
(c)
x c
< /3
for all n, m > N
1
and all x [a, b]. Now let m go to innity, to get
(1)
f(x) f(c)
x c

f
n
(x) f
n
(c)
x c
< /3
for all n > N
1
and any x [a, b].
Since lim
n
f
n
= g, there is an N
2
large enough so that,
(2) [f
n
(c) g(c)[ < /3
for all n > N
2
.
Let n = maxN
1
, N
2
+ 1. Note that (1) and (2) hold for this n. Choose
> 0 such that if x [a, b] and [x c[ < , then
(3)
f
n
(x) f
n
(c)
x c
f
n
(c)
< /3.
Now (1), (2) and (3) give us the result.
In the theorem above we do not really need the pointwise convergence of
(f
n
)
n
on [a, b], it is enough to know that (f
n
(x
0
))
n
converges for some n to get
the pointwise, in fact the uniform convergence on [a, b].
Theorem 12.6.2 Let (f
n
)
n
be a sequence of dierentiable functions dened
and assume that the sequence (f
n
)
n
converges uniformly on [a, b]. If there exists
a point x
0
[a, b] such that (f
n
(x
0
))
n
is convergent. Then (f
n
)
n
converges uni-
formly on [a, b] and lim
n
f
n
is dierentiable and (lim
n
f
n
)
= lim
n
f
n
.
Proof: In view of the previous theorem, it is enough to prove the uniform
convergence of (f
n
)
n
on [a, b].
COMPLETE THE PROOF.
Exercises.
12.7 Second and Further Derivatives
Chapter 13
Analytic Functions
13.1 Power Series
We collect the main results and a few more about power series. This will show
that the power series whenever and wherever they converge behave like
polynomials, as much as they can.
Theorem 13.1.1 Let

i=0
a
i
z
i
be a power series with R = lim1/[a
i
[
1/i

R
0
its radius of convergence. Let 0 < r < R. Then the following
hold.
i. If [z[ < R, then
i=0
a
i
z
i
converges absolutely. If [z[ > R, then
i=0
a
i
z
i
diverges.
ii.

i=0
a
i
z
i
converges uniformly on B(0, r).
iii. The radius of convergence of
i=1
ia
i
z
i1
iv.

i=0
a
i
z
i
is innitely dierentiable in B(0, R) and
(
i=0
a
i
z
i
)
i=1
ia
i
z
i1
.
v. If f(z) =

i=0
a
i
z
i
in B(0, R), then f is innitely dierentiable in
B(0, R) and a
i
= f
(i)
(0)/i!. In other words, the coecients of a power series
are unique and if the function f(z) is a power series, then
f(z) =
i=0
f
(i)
i!
z
i
.
Proof: Part (i) is Corollary 7.5.6.
Part (ii) is by Proposition 6.8.3 and part (i).
iii. Since

i=0
a
i
z
i
converges absolutely for [z[ = r, the series

i=0
a
i
r
i
converges. By Weierstrass M-Test (Theorem 9.3.3), the series f =

i=0
a
i
z
i
137
138 CHAPTER 13. ANALYTIC FUNCTIONS
iv. By parts (i, iii), the series

i=1
ia
i
z
i1
Let f
n
=

n
i=0
a
i
z
i
. By Theorem 12.6.1, f =

i=0
a
i
z
i
is dierentiable and
f
= lim
n
f
n
= lim
n
n
i=1
ia
i
z
i1
=
i=1
ia
i
z
i1
for z B(0, r), so also
for all z B(0, R).
v. Direct application of the previous facts.
Corollary 13.1.2 Consider the innite series

i=0
a
i
(z a)
i
with 0 < R =
lim1/[a
i
[
1/i
R
0
its radius of convergence. Let 0 < r < R.
Then the following hold.
i. If z B(a, R), then

i=0
a
i
(z a)
i
converges absolutely. If z , B(a, R),
then

i=0
a
i
(z a)
i
diverges.
ii. The radius of convergence of

i=0
a
i
(z a)
i
is also R.
iii.

i=0
a
i
(z a)
i
converges uniformly on B(a, r).
iv.

i=0
a
i
(z a)
i
is innitely dierentiable in B(a, R) and
(
i=0
a
i
(z a)
i
)
i=1
ia
i
(z a)
i1
,
the convergence being uniform in B(a, r).
v. If f(z) =
i=0
a
i
(za)
i
for z B(a, R), then f is innitely dierentiable
in B(a, R) and a
i
= f
(i)
(a)/i!.
Exercises.
i. Let
i=1
a
i
/n
i
be convergent for all n > 1. Show that lim
n
i=0
a
i
/n
i
=
a
0
. (Solution: Consider the power series f(x) :=
i=1
a
i
x
i
which is con-
vergent at x = 1/2, thus its radius of convergence is at least 1/2. Thus f(x)
is continuous at 0. It follows that for all > 0 there is a > 0 such that if
[x[ < and x ,= 0 then [f(x) f(0)[ < . Let 1/ 1 < N = [1/] 1/.
Then for n > N, 1/n 1/(N + 1) < and so [f(1/n) f(0)[ < , i.e.
[
i=1
a
i
/n
i
a
0
[ < . Thus lim
n
i=0
a
i
/n
i
= a
0
.)
13.2 Taylor Series
As seen above, the power series

i=0
a
i
z
i
and their cousin

i=0
a
i
(z a)
i
are
easy to handle, they behave almost like polynomials. According to these results,
it would be helpful if every function could be expressed as such an innite series
around any a. But since such a series is innitely dierentiable if it converges
around a, a function which is not innitely dierentiable around a cannot be
expressed as such a series. As it happens, even this condition is not sucient
for a function to be expressed as such a series. We will see an example of such
a function later.
A function f dened on an open subset U (of R or of C) is said to be analytic
at a point a U if there is a sequence (a
i
)
i
such that f(z) =

i=0
a
i
(z a)
i
for z B(a, R) for some R > 0. The coecients a
i
are unique and equal to
13.2. TAYLOR SERIES 139
(i)
(a)/i! as Theorem 13.1.1 shows. Also we can nd a maximal R where the
equality f(z) =
i=0
a
i
(z a)
i
holds.
If f is a complex function, then the fact that f is dierentiable on U once
is enough to insure that for any a U, f =
i=0
a
i
(z a)
i
for z B(a, R) for
some R > 0. This result is part of a subject called complex analysis and will
be seen somewhere else. The situation in R is much more mysterious and is the
subject of this section.
Lemma 13.2.1 Let U be an open subset of R and f : U R an (n+1)-times
dierentiable function. Let [a, b] U. Then
f(b) = f(a)+
f
(a)
1!
(ba)+
f
(a)
2!
(ba)
2
+. . . +
f
(n)
(a)
n!
(ba)
n
+
f
(n+1)
(c)
(n + 1)!
(ba)
n+1
for some c (a, b).
Proof: Let
g(x) = f(b)f(x)(bx)f
(x)
(b x)
2
f
(x)
2!
. . .
(b x)
n
f
(n)
(x)
n!

(b x)
n+1
(n + 1)!
where is chosen so that g(a) = 0. Since g(b) = 0 also, by Rolles Theorem
(12.5.2), there is a c (a, b) such that g
(c) = 0. Note that, after simplication,

g
(x) = [f(b)
n
i=0
(bx)
i
f
(i)
(x)
i!

(bx)
n+1
(n+1)!
]
=

n
i=1
(bx)
i1
f
(i)
(x)
(i1)!

n
i=0
(bx)
i
f
(i+1)
(x)
i!
+
(bx)
n
n!
=
(bx)
n
f
(n+1)
(x)
n!
+
(bx)
n
n!
.
Thus = f
(n+1)
(c). Now the equation g(a) = 0 gives the desired result.
Corollary 13.2.2 Let U be an open subset of R and f : U R an (n + 1)-
times dierentiable function. Let [a, b] U. Then for all x [a, b) there is a
(x, b) (that depends on a and x) such that
f(x) = f(a)+
f
(a)
1!
(xa)+
f
(a)
2!
(xa)
2
+. . .+
f
(n)
(a)
n!
(xa)
n
+
f
(n+1)
()
(n + 1)!
(xa)
n+1
.
It follows that the innitely dierentiable real function f is equal to its
Taylor series
i=0
f
(i)
i!
(z a)
i
around a (so is analytic) if and only if [
f
(n+1)
()
(n+1)!
(z a)
n+1
[ converges to 0 as
n goes to innity. The term (n + 1)! is helpful to make the quantity small. If
z is close to a, (z a)
n+1
can also be made small. Only the term f
(n+1)
() is
bothersome. Note that here depends on n, a and x.
13.2.1 Calculating Taylor Polynomials
Theorem 13.2.3 Suppose that on an open subset we have
f(x) =
n
i=0
a
i
x
i
+x
n
(x)
where lim
x
(x) = 0 and f is (n + 1)-times dierentiable. Then Tf
n
(x) =
n
i=0
a
i
x
i
. ????
Taylor Polynomials of 1/ cos x
We start noting that since 1/ cos x is an even function, its derivatives of odd
degree must be 0 at x = 0.
First Method. Let f(x) = 1/ cos x and start computing its derived series.
This may take a long time and is not advised.
Second Method. Let us compute the rst 6 (or 7) terms of the Taylor
series of 1/ cos x around 0. For this we take the inverse of the Taylor series for
cos x around 0.
1
cos x
=
1
1x
2
/2!+x
4
/4!x
6
/6!+x
7
f(x)
=
1
1(x
2
/2!x
4
/4!+x
6
/6!x
7
f(x))
= 1 + (x
2
/2! x
4
/4! +x
6
/6! x
7
f(x))+
+(x
2
/2! x
4
/4! +x
6
/6! x
7
f(x))
2
+
+(x
2
/2! x
4
/4! +x
6
/6! x
7
f(x))
3
= 1 + (x
2
/2! x
4
/4! +x
6
/6!) + (x
4
/4 x
6
/4!) +x
6
/8 +x
7
g(x)
= 1 +x
2
/2 x
4
(1/4! 1/4) +x
6
(1/6! 1/4! + 1/8) +x
7
g(x)
= 1 +x
2
/2 + 5x
4
/24 + 61x
6
/720 +x
7
g(x),
where f and g are functions that converge to 0 when x goes to 0.
Taylor Polynomial of tanx
Note rst that since tan x is an odd function, its derivatives of even degree must
be 0.
First Method. Let f(x) = tanx and start computing its derived series.
This is long and tiring and is not advised.
Second Method. We use the Taylor polynomial of 1/ cos x that has been
computed above:
tanx = sinx
1
cos x
=
_
x
x
3
3!
+
x
5
5!
+x
6
f(x)
__
1 +
x
2
2
+
5x
4
24
+
61x
6
720
+x
6
g(x)
_
= x +
_
1
3!
+
1
2
_
x
3
+
_
1
5!

1
12
+
5
24
_
x
5
+x
6
h(x)
= x +
1
3
x
3
+
2
15
x
5
,
where f, g and h are functions that converge to 0 when x goes to 0.
Third Method for those who know how to integrate. Since tan
x =
1/ cos
2
x, if we know the Taylor polynomial of 1/ cos
2
x, we can integrate to
13.2. TAYLOR SERIES 141
nd the Taylor polynomial of tanx, except may be the rst term. But since
tan0 = 0, the rst term must be 0.
Fourth Method. Since tanx is (ln(cos x))
= tanx, if we know the

Taylor polynomial of ln(cos x), then by dierentiating, we can nd the Taylor
polynomial of tanx:
ln(cos x) = ln
_
1
x
2
2!
+
x
4
4!

x
6
6!
x
7
f(x)
_
= ln
_
1
_
x
2
2!

x
4
4!
+
x
6
6!
+x
7
f(x)
__
=
_
x
2
2!

x
4
4!
+
x
6
6!
+x
7
f(x)
_
1
2
_
x
2
2!

x
4
4!
+
x
6
6!
+x
7
f(x)
_
2
1
3
_
x
2
2!

x
4
4!
+
x
6
6!
+x
7
f(x)
_
3
=
1
2
x
2
1
12
x
4
2
90
x
6
+x
7
g(x)
Hence
tanx = (ln(cos x))
=
_
1
2
x
2
+
1
12
x
4
+
2
90
x
6
+x
7
g(x)
_
= x+
1
3
x
3
+
2
15
x
5
+x
6
h(x)
where the limits of g and h when x goes to 0 are 0.
Taylor Polynomial of sin
3
x
We compute up to ninth degree:
sin
3
(x) =
_
x
1!

x
3
3!
+
x
5
5!

x
7
7!
+x
8
f(x)
_
3
= x
3
_
1
1!

x
2
3!
+
x
4
5!

x
6
7!
+x
7
f(x)
_
3
= x
3
_
1 + 3
11(1)
1!1!3!
x
2
+
_
3
111
1!1!5!
+ 3
1(1)(1)
1!3!3!
_
x
4
+
+
_
(1)(1)(1)
3!3!3!
+ 3!
1(1)1
1!3!5!
+ 3
11(1)
1!1!7!
_
x
6
+x
7
(x)
_
= x
3
_
1
1
2
x
2
+
13
120
x
4
41
3024
x
6
+x
7
(x)
_
= x
3
1
2
x
5
+
13
120
x
7
41
3024
x
9
+x
10
(x)
where the limits of f, and
when x goes to 0 are 0.

Exercises.
i. Show that the derivatives of even (resp. odd) degree of an analytic function
which is odd (resp. even) must be 0.
ii. Let f : R
>0
R be twice dierentiable. Assume that f > 0 and f
< 0
then f
cannot be always negative.

Proof: Assume f
(x) < 0 for all x. Let a R

>0
be xed. Then for all
x > a, f(x) = f(a) + (x a)f
(a) + (b x)
2
f
(c)/2! for some c. We can

choose x large enough so that f(a) +(x a)f
(a) < 0. Then f(x) < 0.

iii. If f(x) =
1
1+x
and U = (1, ), solve the rst two questions.
iv. a. Estimate the error made in replacing the function exp on the interval
[0, 1] by its Taylor polynomial of degree 10.
b. On what interval [0, h] does the function exp dier from its Taylor
polynomial of degree 10 by no more than 10
7
?
c. For what value of n does the function exp dier from its Taylor poly-
nomial of degree n by no more than 10
7
on the interval [0, 1]?
13.3 Analytic Functions
If f is innitely dierentiable and a, x U, we set
(Tf)(x) =
n=0
f
(n)
(a)
n!
(x a)
n
.
Note that (Tf)(x) may or may not be convergent for a given x. But if Tf is
convergent, then Tf and f have the same n
th
-derivatives at 0 for all n. Does
this condition implies that Tf = f. Thus we have two questions:
1. For what values of x is (Tf)(x) convergent? Is Tf convergent for all
x U.
2. In case (Tf)(x) is convergent on U, do we have (Tf)(x) = f(x) for all x?
The answer to both questions is negative as the next examples show.
Examples.
i. Let f be the real function f(x) =
1
1+x
2
. Then Tf(x) =

i=0
(1)
i
x
2i
.
It can be easily checked that Tf converges for x (1, 1) and diverges
elsewhere, although f is dened everywhere. Thus the answer to the rst
question is negative.
ii. Let
f(x) =
_
e
1/x
2
if x ,= 0
0 if x = 0
Then f is innitely dierentiable and f
(n)
(0) = 0 for all n N. Therefore
Tf = 0 and Tf(x) = f(x) if and only if x = 0. Thus the answer to the
second question is also negative.
A function f dened on an open subset U (of R or of C) is said to be analytic
at a point a U if there is a sequence (a
i
)
i
such that f(z) =

i=0
a
i
(z a)
i
for z B(a, R) for some R > 0. The coecients a
i
are unique and equal to
(i)
(a)/i! as Theorem 13.1.1 shows. Also we can nd a maximal R where the
equality f(z) =
i=0
a
i
(z a)
i
holds.
Proposition 13.3.1 If f : B(0, R) C is analytic and f = 0 on RB(0, R)
then f = 0.
Proof: . By Exercise i, page 132, f
= 0 on R. Thus f
(n)
(0) = 0 for all n and
so f = Tf = 0.
13.4. TRANSCENDENTAL FUNCTIONS 143
13.4 Transcendental Functions
We dene
exp(z) =

n=0
z
n
/n!
sin(z) =

n=0
(1)
n
z
2n+1
/(2n + 1)!
cos(z) =

n=0
(1)
n
z
2n
/(2n)!
cosh(z) =

n=0
z
2n+1
/(2n + 1)!
sinh(z) =

n=0
z
2n
/(2n)!
Each of these functions are analytic on C as it can be checked by using
Theorem 13.1.1. Also they all take real values at real numbers, i.e. they can
also be considered as real functions.
Exercises.
i. Show that lim
n
(1 +z/n)
n
= exp(z). (See Example i, page 111).
ii. Show that sinz =
exp(iz)exp(iz)
2i
.
13.4.1 Exponentiation and Trigonometric Functions
We have exp(0) = 1, cos(0) = 1, sin(0) = 0. By Theorem 7.5.7, exp
= exp,
sin
= cos, cos
= sin, cosh
= sinh, sinh
= cosh
, cos
2
(x) + sin
2
(x) = 1.
Theorem 13.4.1 exp(x +y) = exp(x) exp(y).
Proof: Since exp(z) =
n=0
z
n
/n!, we have to prove that the limit of
(
n
i=0
x
i
/i!)(
n
i=0
y
i
/i!)
n
i=0
(x +y)
i
/i!
is 0 as n . We rst compute the left hand side:
(
n
i=0
x
i
/i!)(
n
i=0
y
i
/i!) =
n
i=0
n
j=0
x
i
y
j
i!j!
.
Now we compute the left hand side:
n
i=0
(x +y)
i
/i! =
n
i=0
i
j=0
_
i
j
_
x
j
y
ij
/i! =
n
i=0
i
j=0
x
j
y
ij
(i j)!j!
=
n
i=0
i
k=0
x
j
y
k
k!j!
.
Another proof of Theorem 13.4.1: Let a R. Let f(x) = exp(x + a)

and g(x) = exp(x) exp(a). Then f(0) = g(0) and f = f
and g = g
. Thus
f
(n)
(0) = g
(n)
(0) all n. So f = g by Theorem 13.1.1.
Theorem 13.4.2 exp(iz) = cos z +i sinz for all z C.
Proof: Easy.
Corollary 13.4.3 sin(x + y) = sin(x) cos(y) + cos(x) sin(y) and cos(x + y) =
cos(x) cos(y) sin(x) sin(y) for all x, y C.
Proof: If x, y R, this follows from Theorems 13.4.2 and 13.4.1. Now ap-
ply Proposition 13.3.1 to the analytic functions sin(x + a) sin(x) cos(a) +
cos(x) sin(a) and cos(x +a) cos(x) cos(a) sin(x) sin(a).
Theorem 13.4.4 The functions sin and cos restricted to R have a common
period.
Proof: Assume cos(x) > 0 for all x. Then sin is increasing. Since sin(0) = 0,
sin(x) > 0 for all x > 0. Thus cos
= sin < 0 on R
>0
. Also, cos
(x) =
cos(x) < 0 and this contradicts Exercise ii, page 141. Thus cos(x) = 0 for
some x > 0. Then sin(x) = 1. So cos(2x) = cos
2
(x) sin
2
(x) = 1 and
sin(2x) = 0. Thus A := R
>0
: sin() = 0 is nonempty. By the formulas
sin(x + y) = sin xcos y + siny cos x and sin(x) = sinx, we see that the set
A of zeroes of sinx = 0 is an additive subgroup of R. Since sin is continuous,
this subgroup is closed. If it had an accumulation point, the accumulation point
itself would be in A, but then 0 would be an accumulation point. Let (a, b) be
an open interval. Assume 0 < a < b. Let A be such that 0 < < b a.
Then n (a, b) A for some n N. So A is dense in R. Since A is also closed,
this implies that A = R and sin = 0, a contradiction. Hence A is discrete. Let
> 0 be the least positive element of A. It is easy to show that the period of
sin is 2.
Also follows from Lemma i.
DETAILS
It follows from the above proof that a subgroup of R is either dense in R or
is generated by one element.
Find an upper bound for .
The number e. We let e = exp(1) =
i=0
1/i!.
Proposition 13.4.5 e is irrational.
Proof: Assume e = m/n for m and n > 0 integers. Then the number
N := n!
_
e
n
i=0
1/i!
_
=
i=n+1
n!/i! > 0
is a positive natural number and we have
N =

i=n+1
n!/i! =
i=1
1/(n + 1) . . . (n +i)
<
i=1
1/(n + 1)
i
=
1
n+1
i=0
1/(n + 1)
i
=
1
n+1
1
1
1
n+1
= 1/n,
13.4. TRANSCENDENTAL FUNCTIONS 145
implying N = 0, a contradiction.
See [AZ, page 28] for the fact that e
q
is irrational for all q Q 0.
The proposition above is a consequence of a more general result:
Theorem 13.4.6 If (a
i
)
i
is a sequence of zeroes and ones which is not even-
tually zero, then

i=0
a
i
/i! is irrational.
Proof: The same as above. Take this proof to its appropriate place.
Exercises.
i. Find cos(15
).
ii. Express sin(4x) and cos(4x) in terms of sinx and cos x. (Prove your
formula).
iii. Let f(x) = x
3
3x + 5. Show that f(ln a) = 6 for some a > 1.
iv. Show that
exp(2iz)1
exp(2iz)+1
= i tanz for all z C.
v. Describe the set R such that the subgroup 1, ) = Z+Z is discrete.
(Solution: If = p/q Q, then it is easy to show that 1, ) = 1/q)
is discrete. Otherwise, the elements of the sequence (n [n])
nN
are
distinct and are all in [0, 1)1, ). Thus 1, ) has an accumulation point
in [0, 1). Thus 0 is an accumulation point of 1, ). Now for 0 <
in R, choose
n
1, ) (0, ). For some k N, < k
n
and
k
n
(, )).
vi. Assuming is irrational, show that sin(Z) is dense in [1, 1]. (Solution:
Let y [1, 1]. Let x R be such that sin x = y. Since Z + 2Z is dense
in R, there is a sequence a
n
+2b
n
with a
n
, b
n
Z such that lim
n
(a
n
+
2b
n
) = x. Since sin is continuous, y = sin x = sin(lim
n
a
n
+ 2b
n
) =
lim
n
sin(a
n
+ 2b
n
) = lim
n
sina
n
.)
vii. Show that

n=1
sin n
n
is convergent. (Solution: We will apply Abels
Theorem (Theorem 7.5.8). We only need to show that the partial sums of
[
m
j=n
sinn[ are bounded. Since sinn = Im(
in
), it is enough to show that
the partial sums [
m
j=n
e
ij
[ are bounded. We will prove that the sums
[
m
j=0
e
ij
[ are bounded. This will prove the result since [
m
j=n
e
ij
[ =
[
m
j=0
e
ij

n1
j=0
e
ij
[ [
m
j=0
e
ij
[ + [
n1
j=0
e
ij
[. Now we compute:
[
m
j=0
e
ij
[ = [
e
i(m+1)
1
e
i
1
[
2
|e
i
1|
.)
viii. Find the radius of convergence of
n=1
sin n
n
z
n
. (Solution: Since
sin n
n
z
n
[z[
n
, the radius of convergence 1. If the radius of convergence is > 1,
then the radius of convergence of the derived series

n=0
sinnz
n1
is
also > 1. But by Exercise vi, page 145, the general term sinnz
n
does not
converge to 0 if [z[ ,< 1).
ix. Show that lim
n
(1 + 1/n)
n
= e.
x. Find the following limits:
lim
x
(1 + 1/x)
x
lim
x
(1 1/x)
x
lim
n
(1 1/n)
n
lim
n
(1 1/n
2
)
n
2
lim
n
(1 1/n
2
)
n
lim
n
(1 1/n)
n
2
lim
x
(1 +c/x)
x
13.4.2 Inverse Trigonometric Functions
13.4.3 Logarithm
13.4.4 Hyperbolic Functions
13.5 Supplement
13.5.1 Trigonometric Functions
By Example i, page 111, lim
n
(1 +z/n)
n
= exp(z). Thus
sinz =
exp(iz) exp(iz)
2i
= lim
n
1
2i
[(1 +iz/n)
n
(1 iz/n)
n
].
On the other hand, we note that (1 + iz/n)
n
(1 iz/n)
n
= 0 if and only if
_
1+iz/n
1iz/n
_
n
= 1 if and only if
_
n+iz
niz
_
n
= 1 if and only if
n+iz
niz
= exp(2ik/n)
for some k = 0, 1, . . . , n 1 if and only if (as an easy calculation shows) z =
n
i
exp(2ik/n)1
exp(2ik/n)+1
= ntankn by Exercise iv, page 145.
ppppp DEVAM EDECEK, VALIRON, SAYFA 41
13.5.2 Series
Proposition 13.5.1 For x ,= 2k (k Z) and (
n
)
n
a decreasing sequence
whose limit is 0, the series

n=0
n
e
inx
is convergent.
Proof: We will apply Theorem 7.5.8. Let x ,= 2k for any k Z and v
n
=
v
n
(x) = e
inx
. For m < n, v
m
+ . . . + v
n
= e
imx
+ . . . + e
inx
= e
imx
(1 + e
ix
+
. . . +e
i(nm)x
= e
imx 1e
i(nm+1)x
1e
i
x
, so that [v
m
+. . . +v
n
[
2
|1e
ix
|
. Hence the
conditions of Theorem 7.5.8 are met.
Corollary 13.5.2 For x ,= 2k (k Z) and (
n
)
n
a decreasing sequence whose
limit is 0, the series

n=0
n
sin(nx) and

i=0
n
cos(nx) are convergent.
13.6. NOTES 147
Exercises.
i. Show that

k=1
sin(kx)/k
2
is uniformly convergent on (, ).
ii. Show that

k=0
(xe
x
)
k
converges uniformly on [0, 2].
iii. Show the following:
sin
2
(z) + cos
2
(z) = 1
exp(iz) = cos(z) +i sin(z)
cos is even, sin is odd
sinh
2
(z) cosh
2
(z) = 1
exp(z) = cosh(z) +i sinh(z).
iv. Let X be a set with n elements. Let 0 p := p(n) 1. We select a
random subset A := A(p) of Xin such a way that the probability that an
element x X is in A is p.
i. Calculate the probability that A has any element at all.
ii. What is the expected number of elements of A?
iii. Let p(n) = 1/n. What is the probability that A(p) ,= when n tends
to innity?
iv. Assume that p = p(n) 1/n. Show that the probability that A(p) =
when n tends to innity is 1.
v. Assume that p = p(n) 1/n. Show that the probability that A(p) =
when n tends to innity is 0.
13.6 Notes
Analytic functions are not the only functions whose sequence (f
(n)
(x
0
))
n
of
derivatives determine the function in a neighborhood of x
0
. The functions de-
ned and innitely dierentiable on [a, b] such that if M
n
= max[f
(n)
(x)[ :
x [a, b], then

n
1/M
1/n
n
is divergent (Denjoy 1921 and Carleman). Such
functions are called quasi analytic.
Chapter 14
Graph Drawing
14.1 Drawing in Cartesian Coordinates
14.1.1 Asymptotes
Exercises.
i. Draw with as much care as possible the graph of f(x) =
x
2
(x1)(x+2)
.
ii. Let f(x) = x + 2x
2
sin(1/x) if x = 0 and f(0) = 0.
i. Show that f
(0) = 1.
ii. Show that any interval containing 0 also contains points where f
(x) <
0, so that f cannot be increasing on any interval around 0 although f
(0) >
0.
14.2 Parametric Equations
14.3 Polar Coordinates
14.4 Geometric Loci
Exercise. [BR] Let A = (1, 0). Find the set of points M of the plane R
2
such
that, if P and Q design the projections of M onto the axes x and y respectively,
the point S of intersection of AQ and PM is at distance 1 from A.
First Solution. Let M = (x
0
, y
0
). Then P = (x
0
, 0) and Q = (0, y
0
). The line
AQ has equation y =
0y0
10
x+y
0
. Thus the point S has coordinates (x
0
, y
0
x
0
+
y
0
). The square of its distance from A(1, 0) is (x
0
1)
2
+(y
0
x
0
+y
0
)
2
, which
we want to be 1. Thus the equation of the curve in Cartesian coordinates is
(x 1)
2
+ (yx + y)
2
= 1, or (x 1)
2
(1 + y
2
) = 1, or y
2
=
1
(x1)
2
1, i.e.
y =
_
1
(x1)
2
1
_
1/2
. Note that x ,= 1 and
1
(x1)
2
1 0, i.e. x [0, 2] 1.
149
150 CHAPTER 14. GRAPH DRAWING
Note also that the curve is symmetric with respect to the x axis. The line x = 1
is an asymptote and the curve is also symmetrical with respect to the line x = 1.
It is possible to draw this curve, even if some painful work is necessary.
Second Solution. Let be the angle PAQ. Then one can see immediately
that
x() = 1 cos
y() = tan()
where [0, 2] /2, 3/2. It is enough to draw the curve when [0, /2)
because of the obvious symmetries.
It is easy to see that
i) x(0) = y(0) = 0.
ii) x and y both increase when increases from 0 to /2.
iii) When goes to /2, x goes to 1 and y goes to . Thus there is an
asymptote at = /2, which is the line x = 1.
Dierentiating with respect to , we nd
dx/d = sin
dy/d = 1/ cos
2
.
Thus
dy/dx = (dy/d)/(dx/d) = 1/ cos
2
sin.
It follows that when [0, /2), y is an increasing function of x. It also follows
that the y axis is tangent to the curve (case = 0).
Now we nd d
2
y/dx
2
:
d
2
y/dx
2
= d(dy/dx)/dx = (d(dy/dx)/d)/(dx/d) = (1/ cos
2
sin)
/ sin
=
2 sin
2
cos
2
cos
3
sin
3
.
There is an inection point when 2 sin
2
cos
2
= 0, i.e. when y = tan() =
1/
2. We can compute the rst coordinate x of the inection point: 1/ cos
2
1 = tan
2
= 1/2, cos =
_
2/3 and x = 1 cos = 1
_
2/3. The graph is
concave down before (1
_
2/3, 1/
2) and concave up afterwards.

Exercise. [BR] To a point M = (X, Y ), Y ,= 0, we associate a point m as
follows: The line OM intersects the line y = a in H. The lines passing from M
and H and parallel to the x and y axes intersect at m. Given m, we can also
nd M.
1. Calculate the coordinates of m in terms of the coordinates of M.
2. Find the curve that the point m traces when M moves on the circle
X
2
+Y
2
2RX = 0.
3. Find the curve that the point M traces when m moves on the circle
x
2
+y
2
2Ry = 0.
Solution.
14.5. SUPPLEMENT 151
14.5 Supplement
Theorem 14.5.1 (Cauchys Mean-Value Theorem) If f and g are both
continuous on [a, b] and dierentiable on (a, b), then there exists a c (a, b)
such that
f
(c)[g(b) g(a)] = g
(c)[f(b) f(a)].
Proof: Let F : [a, b] R be dened by
F(x) = f(x)[g(b) g(a)] g(x)[f(b) f(a)].
Then F is continuous on [a, b] and dierentiable on (a, b). It is easy to check
that F(a) = F(b). By Rolles Theorem (Theorem 12.5.2), there is c (a, b)
such that F
(c) = 0.
Note that Theorem 12.5.3 follows from the above one by taking g(x) = x.
Lemma 14.5.2 If f is dierentiable then f
satises the Mean Value Theorem,

i.e. if a, b U, f
(a) < < f
(b), then f
(c) = for some c (a, b).

Proof: Since f
(a) < < f
(b), there is an h > 0 such that

f(a +h) f(a)
h
< <
f(b +h) f(b)
h
.
Fix such an h > 0. The function x
f(x+h)f(x)
h
is continuous on (a, b). By
the Intermediate Value Theorem (Theorem 11.3.3) there is an x = x(h) such
that
f(x+h)f(x)
h
= . Also, by Lemma 12.5.3, there is a c (x, x+h) such that
f(x+h)f(x)
h
= f
(c). Thus = f
(c).
Theorem 14.5.3 If f
exists and is bounded on some interval I, then f is

uniformly continuous on I.
14.5.1 Lipschitz Condition
14.5.2 A Metric On R
n
Lemma 14.5.4 i. For i = 1, . . . , n, let (X
i
,
i
) be a metric space. Let X =
X
1
. . . X
n
. Let p > 1 be any real number. For x, y X let d
p
(x, y) =
(
n
i=1
i
(x
i
, y
i
)
p
)
1/p
. Then (X, d
p
ii. Let everything be as above. Set d
(x, y) = max
i
(x
i
, y
i
) : i = 1, . . . , n.
Then (X, d
Proof: i. Setting a
i
=
i
(x, y), b
i
=
i
(x, z), c
i
=
i
(y, z), it is enough to show
that if b
i
, c
i
0, then
(
n
i=1
(b
i
+c
i
)
p
)
1/p
(
n
i=1
b
p
i
)
1/p
+ (
n
i=1
c
p
i
)
1/p
.
Taking p-th powers, we need to show that
n
i=1
(b
i
+c
i
)
p
((
n
i=1
b
p
i
)
1/p
+ (
n
i=1
c
p
i
)
1/p
)
p
.
We rst compute the left hand side:
n
i=1
(b
i
+c
i
)
p
=
n
i=1
p
j=0
_
p
j
_
b
j
i
c
pj
i
=
n
i=1
b
p
i
+
n
i=1
c
p
i
+
p1
j=1
_
p
j
_
n
i=1
b
j
i
c
pj
i
.
We now compute the right hand side:
((
n
i=1
b
p
i
)
1/p
+ (
n
i=1
c
p
i
)
1/p
)
p
=
p
j=0
_
p
j
_
(
n
i=1
b
p
i
)
j/p
(
n
i=1
c
p
i
)
(pj)/p
=
=
n
i=1
b
p
i
+
n
i=1
c
p
i
+
p1
j=1
_
p
j
_
(
n
i=1
b
p
i
)
j/p
(
n
i=1
c
p
i
)
(pj)/p
.
Simplifying, we see that we have to show that
n
i=1
b
j
i
c
pj
i
(
n
i=1
b
p
i
)
j/p
(
n
i=1
c
p
i
)
(pj)/p
for all j = 1, . . . , p 1. Setting r = b
p
i
, s = c
p
i
, = j/p and = (p j)/p, we
see that it is enough to show that
n
i=1
r
i
s
i
(
n
i=1
r)
(
n
i=1
s)
if 0 r
i
, 0 s
i
, 0 < , < 1 and + = 1.
We may assume that some r
i
,= 0 and some s
j
,= 0, thus
n
i=1
r
i
and
n
i=1
r
i
are nonzero. Dividing both sides by (
n
i=1
r
i
)
n
i=1
r
i
)
, we see that we have

to show that
n
i=1
(
r
i
n
i=1
r
i
)
(
s
i
n
i=1
s
i
)
1.
By setting
u
i
=
r
i
n
i=1
r
i
and v
i
=
r
i
n
i=1
s
i
,
we see that we have to prove that
n
i=1
u
i
v
i
= 1
if 0 u
i
, 0 v
i
,

n
i=1
u
i
=
n
i=1
v
i
= 1, 0 < , < 1 and + = 1.
14.5. SUPPLEMENT 153
Set f() =

n
i=1
u
i
v
i
= 1. Note that f(0) = f(1) = 1. Then an easy
calculation shows that
f
() =
i = 1
n
(ln(u
i
)u
i
v
1
i
ln(v
i
)u
i
v
1
i
) =
i = 1
n
u
i
v
1
i
(ln(u
i
)ln(v
i
)),
and that
f
() =
n
i=1
(ln(u
i
) ln(v
i
))
2
u
i
v
1
i
).
Thus f
is concave up, hence f() f(0) = f(1) = 1 for all (0, 1).
Chapter 15
Riemann Integral
15.2 Fundamental Theorem of Calculus
15.3 How To Integrate?
15.3.1 Power Series
How to integrate polynomials and power series...
15.3.2 Trigonometric Functions
Exercises.
i. Let I
n
=
_
/2
0
sin
n
xdx. Show that
I
2n
=
1 3 5 . . . (2n 1)
2 4 6 . . . (2n)
2
and that
I
2n+1
=
2 4 6 . . . (2n)
1 3 5 . . . (2n + 1)
.
ii. [Wallis Formula] Let I
n
be as above. Show that (I
n
)
n
is a decreasing se-
quence. (Hint: (sin
n
x)
n
is a decreasing sequence of functions). Conclude
that
(2 4 6 . . . (2n))
2
(3 5 . . . (2n 1))
2
(2n + 1)
<

2
<
(2 4 6 . . . (2n 2))
2
(2n)
(3 5 . . . (2n 1))
2
.
Conclude that
2
= lim
n
(2 4 6 . . . (2n))
2
(3 5 . . . (2n 1))
2
2n
155
156 CHAPTER 15. RIEMANN INTEGRAL
and that
= lim
n
(2 4 6 . . . (2n))
4
n((2n)!)
2
= lim
n
2
4n
(n!)
4
n((2n)!)
2
.
The last equality is known as Wallis formula.
15.4 Integration of Complex Function
Dene
_
b
a
f(t, z)dt.
15.4.1 Functions Dened by Integration
Theorem 15.4.1 . Let f(z, t) be a function where z A C and t I, an
interval of R. Suppose
_
I
f(z, t)dt exists for all z A. Then z
_
I
f(z, t)dt is
a continuous function from A into C.
15.5 Applications
15.5.1 Application to Series
Theorem 15.5.1 (Integral Test, Cauchy) Let f be a nonincreasing positive
real valued function which is dened for x 1. Then the series

n=1
f(n)
_
1
f(x)dx exists. Also,
_

1
f(x)dx
n=1
f(n) f(1) +
_
n
1
f(x)dx
and
n=1
f(n) =
n
i=1
f(i) +R
n
where
_
n+1
f(x)dx R
n

_
n
f(x)dx.
Proof: For n < m we have,
_
m
n
f(x)dx f(n) +f(n + 1) +. . . +f(m1)
_
m1
n1
f(x)dx.
This proves the rst part. The second part follows by taking n = 1 and 2 and
sending m to innity in the above inequalities. The third part is easy as well.
Assume now the series diverges but that f(x) converges to 0 when x goes to
innity. Let m = m(n) n be any integer valued function. Then
0 f(n)+f(n+1)+. . . +f(m1)
_
m
n
f(x)dx
_
n
n1
f(x)dx
_
m
m1
f(x)dx,
15.5. APPLICATIONS 157
so that when n tends to innity, we get,
lim
n
_
f(n) +f(n + 1) +. . . +f(m1)
_
m
n
f(x)dx
_
= 0.
For example if we take m = 2n and f(x) = 1/x,
lim
n
_
1
n
+
1
1 +n
+. . . +
1
2n 1

_
2n
n
dx/x
_
= 0.
Since
_
2n
n
dx/x = ln(2n) ln(n) = ln(2), we have,
lim
n
_
1
n
+
1
1 +n
+. . . +
1
2n
_
= ln(2).
More generally, let
n
= f(1) +f(2) +. . . +f(n)
_
n
1
f(x)dx.
For m > n we have
m
=
_
m
n
f(x)dx (f(n + 1) +. . . +f(m)).
By the theorem this number is nonnegative (replace n and m by n+1 and m+1
respectively) and
[
n
m
[ =
n
m
=
_
m
n
f(x)dx (f(n) +. . . +f(m1)) +f(n) f(n).
Since lim
n
f(n) = 0, the double sequence (
m

n
)
m,n
converges to 0.
Therefore the sequence (
n
)
n
has a limit, say c
f
. Thus
lim
n
_
n
i=1
f(i)
_
n
1
f(x)dx
_
= c
f
.
In particular if f(x) = 1/x, we see that
lim
n
_
1 +
1
2
+. . . +. . . +
1
n
ln(n)
_
exists. This constant is called Eulers constant or Euler-Mascheroni con-
stant. Since the sequence is increasing, the Euler constant is a positive number.
We can now state the following result:
Theorem 15.5.2 Let f(x) be a positive real valued nonincreasing function de-
ned for x 1 and which converges to 0 as x goes to innity. Then
lim
n
_
n
i=1
f(i)
_
n
1
f(x)dx
_
exists. In particular
lim
n
_
1 +
1
2
+. . . +. . . +
1
n
ln(n)
_
= 0.57721 . . .
exists.
Corollary 15.5.3 The series

n=1
1
n
1+
converges if > 0 and diverges if
0.
Proof: Follows from Theorem 15.5.1.
Corollary 15.5.4 If lim
n
_
1 +
ln |u
n
|
ln n
_
< 0 then the series
n
u
n
converges
absolutely. If lim
n
_
1 +
ln |un|
ln n
_
< 0 then the series

n
u
n
converges abso-
lutely.
Proof: Follows from Corollary 15.5.3.
Theorem 15.5.5 (Raabe and Duhamel) If limn
_
un
u
n+1
1
_
> 1, then the
series

n
u
n
converges absolutely. If u
n
> 0 and limn
_
u
n
un+1
1
_
< 1 then
the series

n
u
n
diverges.
Proof: In the rst case, there is an > 0 such that n
_
u
n
u
n+1
1
_
> 1 +
eventually. Thus
un
u
n+1
> 1 +
1+
n
eventually. Let 0 < < . Then 1 +
1+
n
>
(1 + 1/n)
+1
eventually. Thus
u
n+1
u
n
<
_
n
n + 1
_
+1
=
1/(n + 1)
+1
1/n
+1
eventually. Since

n
1/n
1+
converges, the result follows from Corollary 7.2.6.
Assume now we are in the second case. Then n
_
vn
vn1
1
_
< 1 eventually.
Thus
vn1
v
n
>
1/(n+1)
1/n
eventually. Since
n
1/n diverges, the second part follows
from Theorem 7.2.6 as well.
Example. Consider the series
1 +z +
( 1)
2!
z
2
+. . .
( 1) . . . ( n + 1)
n!
z
n
+. . .
where C. The series converges for [z[ < 1 according to dAlembert (Corol-
lary 7.5.2). Suppose now [z[ = 1. If u
n
is the term of z
n
, then
u
n
u
n+1
n + 1
n
.
Thus when [z[ = 1, by Raabe and Duhamels Convergence Rule, the series
converges absolutely for '() > 0, because a simple computation shows that
n
_
u
n
un1
1
_
converges to 1 +'().
If R
<0
and x = 1 the series diverges as it can be checked as above.
Exercises.
i. Show that the Euler constant is < 1.
15.5.2 Stirling Formula
We will try to nd an approximation for n!.
Our rst try:
Lemma 15.5.6 lim
n
ln(n!)
nln nn
= lim
n
ln(n!)
nln
n
e
= 1.
Proof: Note that ln(n!) =
n
i=1
lni. By the integration method, we have
nlnn n < nlnn n + 1 =
_
n
1
lnxdx

n
i=1
lni
_
n+1
1
lnxdx = (n + 1) ln(n + 1) n;
furthermore the quotient of the rst term to the last term converges to 1:
nln nn
(n+1] ln(n+1)n
=
ln n1
(1+1/n) ln(n+1)1

ln n1
ln(n+1)1
=
11 ln n
ln(n+1)/ ln n1/ ln n

ln n
ln(n+1)

1/n
1/n+1
1.
This gives the result.
Theorem 15.5.7 lim
n
n!
(n/e)
n
2n
= 1.
We can obtain a better approximation:
Theorem 15.5.8 n! = (n/e)
n
2n(1+1/12n+f(n)/n
2
) where f is a function
that is bounded.
The proof will show that we can extend the approximation. But this will be
done by another method in the next subsection.
Question: Does lim
n
f(n) exist?
Proof: Consider the sequence
s
n
= ln(n!) n(lnn 1) = ln(n!) nln
n
e
.
For n > 1, we have,
u
n
:= s
n
s
n1
= 1 + (n 1) ln(1 1/n)
= 1 (n 1)(1/n + 1/2n
2
+ 1/3n
3
+. . .)
=

i=1
1
i(i+1)n
i
= 1/2n + 1/6n
2
+g(n)/n
3
where g(n) =
i=3
1/i(i+1)n
i3
is a positive valued function whose limit when
n goes to innity is nite (see Exercise i, page 138).
We now introduce
(1) S
n
:= s
n
1
2
lnn = ln(n!) n(ln n 1)
1
2
lnn
(See Exercise i, page 161). Then
v
n
:= S
n
S
n1
= u
n
+
1
2
ln
n1
n
= 1/2n + 1/6n
2
+g(n)/n
3
+
1
2
ln(1 1/n)
= 1/2n + 1/6n
2
+g(n)/n
3
1
2
(1/n + 1/2n
2
+ 1/3n
3
+. . .)
= (1/6n
2
1/4n
2
) +
g(n)(1/3+1/4n+1/5n
2
+...)/2
n
3
= 1/12n
2
+h(n)/n
3
where lim
n
h(n) is a constant. Setting v
1
= 1, we have
S
n
=
n
i=1
v
i
=
1
12
n
i=1
1/i
2
+
n
i=2
h(i)
i
3
.
Thus the sequence (S
n
)
n
, i.e. the series

i=1
v
i
has a limit say S. Then S
n
=
S R
n
where
R
n
=
i=n+1
v
i
=
1
12
i=n+1
1
i
2
+
i=n+1
h(i)
i
3
.
Now we have two remarks: First we show that n
2
i=n+1
h(i)
i
3
converges to a
nite number when n goes to innity. Since (h(i))
i
is bounded, we may suppose
h(i) = 1. By the integral test, we have,
n
2
i=n+1
1
i
3
= n
2
i=1
1
(n +i)
3
< n
2
_

2
1
(n +x)
3
dx =
n
2
2(n + 2)
2
,
which shows what we want. Thus we may replace

i=n+1
h(i)
i
3
in the formula
above by t(n)/n
2
where t(n) is a function with a nite limit:
(2) R
n
=
i=n+1
v
i
=
1
12
i=n+1
1
i
2
+
t(n)
n
2
.
Here is our second remark:
1
i

1
i + 1
=
1
i(i + 1)
<
1
i
2
<
1
i(i 1)
=
1
i 1

1
i
,
so that
(3)
1
n + 1
=
i=n+1
_
1
i

1
i + 1
_
<
i=n+1
1
i
2
<
i=n+1
_
1
i 1

1
i
_
=
1
n
.
Let
k(n) =
n
2
12
_
1
n

i=n+1
1
i
2
_
.
By (3), 0 < k(n) <
n
2
12
(
1
n

1
n+1
) =
1
12(1+1/n)
< 1. Thus k(n) remains bounded.
(Question: Does it have a limit?) Note that
(4)
i=n+1
1
i
2
=
1
n

12k(n)
n
2
.
Now by (1), (2) and (4),
ln(n!)
(1)
= S
n
+nln
_
n
e
_
+
1
2
lnn = S R
n
+nln
_
n
e
_
+
1
2
lnn
(2)
= S +
1
12
i=n+1
1/i
2
t(n)
n
2
+nln
_
n
e
_
+
1
2
lnn
(4)
= S +
1
12n

k(n)
n
2

t(n)
n
2
+nln
_
n
e
_
+
1
2
lnn.
By exponentiating and using e
x
= 1 +x +x
2
/2! +. . ., we get,
n! = e
S
_
n
e
_
n
n
_
1 +
1
12n
+
f(n)
n
2
_
for some f(n) that remains bounded when n goes to innity. It remains to
evaluate A := e
S
and we use Wallis formula for this. We replace n! by the
formula found above in Wallis formula, which we recall from page 155:
= lim
n
2
4n
(n!)
4
n((2n)!)
2
.
We easily nd A =
2. Thus
n! =
2n
_
n
e
_
n
_
1 +
1
12n
+
f(n)
n
2
_
.
Exercises.
i. Show that s
n
1
2
lnn can be made as close to a constant. (Solution: We
continue with the notation of the proof of the theorem above. Note that
s
n
1 = s
n
s
1
=
n
i=2
u
i
=
n
i=2
(1/2i+1/6i
2
+g(i)/i
3
) =
1
2
n
i=2
1/i+
1
6
n
i=2
1/i
2
+
n
i=2
g(i)/i
3
. Since g(i) is bounded and since

i=2
1/i
2
and

i=2
1/i
3
converge, for n large enough
1
6
n
i=2
1/i
2
+
n
i=2
g(i)/i
3
can be made as close to a constant as we wish to. On the other hand
n
i=2
1/i =

n
i=1
1/i 1, can be made as close to lnn 1 as we wish
to (Theorem 15.5.2). Thus s
n
can be made close to A +
1
2
lnn for some
constant A.)
15.5.3 Eulers Function
Consider the integral
(z) =
_

0
e
t
t
z1
dt
where z C (but the reader may opt to take z R) and, for 0 < a < b, the
function
F(z) = F
a,b
(z) =
_
b
a
e
t
t
z1
dt,
which is a continuous function of z by Theorem 15.4.1.
Fixing a > 0 and regarding this as a family of functions parametrized by
b > a, we claim that for z varying over a set whose real part is bounded above,
F(z) converges uniformly when b . We have to show that for all > 0 there
is b
0
such that for all b, b
> b
0
and all z in the right domain,
_
b
b
e
t
t
z1
dt
< .
Indeed, let > 0 and A be the upper bound for the real part of z. Let b
0
be
such that t
A1
< e
t/2
for t > b
0
and e
b
0
/2
< /4. Then we have
[e
t
t
z1
[ = e
t
t
x1
[t
iy
[ = e
t
t
x1
[e
iy ln t
[ = e
t
t
x1
e
t
t
A1
< e
t/2
,
so that for b, b
> b
0
,
_
b
b
e
t
t
z1
dt

_
b
e
t
t
z1
dt <
_
b
b
e
t/2
dt
= 2e
b
/2
+ 2e
b/2
< 4e
b
0
/2
< .
It follows that
_

a
e
t
t
z1
dt
is a continuous function of z for a > 0.
Now we x b and view F(z) as a family of functions parametrized by 0 <
a < b. We claim that the family F(z) converges uniformly when a 0 on any
set in which the real part of z is bounded below. For this we have to show that
for any > 0 there is a > 0 such that for all 0 < a < a
< and for all z C

in the right domain (the real part above some xed A > 0),
_
a
a
e
t
t
z1
dt
< .
Indeed let > 0. Let A be the lower bound for the real part of z. For t [0, 1),
we have
[e
t
t
z1
[ = e
t
t
x1
< t
x1
t
A1
.
Let < 1 be such that
A
< A/2. Then for all 0 < a < a
< < 1,
_
a
a
e
t
t
z1
dt

_
a
e
t
t
z1
dt
_
a
a
t
A1
dt
= a
A
/Aa
A
/A < 2a
A
/A < 2
A
/A < .
Thus
_
b
0
e
t
t
z1
dt
is a continuous function of z for a > 0.
It follows from above that (z) is well-dened and is a continuous function
of z on the domain z C : '(z) > 0.
By integrating by parts, we have (du = e
t
dt, v = t
z
),
(z + 1) =
_
0
e
t
t
z
dt = [e
t
]
0

_
0
(e
t
)zt
z1
dt
= [1 + 1] +z
_
0
e
t
t
z1
dt = z(z).
We deduce that for n N 0,
(z +n) = (z +n 1)(z +n 2) z(z).
On the other hand (1) =
_
0
e
t
dt = [e
t
]
0
= 1. Thus, setting z = 1 above
we get,
(n + 1) = n!
We proved the following:
Theorem 15.5.9 The function
(z) =
_

0
e
t
t
z1
dt
where z C is well-dened and continuous for '(z) > 0. Furthermore
(z +n) = (z +n 1)(z +n 2) z(z)
and (n + 1) = n!
The function is called Eulers Gamma function. We will soon extend
its domain of denition from '(z) R
>0
to '(z) R (N). For this purpose
we write
(z) =
_
1
0
e
t
t
z1
dt +
_

1
e
t
t
z1
dt.
As we have seen above the second part is a continuous function of z for all
z C. We deal with the rst one. Let us replace e
t
by its Taylor series
n=0
(1)
n
t
n
/n!. This series converges uniformly for t [0, 1], thus also when
we multiply it with t
z1
, hence we can integrate it term per term to get:
_
1
0
e
t
t
z1
dt =
_
1
0
(
n=0
(1)
n
t
n+z1
/n!)dt =
n=0
[
_
1
0
(1)
n
t
n+z1
/n!]dt
=

n=0
[(1)
n
t
n+z
/n!(n +z)]
1
0
=

n=0
(1)
n
/n!(n +z)
Such a series converges uniformly on every compact subset ( closed bounded
subset) not containing the nonpositive integers according to Weierstrass M-test
(Theorem 9.3.3), because then [n +z[ > > 0 for some and so [(1)
n
/n!(n +
z)[ = 1/n![n +z[ < 1/n! and

n=0
1/n! converges. Thus
(z) =
n=0
(1)
n
/n!(n +z) +
_

1
e
t
t
z1
dt.
But this new expression is well-dened and continuous for all z C (N). We
still call it (z).
Now we check that the functional equality
(z +n) = (z +n 1)(z +n 2) z(z)
holds for the new denition for all z C (N). It is enough to show it for
n = 1. Indeed,
(z + 1) =

n=0
(1)
n
n!(n+z+1)
+
_
1
e
t
t
z
dt
=

n=0
(1)
n
((n+z+1)z)
(n+1)!(n+z+1)
+
_
1
e
t
t
z
dt
=

n=0
(1)
n
((n+z+1)z)
(n+1)!(n+z+1)
+ [e
t
t
z
]
1
+z
_
1
e
t
t
z1
dt
=

n=0
(1)
n
(n+1)!
+z
n=1
(1)
n
n!(n+z)
+e
1
+z
_
1
e
t
t
z1
dt
= z
n=1
(1)
n
n!(n+z)
+z
_
1
e
t
t
z1
dt = z(z).
Chapter 16
Supplements
16.1 Stone-Weierstrass Theorem
Theorem 16.1.1 Let K be a compact metric space. Let A C(K) be a subal-
gebra. Suppose
1) A separates points.
2) A contains the constant functions.
Then A is dense in C(K) with respect to the uniform norm.
Ex: Find maximal ideals of C([a, b]). Answer: Only the expected ones.
165
166 CHAPTER 16. SUPPLEMENTS
Chapter 17
Fourier Series
17.1 Hilbert Spaces
Let H be a real or complex vector space with a scalar product ( , ). Then
[x[ =
_
(x, x) denes a normed vector space. Suppose H is separable (has
a countable dense subset) and complete. H is called a separable Hilbert
space. Choose a countable dense subset, then a countable maximal linearly
independent subset, then orthogonalize this subset by Gram-Schmidt.
Theorem 17.1.1 Let H be a Hilbert space and M H be a closed subspace.
Then there is a unique y M such that [x y[ = d(x, M).
Theorem 17.1.2 Any Hilbert space is isomorphic to
2
.
17.2 Fourier Series
Theorem 17.2.1 The function
d
2
(f, g) = [f g[
2
=
1
2
_

[f(x) g(x)[
2
dx
denes a pseudometric on the set of integrable functions on the interval [, ].
Theorem 17.2.2 Let f be an integrable function on [, ]. Let
n
(x) =
exp(inx) and
c
n
=
1
2
_

f(x) exp(inx)dx = f,
n
).
Let s
n
(f; x) =
N
N
c
n
n
(x). Then
lim
N
_

[f(x) s
N
(f; x)[
2
dx
_
= 0.
167
168 CHAPTER 17. FOURIER SERIES
Furthermore
m
,
n
) =
n,m
for all n, m Z
The L
2
-norm: [f[
2
=
_
[f(x)[
2
dx.
Lemma 17.2.3 f bounded integrable. > 0. Then there is a polygonal function
p on [, ] such [f p[
2
< .
Proof: Trivial.
Theorem 17.2.4 Let f and g be two integrable functions on [, pi]. Let
f(x)
c
n
n
and g
d
n
n
be its Fourier series. Then
f, g) =
n=infty
c
n
d
n
and
[f[
2
=
n=infty
[c
n
[
2
.
Lemma 17.2.5 If f is a bounded function on [, ] which is integrable then
for all > 0 there is an h C([, ]) such that [f h[
2
< .
R = Integrable functions.
C([, ]) = Continuous functions.
F = Finite linear combinations of
n
=
n
: n Z).
s
N
(f; x) F
N
=
n
: N n N).
L
2
= f :
_
[f[
2
dx < .
It t
N
F
N
then [f s
N
[ [f t
N
[.
Chapter 18
Topological Spaces
(continued)
18.1 Product Topology
For each i in an index set I, let X
i
be a topological space. Let X =

iI
X
i
.
Consider the smallest topology on X that makes all the projection maps
i
:
X X
i
continuous. This is called the product topology.
Exercises.
i. The open sets
x
iI
X
i
: x
i
1
U
i
1
, . . . , x
i
n
U
i
n
(n N and U
i
j
X
i
j
open) form a base of the topology.
ii. If the topological spaces X
i
are discrete, then the open sets
x
iI
X
i
: x
i
1
= a
i
1
, . . . , x
i
n
= a
i
n
(n N and a
i
j
X
i
j
) form a base of the product topology.
iii. If I = N and the topological spaces X
i
are discrete, then the open sets
x
iI
X
i
: x
1
= a
1
, . . . , x
n
= a
n
(n N and a
i
X
i
iv. If X
i
= R (with the usual topology), then the open sets
x
iI
X
i
: x
i
1
(a
i
1
, b
i
1
), . . . , x
i
n
(a
i
n
, b
i
1
)
169
170 CHAPTER 18. TOPOLOGICAL SPACES (CONTINUED)
(n N and a
i
j
, b
i
j
X
i
j
v. Assume I is nite and each X
i
is a metric space. Show that the product
topology on

I
X
i
is given by any of the product metrics.
vi. Suppose I = = N and X
i
= N has the discrete topology. Then, viewing
X
i
as the set of function from N into N, we have:
a.

X
i
is metrisable by the following metric d(x, y) = 1/(n + 1) if n is
the smallest integer for which x
n
,= y
n
.
b. The set of injective functions from into is a closed subset of
X
i
.
c. Neither the set of surjective functions nor Sym() is a closed subset of
X
i
.
Problem 18.1.1 What is the closure of Sym() in

?
Problem 18.1.2 Let Sym
f
() be the set of permutations of Sym() that move
only nitely many elements of . Clearly Sym
f
() is a subgroup of Sym().
What is its closure in Sym()?
Hausdor Spaces. Let X be a topological space. If for any two distinct
points x and y of X there are disjoint open sets U
x
and U
y
containing x and y,
then we say that X is Hausdor.
The coarsest topology on a set with at least two points is not Hausdor.
Proposition 18.1.3 A metric space is Hausdor.
Exercises.
i. Assume that each X
i
is a Hausdor space. Show that the product topology
iI
X
i
is Hausdor.
18.2 Homeomorphisms
The notion of isomorphism between topological spaces is dened as follows: Two
topological spaces X and Y are called homeomorphic (i.e. isomorphic) if there
is continuous bijection f : X Y whose inverse is also continuous. Such a
map is called a homeomorphism.
18.3 Sequences in Topological Spaces
Let X be a set, (x
n
)
n
a sequence in X and x X. We say that x is a limit
of the sequence (x
n
)
n
if for any open subset U containing x, there is a natural
number N such that x
n
U whenever n > N.
Lemma 18.3.1 Let A X be closed, (a
n
)
n
a sequence from A and x a limit
of (a
n
)
n
. Then x A.
18.4. SEQUENTIAL COMPACTNESS 171
Examples. 1. If the sequence is eventually constant, i.e. if there is an x
and a natural number N such that x
n
= x for n > N, then x is a limit of the
sequence (x
n
)
n
.
2. Let X have the coarsest topology. Then any point of X is a limit of any
sequence of X.
3. Let X have the discrete topology. Then a sequence has a limit if and only
if the sequence is eventually constant, i.e. if there is an x and a natural number
N such that x
n
= x for n > N.
Proposition 18.3.2 In a Hausdor topological space a sequence has at most
one limit.
In this case we say that (x
n
)
n
is a converging sequence and we write lim
n
x
n
=
x.
Proposition 18.3.3 Let X and Y be two topological spaces. Let f : X Y .
Show that f is continuous if and only if for any convergent sequence (x
n
)
n
of
X, (f(x
n
))
n
is convergent and f(lim
n
x
n
) = f(lim
n
f(x
n
)).
18.4 Sequential Compactness
Theorem 18.4.1 A metric space is compact if and only if it is sequentially
compact.
The converse fails:
Example.
18.5 Supplements
18.5.1 T
0
-Identication
A topological space X is called T
0
if for any distinct x, y X there is an open
subset that contains only one of the two points.
Lemma 18.5.1 A topological space X id T
0
if and only if for any two distinct
x and y in X, x ,= x.
In a topological space dene x y if and only if x = x. This is
an equivalence relation. On X/ put the largest topology that makes the
projection map X X/ continuous.
Proposition 18.5.2 X/ is a T
0
-topological space.
Proof: HW.
Problem 1. Find the universal property of X/ that characterizes it.
Problem 2. Try to do the same with T
1
-spaces.
172 CHAPTER 18. TOPOLOGICAL SPACES (CONTINUED)
Chapter 19
Exams
19.1 Midterm Math 121 (November 2002)
i. Which of the following are not vector spaces over R (with the componen-
twise addition and scalar multiplication) and why?
V
1
= (x, y, z) R
3
: xy 0
V
2
= (x, y, z) R
3
: 3x 2y +z = 0
V
3
= (x, y, z) R
3
: xyz Q
V
4
= (x, y) R
3
: x +y 0
V
5
= (x, y) R
2
: x
2
+y
2
= 0
V
6
= (x, y) C
2
: x
2
+y
2
= 0
(2+2+2+2+2+5 pts.)
Answers. V
1
is not a vector space because e.g. (1, 0, 0) V
1
, (0, 1, 0)
V
2
but their sum (1, 1, 0) , V
1
.
V
2
is a vector space.
V
3
3
, but

2(1, 1, 1) , V
3
.
V
4
4
, but (1, 2, 1) , V
4
.
V
5
is a vector space because V
5
= (0, 0, 0).
V
6
is not a vector space because e.g. (1, i) V
6
, (1, i) V
6
, but their
sum (1, 0) , V
6
.
ii. On the set X = 2, 3 . . . , 100 dene the relation x y by x ,= y and x
divides y.
a) Show that this denes a partial order on X. (3 pts.)
b) Is this a linear order? (2 pts.)
c) Find all the maximal and minimal elements of this poset. (5 pts.)
Answers. a) Yes, this is a partial order: Clearly x , x for any x. Since
division is transitive, is transitive as well. (The details are left).
173
174 CHAPTER 19. EXAMS
b) No, because 2 and 3 are not comparable.
c) The prime numbers are minimal elements. The maximal elements are
the numbers which are greater than 50. For example 53 is both minimal
and maximal.
iii. On R R dene the relation as follows (x, y) (x
1
, y
1
) by either
y < y
1
, or y = y
1
and x < x
1
.
a) Show that this is a linear order. (5 pts.)
least upper bound? (5 pts.)
Answers. a) Yes!
b) No. For example the set R0 is bounded above by (0, 1) but it does
not have a least upper bound.
iv. For each n N, let a
n
and b
n
be two real numbers. Assume that for each
n, a
n
a
n+1
b
n+1
b
n
. Show that
nN
[a
n
, b
n
] = [a, b] for some real
numbers a and b. (10 pts.)
Proof: Since the set a
n
: n N is bounded above by b
0
, it has a least
upper bound, say a. Similarly the set b
n
: n N has a greatest lower
bound, say b. I claim that
nN
[a
n
, b
n
] = [a, b].
If x a, then x a
n
for all n. Likewise, if x b, then x b
n
for all n.
Hence, if x [a, b], the x [a
n
, b
n
] for all n.
Conversely, let x
nN
[a
n
, b
n
]. Then a
n
x b
n
for all n. Thus x is
an upper bound for a
n
: n N and a lower bound for b
n
: n N.
Hence a x b.
v. Show that for any natural number n and for any real number x [0, 1),
(1 x)
n
1 nx +
n(n 1)
2
x
2
.
(10 pts.)
Proof: We proceed by induction on n.
If n = 0, then both sides are equal to 1.
Suppose we know the result for n, i.e. suppose we know that for any real
number x [0, 1),
(1 x)
n
1 nx +
n(n 1)
2
x
2
.
We will prove that for any real number x [0, 1),
(1 x)
n+1
1 (n + 1)x +
(n + 1)n
2
x
2
.
19.1. MIDTERM MATH 121 (NOVEMBER 2002) 175
Let x [0, 1). Since (1 x)
n
1 nx +
n(n1)
2
x
2
and since 1 x > 0,
multiplying by 1 x both sides we get (1 x)
n+1
= (1 x)
n
(1 x)
(1 nx +
n(n1)
2
x
2
)(1 x) = 1 (n + 1)x +
(n+1)n
2
x
2
n(n1)
2
x
3
. Thus
(1x)
n+1
1(n+1)x+
(n+1)n
2
x
2
n(n1)
2
x
3
. Since x 0, (1x)
n+1
1 (n + 1)x +
(n+1)n
2
x
2
n(n1)
2
x
3
1 (n + 1)x +
(n+1)n
2
x
2
.
vi. a) Show that for any complex number there is a polynomial of the form
p(X) = X
2
+ aX + b R[X] such that p() = 0. (Note: a and b should
be real numbers). (10 pts.)
b) What can you say about a and b if = u + iv for some u, v Z? (5
pts.)
Proof: a) Let be a complex number. Then p(x) := (x )(x ) =
x
2
( +)x + R[x] and it is easy to check that p() = 0.
b) It is clear that if = u +iv for some u, v Z, then p(x) Z[x].
Second Proof of part a. Write = u + iv where u and v are real
numbers. Then
2
= u
2
v
2
+2uvi. Thus
2
2u = (u
2
v
2
+2uvi)
2u(u + iv) = u
2
v
2
, so that
2
2u + (u
2
+ b
2
) = 0. Hence is a
root of the polynomial p(x) = x
2
2ux + (u
2
+b
2
) R[x].
Part b follows from this immediately.
vii. a) Show that for any C there is a C such that
2
= . (15 pts.)
b) Show that for any , C there is an x C such that x
2
+x+ = 0.
(10 pts.)
Proof: a. Let = a + bi. We try to nd C such that
2
= , i.e.
we try to nd two real numbers x and y such that (x +iy)
2
= a +bi. We
may assume that ,= 0 (otherwise take = 0). Thus a and b cannot be
both 0. After multiplying out, we see that this equation is equivalent to
the system
x
2
y
2
= a
2xy = b
Since x = 0 implies a = 0 = y = b, x must be nonzero. Thus we have
y = b/2x and so the above system is equivalent to the following:
x
2
(b/2x)
2
= a
y = b/2x
Equalizing the denominators in the rst one, we get the following equiva-
lent system:
4x
4
4ax
2
b
2
= 0
y = b/2x
So now the problem is about the solvability of the rst equation 4x
4
4ax
2
b
2
= 0. (Once we nd x, which is necessarily nonzero, we set
y = b/2x). Setting z = x
2
, we see that the solvability of 4x
4
4ax
2
b
2
= 0
is equivalent to the question of whether 4z
2
4azb
2
= 0 has a nonnegative
solution. Since the last one is a quadratic equation over R, it is easy to
answer this question. There are two possible solutions: z = a
a
2
+b
2
and one of them z = a +
a
2
+b
2
is nonnegative (even if a is negative).
Thus we can take
x =
_
a +
_
a
2
+b
2
and
y = b/2x.
b. We rst compute as follows: 0 = x
2
+x + = x
2
+x +
2
/4 +(
2
/4) = (x+/2)
2
+(
2
/4). Thus if z C is such that z
2
=
2
/4
(by part a there is such a z), then x := z /2 is a root of x
2
+x +.
viii. Suppose X and Y are two subsets of R that have least upper bounds.
Show that the set X + Y := x + y : x X, y Y has a least upper
bound and that sup(X +Y ) = sup(X) + sup(Y ). (15 pts.)
Proof: Let a and b be the least upper bounds of X and Y respectively.
Thus x a for all x X and y b for all y Y . It follows that
x + y a + b for all x X and y Y , meaning exactly that a + b is an
upper bound of X +Y . Now we show that a +b is the least upper bound
of X + Y . Let > 0 be any. We need to show that a + b < x + y
for some x X and y Y . Since a is the least upper bound of X, there
is an x X such that a /2 < x. Similarly there is a y Y such that
b /2 < y. Summing these two, we get a +b < x +y.
ix. We consider the subset X = 1/2
n
: n N 0 of R together with the
usual metric, i.e. for x, y X, d(x, y) is dened to be [x y[. Show that
the open subsets of X are the conite subsets
1
of X and the ones that do
not contain 0. (20 pts.)
Proof: We rst show that the singleton set 1/2
n
is an open subset of X.
This is clear because B(1/2
n
, 1/2
n+1
) = 1/2
n
. It follows that any subset
of X that does not contain 0 is open. Now let U be any conite subset of
X. We proceed to show that U is open. If 0 , U, then we are done by the
preceding. Assume 0 U. Since U is conite, there is a natural number
n
such that for all n n
, 1/2
n
U, i.e. B(0, 1/2
n
) U. Now U is
the union of B(0, 1/2
n
) and of a nite subset not containing 0. Thus U

is open.
For the converse, we rst show that a nonempty open ball is of the form
described in the statement of the question. If the center of the ball is 0,
then the ball is conite by the Archimedean property. If the center of the
ball is not 0, then either the ball does not contain 0 or else it does contain
0, in which case the ball must be conite.
To nish the proof, we must show that an arbitrary union of open balls
each of which does not contain 0 cannot contain 0. But this is clear!
1
A subset Y of X is called conite if X \ Y is nite.
19.2. FINAL MATH 121 (JANUARY 2003) 177
19.2 Final Math 121 (January 2003)
Justify all your answers. A nonjustied answer will not receive any grade what-
soever, even if the answer is correct. DO NOT use symbols such as , , .
Make full sentences with correct punctuation. You may write in Turkish or in
English.
i. Let (a
n
)
n
be a convergent sequence of real numbers.
a. Does the sequence (a
2n
)
n
converge necessarily? (2 pts.)
Since (a
2n
)
n
is a subsequence of the converging sequence (a
n
)
n
, both se-
quences converge to the same limit.
Remarks. We have seen in class that a subsequence of a converging
sequence converges.
Contrary to what some of you think, a
2n
,= 2a
n
!
b. Assume a
n
,= 0 for all n. Does the sequence (a
n
/a
n+1
)
n
converge
necessarily? (2 pts.)
No, the sequence (a
n
/a
n+1
)
n
n
a
n
= 0. For
example, choose
a
n
=
_
1/n if n is even
1/n
2
if n is odd
Clearly lim
n
a
n
= 0, but
a
n
a
n+1
=
_
_
_
(n+1)
2
n
if n is even
n+1
n
2
if n is odd
And the subsequence
(n+1)
2
n
n+1
n
2
converges to 0.
On the other hand, if the limit of the sequence (a
n
)
n
is nonzero, say ,
then the sequence (a
n
/a
n+1
)
n
converges to 1 because lim
n
a
n
/a
n+1
=
lim
n
a
n
/ lim
n
a
n+1
= / = 1. Note that the last part uses the fact
that is nonzero.
ii. Find the following limits and prove your result using only the denition.
(30 pts.)
a. lim
n
2n5
5n+2
.
We claim that the limit is 2/5. Let > 0. Since R is Archimedean, there
is an N such that 2 < N. Now for n > N, [
2n5
5n+2

2
5
[ =
29
5(5n+2)
<
29
25n
<
2/n < 2/N < .
b. lim
n
2n
2
5
5n+2
.
We claim that the limit is . For this it is enough to prove that
lim
n
2n
2
5
5n2
= . Let A be any real number. Let N = max(5A, 2).
For n > N we have
2n
2
5
5n2
>
2n
2
5
5n
>
2n
2
n
2
5n
=
n
2
5n
=
n
5
>
N
5
A. This
proves that lim
n
2n
2
5
5n2
= .
c. lim
n
2n
2
5
n
3
+2
.
We claim that the limit is 0. Let > 0. Let N
1
be such that 2 > N
1
(Archimedean property of R). Let N = max(1, N). Then for all n > N,
2n
2
5
n
3
+ 2
=
2n
2
5
n
3
+ 2
<
2n
2
n
3
+ 2

2n
2
n
3
=
2
n
<
2
N
< .
Note that the rst equality is valid because n
3
5 0, the second in-
equality is valid because n
3
+ 2 > 0.
iii. Find (16 pts. Justify your answers).
a. lim
n
(1/2 + 1/n)
n
.
Note that for n 3, 0 < 1/2 +1/n 1/2 +1/3 = 5/6. Thus the sequence
((1/2+1/n)
n
)
n
is eventually squeezed between the zero constant sequence
and the sequence ((5/6)
n
)
n
. Since lim
n
(5/6)
n
= 0 (because 5/6 < 1),
lim
n
(1/2 + 1/n)
n
= 0.
b. lim
n
(3/2 7/n)
n
.
Since 3/2 > 1 and lim
n
7/n = 0, there is an N such that 7/N <
1/2 = 3/2 1. In fact, it is enough to take N = 15. Then for all n N,
3/27/n 3/27/N > 1 and so (3/27/n)
n
(3/27/N)
n
. Therefore
the sequence ((3/27/n)
n
)
n
is greater than the sequence ((3/27/N)
n
)
n
.
Since 3/2 7/N > 1, the sequence ((3/2 7/N)
n
)
n
diverge to . Hence
lim
n
(3/2 7/n)
n
= .
iv. Find lim
n
_
n
2
1
n
3
n5
_
n
2
1
2n3
. (10 pts. Justify your answer).
Assume n > 5. Then we have, n
3
n5 > n
3
2n > n
3
n
3
/2 = n
3
/2 > 0.
Therefore,
n
2
1
n
3
n 5
=
n
2
1
n
3
n 5
<
n
2
n
3
n 5
<
n
2
n
3
/2
= 2/n.
Also
n
2
1
2n3
>
n
2
1
2n
>
n
2
n
2n
=
n1
2
> 2. Hence
_
n
2
1
n
3
n 5
_
n
2
1
2n3
< (2/n)
2
.
It follows that lim
n
_
n
2
1
n
3
n5
_
n
2
1
2n3
= 0.
v. Show that the series

n=1
(1/n)
n
converges. Find an upper bound for the
sum. (10 pts.)
Since for n 2, 1/n 1/2, we have

n=1
(1/n)
n
1 +
n=2
(1/n)
n
1 +
n=1
(1/2)
n
= 1 +
1
2
n=0
(1/2)
n
= 1 + 1/2 = 3/2.
19.3. RESIT OF MATH 121, FEBRUARY 2003 179
vi. Let (a
n
)
n
be a sequence of nonnegative real numbers. Suppose that the
sequence (a
2
n
)
n
converges to a. Show that the sequence (a
n
)
n
converges to
a. (15 pts.)
Note rst that, since a
n
0, a 0 as well.
Let > 0.
Case 1: a > 0. Since lim
n
a
2
n
= a, there is an N
2
such that for all
n > N, [a
2
n
a[ < a. Now for all n > N, [a
n
a[ =
|a
2
n
a|
a
n
+a

|a
2
n
a|
a
< .
Case 2. a = 0.
Since the sequence (a
2
n
)
n
converges to 0, there is an N such that for all
n > N, a
2
n
<
2
. So ( + a
n
)( a
n
) =
2
a
2
n
> 0. Since > 0 and
a
n
0, we can divide both sides by + a
n
to get a
n
> 0, i.e. a
n
< .
Since a
n
0, this implies [a
n
[ < .
vii. We have seen in class that the sequence given by a
n
= ((1 + 1/n)
n
)
n
converges to a real number > 1. Let e be this limit. Do the following
sequences converge? If so nd their limit. (15 pts. Justify your answers).
a) lim
n
_
1 +
1
n+1
_
n
.
lim
n
_
1 +
1
n+1
_
n
= lim
n
(1+
1
n+1
)
n+1
1+
1
n+1
=
lim
n(1+
1
n+1
)
n+1
lim
n
1+
1
n+1
=
e
1
=
e. The rst equality is algebraic. The second equality holds because the
limits of the numerator and the denominator exist and they are nonzero.
The third equality holds because
_
_
1 +
1
n+1
_
n+1
_
n
is a subsequence of
__
1 +
1
n
_
n
_
n
.
b) lim
n
_
1 +
1
2n
_
3n
.
lim
n
_
1 +
1
2n
_
3n
= lim
n
_
_
1 +
1
2n
_
2n
_
3/2
= e
3/2
by Question vi.
c) lim
n
_
1 +
1
n
_
n
2
.
Since lim
n
_
1 +
1
n
_
n
= e > 1, there is an N
1
such that for all n > N
1
,
[
_
1 +
1
n
_
n
e[ <
e1
2
, so 1 < r :=
e
2
+
1
2
= e
e1
2
<
_
1 +
1
n
_
n
. Thus,
r
n
<
__
1 +
1
n
_
n
_
n
=
_
1 +
1
n
_
n
2
. Thus = lim
r
r
n
_
1 +
1
n
_
n
2
(because r > 1). It follows that
_
1 +
1
n
_
n
2
= .
19.3 Resit of Math 121, February 2003
i. Find a sequence neither decreasing nor increasing that converges to 1. (2
pts.)
ii. Let (a
n
)
n
be a convergent sequence of real numbers. Suppose that a
n
Z
for all n. Is it true that lim
n
a
n
Z? (4 pts.)
iii. Let (a
n
)
n
n
Q
n
a
n
Q? (3 pts.)
iv. Let (a
n
)
n
be a convergent sequence of real numbers. Suppose that 5a
n
/2
N for all n. What can you say about lim
n
a
n
? (2 pts.)
v. Let (a
n
)
n
be a sequence of real numbers such that the subsequence (a
2n
)
n
converges. Does the sequence (a
n
)
n
vi. Let (a
n
)
n
be a sequence of real numbers such that the sequence (a
2
n
)
n
converges to 1. Does the sequence (a
n
)
n
vii. Let (a
n
)
n
be a sequence of real numbers such that the subsequences (a
2n
)
n
and (a
2n+1
)
n
both converge. Does the sequence (a
n
)
n
converge necessar-
ily? (2 pts.)
viii. Let (a
n
)
n
2
n
)
n
n
)
n
ix. Let (a
n
)
n
be a sequence of real numbers such that lim
n
a
n
= . Is it
true that lim
n
a
2n
= ? (3 pts.)
x. Assume lim
n
a
n
exists and a
n
,= 0 for all n. Does the sequence
(a
2n
/a
2n+1
)
n
xi. Find the following limits and prove your result using only the denition.
(30 pts.)
a. lim
n
3n+105
5n79
b. lim
n
n
2
5n+3
100n+2
c. lim
n
n8
2n
3
89
xii. Find (16 pts. Justify your answers).
a. lim
n
_
2
3
+
6n
n
2
+1
_
3n
b. lim
n
_
5
4

7
n
5
_
n
n
xiii. Find lim
n
_
n
2
1
n
3
n5
_
n
2
1
2n3
xiv. Show that the series

n=1
_
n
n
2
+1
_
n/3
converges. Find an upper bound
for the sum. (10 pts.)
xv. Let (a
n
)
n
be a sequence of real numbers. Assume that there is an r > 1
such that [a
n+1
[ r[a
n
[ for all n. What can you say about the convergence
or the divergence of (a
n
)
n
? (6 pts.)
19.4. CORRECTION OF THE RESIT OF MATH 121, FEBRUARY 2003181
19.4 Correction of the Resit of Math 121, Febru-
ary 2003
i. Find a sequence neither decreasing nor increasing that converges to 1. (2
pts.)
Answer: Let a
n
= 1 +
(1)
n
n
. It is clear that lim
n
a
n
= 1. Since
the subsequence (a
2n
)
n
is decreasing and converges to 1 and the subse-
quence (a
2n
)
n
is increasing and converges to 1, the sequence (a
n
) is neither
increasing nor decreasing.
ii. Let (a
n
)
n
n
Z
n
a
n
Z? (4 pts.)
Answer: Yes, it is true. In fact this is true even for Cauchy sequences:
A Cauchy sequence (a
n
)
n
whose terms are in Z is eventually constant, i.e.
there is an N such that a
n
= a
N
for all n N, and this implies of course
that lim
n
a
n
= a
N
Z. So, let us show that the Cauchy sequence
(a
n
)
n
is eventually constant.
In the denition of Cauchy sequences, take = 1/2. Thus, there is an M
such that for all n, m > M, [a
n
a
m
[ < 1/2. But since a
n
and a
m
are in
Z, this means that for all n, m > M, [a
n
a
m
[ = 0, i.e. that a
n
= a
m
.
Now take N = M + 1.
iii. Let (q
n
)
n
be a convergent sequence of real numbers. Suppose that q
n
Q
n
q
n
Q? (3 pts.)
Answer: Of course not! In fact every real number is the limit of a rational
sequence. Indeed, let r R. Let n N 0. Since Q is dense in R, there
is a rational number q
n
(r 1/n, r). Since r 1/n < q
n
< r, by the
Sandwich Lemma, lim
n
q
n
= r.
iv. Let (a
n
)
n
be a convergent sequence of real numbers. Suppose that 5a
n
/2
N for all n. What can you say about lim
n
a
n
? (4 pts.)
Answer: Let lim
n
a
n
= r. Then lim
n
5a
n
/2 = 5r/2. By hypothe-
sis and by part ii, 5r/2 Z. Thus r = 2n/5 for some n N.
v. Let (a
n
)
n
be a sequence of real numbers such that the subsequence (a
2n
)
n
converges. Does the sequence (a
n
)
n
Answer: Of course not! We can have a
2n
= 1/n and a
2n+1
= n.
vi. Let (a
n
)
n
2
n
)
n
n
)
n
n
= (1)
n
. Then (a
n
)
n
is a
sequence of alternating ones and minus ones, so that it diverges. And
since a
2
n
= 1, the sequence (a
2
n
)
n
converges to 1.
vii. Let (a
n
)
n
2n
)
n
and (a
2n+1
)
n
both converge. Does the sequence (a
n
)
n
converge necessar-
ily? (2 pts.)
n
= (1)
n
. Then (a
n
)
n
is a
sequence of alternating ones and minus ones, so that it diverges. And
since a
2n
= 1 and a
2n+1
= 1, the sequence (a
2n
)
n
converges to 1 and
the sequence (a
2n+1
)
n
converges to 1.
viii. Let (a
n
)
n
2
n
)
n
n
)
n
Answer: Yes! Let > 0. Let =
. Since the sequence (a

2
n
)
n
converges
to 0, there is an N such that for all n > N, [a
2
n
[ < , i.e. [a
n
[
2
<
2
. Since
[a
n
[ and are positive, this implies that [a
n
[ < . Thus there is an N such
that for all n > N, [a
n
[ < ; i.e. the sequence (a
n
)
n
converges to 0.
ix. Let (a
n
)
n
n
a
n
= . Is it
true that lim
n
a
2n
= ? (3 pts.)
Answer: Yes! Let A be any real number. lim
n
a
n
= , there is an
N such that for all n > N, a
n
> A. Then for 2n > N, a
2n
> A.
x. Assume lim
n
a
n
exists and a
n
,= 0 for all n. Does the sequence
(a
2n
/a
2n+1
)
n
Answer: No, the sequence (a
2n
/a
2n+1
)
n
n
a
n
=
0. For example, choose
a
n
=
_
1/n if n is even
1/n
2
if n is odd
Clearly lim
n
a
n
= 0, but
a
n
a
n+1
=
_
_
_
(n+1)
2
n
if n is even
n+1
n
2
if n is odd
And the subsequence
(n+1)
2
n
n+1
n
2
converges to 0.
On the other hand, if the limit of the sequence (a
n
)
n
is nonzero, say , then
the sequence (a
2n
/a
2n+1
)
n
converges to 1 because lim
n
a
2n
/a
2n+1
=
lim
n
a
2n
/ lim
n
a
2n+1
= / = 1. Note that the last part uses the
fact that is nonzero.
xi. Find the following limits and prove your result using only the denition.
(30 pts.)
a. lim
n
3n+105
5n79
Answer: lim
n
3n+105
5n79
=
3
5
.
19.4. CORRECTION OF THE RESIT OF MATH 121, FEBRUARY 2003183
Proof: Let > 0. Let N
1
be such that 32 < N
1
. Let N = max(N
1
, 395).
Now for n > N, we have,
3n + 105
5n 79

3
5
762
25n 395
=
762
25n 395

762
24n
<
32
n
<
32
N
1
< .
The rst equality is simple computation. The second equality follows from
the fact n > N 395 > 16 (so that 25n 395 > 0). The third inequality
follows from the fact that n > N 395, so that 25n 395 25n n =
24n.The fourth inequality is also a simple computation.
b. lim
n
n
2
5n+3
100n+2
Answer: lim
n
n
2
5n+3
100n+2
= .
Proof: It is enough to show that lim
n
n
2
5n+3
100n2
= .
We rst note that the two roots of n
2
5n +3 are
5
2512
2
=
5
13
2
, so
that if n 5 >
9
2
=
5+
16
2
>
5+
13
2
, then n
2
5n + 3 > 0.
Now let A R be any real number. Let N = max(100A+ 5, 5). Now, for
all n > N,
n
2
5n + 3
100n 2
>
n
2
5n + 3
100n
>
n
2
5n
100n
=
n 5
100
>
N 5
100
= A.
Here, the rst inequality follows from the fact that n > N 5, so that
n
2
5n + 3 > 0.
c. lim
n
n8
2n
3
89
.
Answer: lim
n
n8
2n
3
89
= 0.
Proof: Let > 0. Let N = max(1/, 89). Now for n > N,
n8
2n
3
89
=
n8
2n
3
89
<
n
2n
3
n
=
1
2n
2
1
<
1
n
2
<
1
n
< .
xii. Find (16 pts. Justify your answers).
a. lim
n
_
2
3
+
6n
n
2
+1
_
3n
.
Answer: lim
n
_
2
3
+
6n
n
2
+1
_
3n
= 0.
Proof: We use the fact that 2/3 < 1. Since lim
n
6n
n
2
+1
= 0, there is
an N such that for all n > N,
6n
n
2
+1
< 1/6. Then 0
_
2
3
+
6n
n
2
+1
_
3n
=
(2/3+1/6)
3n
= (5/6)
3n
. By Sandwich Lemma lim
n
_
2
3
+
6n
n
2
+1
_
3n
= 0.
b. lim
n
_
5
4

7
n
5
_
n
n
.
Answer: lim
n
_
5
4

7
n
5
_
n
n
= .
Proof: We use the fact that 5/4 > 1. Since lim
n
7
n
5
= 0, there is an N
such that for all n > N,
7
n
5
< 1/8. Then
_
5
4

7
n
5
_
n
n
> (5/4 1/8)
n
n
=
(9/8)
n
n
(9/8)
n
. Since (9/8) > 1, lim
n
(9/8)
n
= . The result
follows.
xiii. Find lim
n
_
n
2
1
n
3
n5
_
n
2
1
2n3
. (10 pts.).
Answer: lim
n
_
n
2
1
n
3
n5
_
n
2
1
2n3
= 0.
Proof: Since lim
n
_
n
2
1
n
3
n5
_
= 0, there is an N
1
such that for all
n > N
1
,
n
2
1
n
3
n5
< 1/2. On the other hand, for n > 3,
n
2
1
2n3
<
n
2
1
n
< n.
Let N = max(3, N
1
). Now for n > N,
_
n
2
1
n
3
n5
_
n
2
1
2n3
< (1/2)
n
2
1
2n3
<
(1/2)
n
. Since the right hand side converges to 0, by Sandwich Lemma,
0 lim
n
_
n
2
1
n
3
n5
_
n
2
1
2n3
= 0. (For the rst inequality, one needs the
fact that n
3
n 5 > 0 for n 2. This follows from the facts that
2
3
2 5 = 1 > 0 and n
3
n 5 < (n +1)
3
(n +1) 5. And this last
inequality is easy to show).
xiv. Show that the series

n=1
_
n
n
2
+1
_
n/3
Answer:

n=1
_
n
n
2
+1
_
n/3
=
n=1
(1/n)
n/3
<
n=1
1/2
n/3
=
n=1
1/2
3n/3
+
n=0
1/2
3n+1
3
+
n=0
1/2
3n+2
3
=
n=1
1/2
n
+
1
2
1/3
n=0
1/2
n
+
1
2
1/3
n=0
1/2
n
= 1/2+2
1/3
+2
2/3
<
5.
xv. Let (a
n
)
n
such that [a
n+1
[ r[a
n
n
)
n
? (6 pts.)
Answer: The sequence diverges. Furthermore the sequence diverges to
if it is eventually positive and to if it is eventually negative.
Proof: One can show by induction on n that [a
n
[ > r
n
[a
0
[. Thus
lim
n
[a
n
[ = (because r > 1). It should now be clear that the an-
swer is valid.
19.5 Second Resit of Math 121, March 2003
i. Let (a
n
)
n
be a converging sequence of nonzero real numbers. Do the
sequences
_
a
n
an+1
_
n
,
_
1a
n
1an+1
_
n
and
_
a
n
1+a
2
n+1
_
n
converge? (6 pts.)
19.5. SECOND RESIT OF MATH 121, MARCH 2003 185
ii. Let (a
n
)
n
be a convergent sequence of natural numbers. Is it true that
lim
n
a
n
N? (4 pts.)
iii. Let x > 1. Discuss the convergence of (x
n
/n!)
n
(4 pts.)
iv. Let x R. Discuss the convergence of (x
n!
/n!)
n
(8 pts.)
v. Let (a
n
)
n
be a convergent sequence of real numbers. Suppose that lim
n
a
n

Q. Is it true that a
n
Q for innitely many n? (3 pts.)
vi. Let (a
n
)
n
2n
)
n
,
(a
2n+1
)
n
and (a
7n+1
)
n
all converge. Does the sequence (a
n
)
n
converge
necessarily? (5 pts.)
vii. Let (a
n
)
n
2
n
)
n
converges. Discuss the convergence of the sequence (a
n
)
n
. (5 pts.)
viii. Let (a
n
)
n
n
a
n
= . Is it
true that lim
n
a
2n
= ? (2 pts.)
ix. Let (a
n
)
n
n
a
n
= .
Discuss the convergence of lim
n
1/a
2n
= 0. (10 pts.)
x. Find the following limits and prove your result using only the denition.
(18 pts.)
a. lim
n
3n
2
5
5n
2
+3n
.
b. lim
n
n
2
5n+3
4n+2
c. lim
n
5n
2
4
2n
3
+2
xi. Find lim
n
_
n
2
1
n
3
n5
_
n
2
1
2n3
xii. Show that the series

n=0
(1)
n
z
2n+1
/(2n + 1)! converges for all z R.
(4 pts.)
xiii. Show that the series

n=1
_
n
n
2
+1
_
n/3
xiv. Let (a
n
)
n
such that [a
n+1
[ r[a
n
n
)
n
? (6 pts.)
19.6 Midterm of Math 152, April 2004
n
1
_
[n
2
2[
.
n
1
n
2
+ 1
.
n
1
_
[n
4
6[
.

n
a
n
n
a
n
= 0.
v. Suppose that (a
n
)
n
series

n
a
n
n
na
n
= 0.
n
)
n

n
a
n
is convergent
but that lim
n
na
n
,= 0.
n
a
n
(b
n
)
n

n
a
n
b
n
viii. Let (a
n
)
n
n=1
[a
n
a
n+1
[ converges. Such
19.7 Final of Math 152, June 2004
I. Convergent Sequences. For each of the topological spaces (X, ), describe
the convergent sequences and discuss the uniqueness of their limits.
i. = (X). ((X) is the set of all subsets of X, 2 pts.).
Answer: Only the eventually constant sequences converging to that con-
stant.
ii. = , X (2 pts.).
Answer: All sequences converge to all elements.
iii. a X is a xed element and is the set of all subsets of X that do not
contain a, together with X of course. (5 pts.)
Answer: First of all, all sequences converge to a. Second: If a sequence
converges to b ,= a, then the sequence must be eventually the constant b.
19.7. FINAL OF MATH 152, JUNE 2004 187
iv. a X is a xed element and is the set of all subsets of X that contain
a, together with of course. (5 pts.)
Answer: Only the eventually constant sequences converge to a. A se-
quence converge to b ,= a if and only if the sequence eventually takes only
the two values a and b.
v. is the set of all conite subsets of X, together with the of course. (6
pts.)
Answer: All the sequences without innitely repeating terms converge
to all elements. Eventually constant sequences converge to the constant.
There are no others.
II. Subgroup Topology on Z. Let = nZ+m : n, m Z, n ,= 0.
We know that (Z, ) is a topological space.
i. Let a Z. Is Z a open in ? (5 pts.)
Answer: Yes.
n=0,1
nZ (3Z + 2) = Z 1.Translating this set by
a 1, we see that Z a is open.
ii. Find an innite non open subset of Z. (5 pts.)
Answer: The set of primes is not an open subset. Because otherwise, for
some a ,= 0 and b Z, the elements of aZ + b would all be primes. So b,
ab +b and 2ab +b would be primes, a contradiction.
iii. Let a, b Z. Is the map f
a,b
: Z Z dened by f
a,b
(z) = az + b
continuous? (Prove or disprove). (10 pts.)
Answer: Translation by b is easily shown to be continuous. Let us con-
sider the map f(z) = az. If a = 0, 1, 1 then clearly f is continuous.
Assume a ,= 0, 1 and that f is continuous. We may assume that a > 1
(why?) Choose a b which is not divisible by a. Then f
1
(bZ) is open,
hence contains a subset of the form cZ + d. Therefore a(cZ + d) bZ.
Therefore ac = b and so a divides b, a contradiction. Hence f is not
continuous unless a = 0, 1.
iv. Is the map f
a,b
: Z Z dened by f(z) = z
2
continuous? (Prove or
disprove). (5 pts.)
Answer: No! Left as an exercise.
v. Is the topological space (Z, ) compact? (Prove or disprove).(15 pts.)
Answer: First Proof: Note rst the complement of open subsets of
the form aZ + b are also open as they are unions of the form aZ + c
for c = 0, 1, . . . , a 1 and c , b mod a. Now consider sets of the form
U
p
= pZ + (p 1)/2 for p an odd prime. Then
p
U
p
= because if
a
p
U
p
then for some x Z 1, 2a + 1 = px + p, so that a is
divisible by all primes p and a = 0. But if a = 0 then (p1)/2 is divisible
by p, a contradiction. On the other hand no nite intersection of the U
p
s
can be emptyset as (aZ + b) (cZ + b) ,= if a and b are prime to each
other (why?) Hence (U
c
p
)
p
is an open cover of Z that does not have a nite
cover. Therefore Z is not compact.
First Proof: Let p be a prime and a = a
0
+a
1
p +a
2
p
2
+. . . be a p-adic
integer which is not in Z. Let
b
n
= a
0
+a
1
p +a
2
p
2
+. . . +a
n1
p
n1
.
Then
n
p
n
Z +b
n
= but no nite intersection is empty. We conclude as
above.
III. Miscellaneous.
i. Let f : R R be the squaring map. Suppose that the arrival set is
endowed with the usual Euclidean topology. Find the smallest topology
on the domain that makes f continuous. (5 pts.)
Answer: The smallest such topology is the set
U U : U open in the usual topology of R.
ii. Let be the topology on R generated by [a, b) : a, b R. Compare this
topology with the Euclidean topology. (3 pts.) Is this topology generated
by a metric? (20 pts.)
Answer: Any open subset of the Euclidean topology is open in this topol-
ogy because (a, b) =
n=1
[a + 1/n, b). But of course [0, 1) is not open in
the usual topology.
Assume a metric generates the topology. Note that [0, ) is open as it
is the union of open sets of the form [0, n) for n N. Thus the sequence
(1/n)
n
cannot converge to 0. In fact for any b R, no sequence can
converge to b from the left. Thus for any b R there is an
b
> 0 such
that B(b,
b
) [b, ). Let b
0
be any point of R. Let
0
> 0 be such that
B(b
0
,
0
) [b
0
, ). Since b
0
is not open, there is b
1
B(b
0
,
0
) b
0
.
Let 0 <
1
<
0
/2 be such that B(b
1
,
1
) [b
1
, ) B(b
0
,
0
). Inductively
we can nd (b
n
)
n
and (
n
)
n
such that B(b
n
,
n
) [b
n
, )B(b
n1
,
n1
)
b
n1
and
n
<
0
/2
n
. Then (b
n
)
n
is a strictly increasing convergent
sequence, a contradiction.
iii. Show that the series

n
i=0
x
n
/n! converges for any x R (10 pts.) Show
that the map exp : R R dened by exp(x) =
n
i=0
x
n
/n! is continuous.
(10 pts.)
Answer: For the rst part show that the sequence of partial sums is
Cauchy. The second part is easy as well, just write down.
Bibliography
[A] Stephen Abbott, Understanding Analysis, Springer 2001.
[AZ] Martin Aigner and G unter M. Ziegler, Proofs from THE BOOK,
second corrected printing, Springer 1999.
[BR] G. Bouligand and J. Rivaud, Lenseignement des mathematiques
generales par les problèmes, Librairie Vuibert, 4th edition, 1968.
[PZ] William R. Parzynski and Philip W. Zipse, Introduction to Mathe-
matical Analysis, McGraw-Hill 1987.
[V] Georges Valiron, Theorie des Fonctions, Masson et C
ie
, 1941, third
edition 1966.
189
Index
<, 13
R, 53
+, 13
, 14
, 13
>, 13
, 13
0
1
, 16
0, 13, 14
0
0
, 30
1, 13
2, 20
Abels Lemma, 120
Abels Theorem, 120
abelian group, 16
absolute convergence, 87
absolute extremum, 144
absolute maximum, 144
absolute minimum, 144
absolute value of a real number, 21
addition, 1416
addition of real numbers, 13
algebraic numbers, 95
alternating series, 88
analytic functions, 150, 154
Archimedean Property, 29
associative binary operation, 9
associativity of series, 85
associativity of the addition, 14
associativity of the multiplication,
16
Banach algebra, 53
Banach Space, 53
base of a topology, 128
binary operation, 9
Binary Relations, 1011
Bolzano-Weierstrass Theorem, 71, 73
bounded away from a, 64
bounded sequence, 59
bounded set, 59
C, 39
cartesian product, 43
Cauchy Condensation Test., 87
Cauchy sequence, 68
Cauchys Mean-Value Theorem, 163
center of a ball, 54
chain, 21
circle, 54
closed subset, 127
closure, 128
coecients of a polynomial, 65
commutative binary operation, 9
commutative group, 16
commutativity, 17
compact, 129
Comparison Test, 86
complete metric space, 79
completeness, 23
complex numbers, 3942
conditional convergence, 92
conjugate of a complex number, 41
connected topological space, 131
constant sequence, 57
continuous at a point, 133
continuous function, 134, 135
contractive sequence, 75
convergence, 57
convergence of series, 83
convergent sequence, 57
cosine, 90
190
INDEX 191
decimal expansion, 92
decreasing sequence, 33
degree, 65
dense, 34
derivative, 143
dierentiable at a point, 143
disconnected, 131
discrete metric, 51
discrete topology, 125
distributivity, 1718
divergence of series, 83
divergence to innity, 76
divergent sequence, 57
division, 36
division of integers, 29
e, 74, 156
equivalence relation, 11
Euclidean metric space, 51
Eulers constant, 169
Eulers Gamma function, 175
Euler-Mascheroni constant, 169
eventually constant sequence, 57
exponentiation, 30, 90
extended real line, 126
extremum, 144
factorial, 31
Fibonacci sequence, 82
eld, 18
nest topology, 125
Gamma function, 175
geometric series, 85
glb(A), 22
group, 14
harmonic series, 85
Hausdor space, 182
Heine-Borel Theorem, 129
Hilbert space, 179
homeomorphism, 182
i, 40
Id
A
, 10
identity element, 14, 16
identity element of a binary opera-
tion, 9
increasing sequence, 33
induced topology, 126
inductive, 27
inf(A), 22
integers, 34
integral part, 29, 36
Integral Test, 168
interior, 127
Intermediate Value Theorem, 138
intervals, 21
inverse, 14, 16
inverse element, 14, 16
inverse element in a binary opera-
tion, 10
inverse of a function (bijection), 10
isolated point, 133
less than, 13
lexicographique ordering, 23
lim, 78
limit, 57, 182
limit at innity, 123
limit inferior, 78
limit of a function, 121, 122
limit point, 121, 133
limit superior, 78
lim, 78
linearly ordered set, 21
Liouville number, 96
Lipschitz condition, 163
local extremum, 144
local maximum, 144
local minimum, 144
lub(A), 22
maximal element, 22
Mean Value Theorem of Dierential
Calculus, 145
metric space, 50
metrisable topology, 125
minus, 14
Monotone Convergence Theorem, 70
monotone sequence, 70
m
th
-root, 31
192 INDEX
multiplication, 1617
multiplication of real numbers, 13
Natural Numbers, 2730
natural numbers, 27
nested intervals property, 33
nonalgebraic numbers, 95
nondecreasing sequence, 33, 70
nonincreasing sequence, 33, 70
nonnegative, 20
normed real vector space, 52
n
th
power, 30
one, 13
open ball, 54
open subset, 125
open subset of a metric space, 54
order, 13, 1921
ordered eld, 19
p-adic metric, 51
partial sums, 83
Peano, 140
pointwise convergence of a family of
functions, 122
pointwise convergence of a series of
functions, 115, 119
pointwise limit, 113, 122
polynomial, 44, 65
power series, 91
product, 13
product topology, 181
R, 13
R
, 16, 20
R
<0
, 20
R
0
, 20
R
>0
, 20
R
0
, 20
Raabe and Duhamels Convergence
Rule, 170
radius of a ball, 54
radius of convergence, 91, 149, 150
Ratio Test, 90
rational numbers, 34
Real Numbers, 1324
real sequence, 33
real vector space, 44
real vector spaces, 4347
rearranging the terms of a series, 84,
87
reexivity, 11
Rolles Theorem, 145
Root Test, 90
scalar multiplication, 43
scalar product, 49
sequence, 33
sequential compactness, 183
series, 83
series of functions, 115
sine, 90
square root, 34, 73
strictly decreasing sequence, 33
strictly increasing sequence, 33
strictly negative, 20
strictly positive, 20
subsequence of a sequence, 33
subspace, 46
sum, 13
sum of vectors, 43
sup(A), 22
sup metric, 51
Sym(A), 10
symmetry, 11
T
0
-topological space, 183
Taylor series, 151
(Tf)(x), 154
topological space, 125
topology generated by, 126, 128
totally ordered set, 21
totaly disconnected, 131
transcendental numbers, 95
transitivity, 11
triangular inequality, 50
ultrametric, 51
uniform convergence, 115
uniform convergence of a family of
functions, 122
uniform convergence of a series of
functions, 119
INDEX 193
uniqueness of the real number sys-
tem, 37
usual metric, 51
vector, 43
vector space over R, 44
Wallis formula, 167
Weierstrass M-Test, 120
, 13
x
1
, 16
x
2
, 19
Z, 34
zero, 13, 14
zero sequence, 45

Analysis (Lecture Notes by Ali Nesin)

Uploaded by

Copyright:

Available Formats

Analysis (Lecture Notes by Ali Nesin)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analysis (Lecture Notes by Ali Nesin)

Uploaded by

Copyright:

Available Formats

Analysis Notes

(only a draft, and the rst one!)

Lemma 2.1.4 If x, y R satisfy x +y = 0, then x = y.

= R 0. Since 1 ,= 0, the element 1 is an element of R

, and 1 respectively. For example, Axiom A1 will then read

, the element y as in M3 is unique.

. Let y be as in M3. Let y

that satises M3 is unique, we can

containing 1, closed under multiplication and

is less than any element of A

(0, s) = (0, r +s)

f(r +s) = f(r) +

) is a structure that satises all the

is just R with its elements renamed:

, contains a dense subset Q

such that for all r, s R,

f(s) (f is order preserving)

. A map f that satises (1) must send 0 to 0

f(0) = f(0) = f(0 + 0) = f(0) +

f(0) and so by simplifying we get

for the binary operation +

f(x) and so f(x) = f(x).

for some x R 0. We will get

and so, for all r R, f(r) = f(r 1) = f(r)

f(y) = f(x +(y)) = f(x y). By

. Since f(1) = f(1 1) = f(1)

f(1), f(1) is either 0

and f is one to one, f(1) = 1

. We can show by induction on n that f(n) N

. It is also clear that f(N) is an inductive subset of N

. By 1e and 1b, f(Z) = Z

. Let n/m Q with n, m Z,

f(n/m) = f(m n/m) = f(n) f(Z) = Z

. Thus f(Q) f(Q

, it is also easy to show that

. Note rst that, in R, x y if

f(z)). Since f(z) ,= 0

, we get f(x) <

. Consider the sets

is an order preserving bi-

such that f(r) < q

= f(q) < f(r), a contradiction. Thus f(r) r

and assuming f(n) has

By induction on n, one can prove that for all n N, f(n) N

is one to one. Assuming f(n) = f(m) for

= f(0) = f(n) = f(m), therefore, by the very denition of f, m cannot

= f(k +1) = f(n) = f(m), therefore m ,= 0 and so m = +1 for

= f(m) = f( +1) = f() +

and so f(k) = f().

, i.e. we let f(n) = n

f(m) and f(n m) =

. Here a/b stands for a

. We should rst check that this map is well-

. We still denote by f this extended map.

is certainly onto: If a/b Q

with a, b N, then f(n) = a and

is one to one, additive, multiplicative

, the sets f(L

satises all the required