Numbers, Sequences and Series

Lecture Notes

Dr. Silvio Fanzon

3 Oct 2024

Academic Year 2024/25

Department of Mathematics

University of Hull
Table of contents
Welcome 5
Digital Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Readings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1 Introduction 6

2 Preliminaries 14
2.1 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Operations on sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1 Union and intersection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.2 Inclusion and equality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.3 Infinite unions and intersections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.4 Complement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.5 Power set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.6 Product of sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Equivalence relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Order relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.7 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.8 Absolute value or Modulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.9 Triangle inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.10 Proofs in Mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.11 Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

3 Real Numbers 46
3.1 Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2 Ordered fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3 Cut Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.4 Supremum and infimum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.4.1 Upper bound, supremum, maximum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.4.2 Lower bound, infimum, minimum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.5 Completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.6 Equivalence of Completeness and Cut Property . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.7 Axioms of Real Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.8 Special subsets of ℝ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.8.1 Natural numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

2
Numbers, Sequences and Series Page 3

3.8.2 Principle of induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

3.8.3 Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.8.4 Rational numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

4 Properties of ℝ 89
4.1 Archimedean Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.2 Nested Interval Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3 Revisiting Sup and inf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.4 Density of ℚ in ℝ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.5 Existence of 𝑘-th Roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.6 Cardinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

5 Complex Numbers 118

5.1 The field ℂ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.1.1 Division in ℂ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.1.2 ℂ is not ordered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.1.3 Completeness of ℂ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.2 Complex conjugates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.3 The complex plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.3.1 Distance on ℂ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.3.2 Properties of modulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.4 Polar coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.5 Exponential form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5.6 Fundamental Theorem of Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.7 Solving polynomial equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5.8 Roots of unity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.9 Roots in ℂ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

6 Sequences in ℝ 158
6.1 Definition of sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.2 Convergent sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
6.3 Divergent sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
6.4 Uniqueness of limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
6.5 Bounded sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
6.6 Algebra of limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6.7 Fractional powers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
6.8 Limit Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
6.8.1 Squeeze Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
6.8.2 Geometric sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
6.8.3 Ratio Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
6.9 Monotone sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
6.9.1 Example: Euler’s Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.10 Some important limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 4

7 Sequences in ℂ 209
7.1 Definition and convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
7.2 Boundedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
7.3 Algebra of limits in ℂ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
7.4 Convergence to zero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
7.5 Geometric sequence Test and Ratio Test in ℂ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
7.6 Convergence of real and imaginary part . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

8 Series 222
8.1 Convergent series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
8.2 Geometric series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
8.3 Algebra of Limits for Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
8.4 Non-negative series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
8.4.1 Cauchy Condensation Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
8.4.2 Comparison Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
8.4.3 Limit Comparison Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
8.4.4 Ratio Test for positive series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
8.5 General series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
8.5.1 Absolute Convergence Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
8.5.2 Ratio Test for general series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
8.5.3 Exponential function and Euler’s Number . . . . . . . . . . . . . . . . . . . . . . . . . 263
8.5.4 Conditional convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
8.5.5 Dirichlet and Alternating Series Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
8.5.6 Abel’s Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

License 273
Reuse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Citation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273

References 274

Dr. Silvio Fanzon [email protected]

Welcome
These are the Lecture Notes of Numbers, Sequences and Series 400297 for 2024/25 at the University of
Hull. I will use this material during lectures. If you have any question or find any typo, please email me at

[email protected]

Up to date information about the course, Tutorials and Homework will be published on the University of Hull
Canvas Website

canvas.hull.ac.uk/courses/73579

Digital Notes
Digital version of these notes available at

silviofanzon.com/2024-NSS-Notes

Readings
We will study the set of real numbers ℝ, and then sequences and series in ℝ. I will follow mainly the textbook
by Bartle and Sherbert [2]. Another good reading is the book by Abbott [1]. I also point out the classic book
by Rudin [3], although this is a more difficult read.

ĺ You are not expected to purchase any of the above books. These lecture notes will cover 100% of the
topics you are expected to known in order to excel in the Homework and Final Exam.

5
1 Introduction
The first aim of this lecture notes is to rigorously introduce the set of Real Numbers, which is denoted by
ℝ. But what do we mean by real numbers? To start our discussion, introduce the set of natural numbers (or
non-negative integers)
ℕ = {0, 1, 2, 3, 4, 5, … }

On this set we have a notion of sum of two numbers, denoted as usual by

𝑛+𝑚

for 𝑛, 𝑚 ∈ ℕ. Here the symbol ∈ denotes that 𝑚 and 𝑛 belong to ℕ. For example 3 + 7 results in 10.

Question 1.1

Can the sum be inverted? That is, given any 𝑛, 𝑚 ∈ ℕ, can you always find 𝑥 ∈ ℕ such that

𝑛+𝑥 = 𝑚? (1.1)

Of course to invert (1.1) we can just perform a subtraction, implying that

𝑥 = 𝑚 −𝑛.

But there is a catch. In general 𝑥 does not need to be in ℕ. For example, take 𝑛 = 10 and 𝑚 = 1. Then 𝑥 = −9,
which does not belong to ℕ. Therefore the answer to Question 1.1 is NO.
To make sure that we can always invert the sum, we need to extend the set ℕ. This is done simply by
introducing the set of integers
ℤ ∶= {−𝑛, 𝑛 ∶ 𝑛 ∈ ℕ} ,

that is, the set

ℤ ∶= {… , −3, −2, −1, 0, 1, 2, 3, …} .

The sum can be extended to ℤ, by defining

(−𝑛) + (−𝑚) ∶= −(𝑚 + 𝑛) (1.2)

6
Numbers, Sequences and Series Page 7

for all 𝑚, 𝑛 ∈ ℕ. Now every element of ℤ possesses an inverse, that is, for each 𝑛 ∈ ℤ, there exists 𝑚 ∈ ℤ,
such that
𝑛 + 𝑚 = 0.

Can we characterize 𝑚 explicitly? Yes; Seeing the definition at (1.2), we simply have

𝑚 = −𝑛 .

On the set ℤ we can also define the operation of multiplication in the usual way. For 𝑛, 𝑚 ∈ ℤ, we denote
the multiplication of 𝑛 times 𝑚 by
𝑛𝑚 or 𝑛 ⋅ 𝑚 .
For example 7 ⋅ 2 = 14 and 1 ⋅ (−1) = −1.

Question 1.2

Can the multiplication in ℤ be inverted? That is, given any 𝑛, 𝑚 ∈ ℤ, can you always find 𝑥 ∈ ℤ such
that
𝑛𝑥 = 𝑚 ? (1.3)

To invert (1.3) if 𝑛 ≠ 0, we can just perform a division, to obtain

𝑚
𝑥= .
𝑛
But again there is a catch. Indeed taking 𝑛 = 2 and 𝑚 = 1 yields 𝑥 = 1/2, which does not belong to ℤ. The
answer to Question 1.2 is therefore NO.
Thus, in order to invert the multiplication, we need to extend the set of integers ℤ. This extension is called
the set of rational numbers, defined by
𝑚
ℚ ∶= { ∶ 𝑚, 𝑛 ∈ ℤ, 𝑛 ≠ 0} .
𝑛
We then extend the operations of sum and multiplication to ℚ by defining
𝑚 𝑝 𝑚𝑞 + 𝑛𝑝
+ ∶=
𝑛 𝑞 𝑛𝑞
and
𝑚 𝑝 𝑚𝑝
⋅ ∶=
𝑛 𝑞 𝑛𝑞
Now the multiplication is invertible in ℚ. Specifically, each non-zero element has an inverse: the inverse of
𝑚/𝑛 is given by 𝑛/𝑚.
To summarize, we have extended ℕ to ℤ, and ℤ to ℚ. By construction we have

ℕ ⊂ ℤ ⊂ ℚ.

Moreover sum and product are invertible in ℚ. Now we are happy right? So and so.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 8

Question 1.3

Can we represent the set ℚ with a drawing?

It is clear how to draw ℤ, as seen below.

Figure 1.1: Representation of integers ℤ

However ℚ is much larger than the set ℤ represented by the ticks in Figure 1.1. What do we mean by larger?
For example, consider 0 ∈ ℚ.

Question 1.4

What is the number 𝑥 ∈ ℚ which is closest to 0?

There is no right answer to the above question, since whichever rational number 𝑚/𝑛 you consider, you can
always squeeze the rational number 𝑚/(2𝑛) in between:
𝑚 𝑚
0< < .
2𝑛 𝑛
For example think about the case of the numbers
1
for 𝑛 ∈ ℕ, 𝑛 ≠ 0 .
𝑛
Such numbers get arbitrarily close to 0, as depicted below.

1
Figure 1.2: Fractions 𝑛
can get arbitrarily close to 0

Maybe if we do the same reasoning with other progressively smaller rational numbers, we manage to fill up
the interval [0, 1]. In other words, we might conjecture the following.

Conjecture 1.5

Maybe ℚ can be represented by a continuous line.

Do you think the above conjecture is true? Conjecture 1.5 is false, as shown by the Theorem below.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 9

Theorem 1.6

The number √2 does not belong to ℚ.

Theorem 1.6 is the reason why √2 is called an irrational number. For reference, a few digits of √2 are given
by
√2 = 1.414213562373095048 …
and the situation is as in the picture below.

Figure 1.3: Representing √2 on the numbers line.

We can therefore see that Conjecture 1.5 is false, and ℚ is not a line: indeed ℚ has a gap at √2. Let us see
why Theorem 1.6 is true.

Proof: Proof of Theorem 1.6

We prove that
√2 ∉ ℚ
by contradiction.
Wait, what does this mean? Proving the claim by contradiction means assuming that the claim is false.
This means we assume that
√2 ∈ ℚ . (1.4)
From this assumption we then start deducing other statements, hoping to encounter a statement which
is FALSE. But if (1.4) leads to a false statement, then it must be that (1.4) was FALSE to begin with. Thus
the contrary of (1.4) must hold, meaning that
√2 ∉ ℚ
as we wanted to show. This would conclude the proof.
Now we need to actually show that assuming (1.4) will lead to a contradiction. Since this is our first proof,
let us take it slowly, step-by-step.
1. Assuming (1.4) just means that there exists 𝑞 ∈ ℚ such that
𝑞 = √2 . (1.5)

2. Since 𝑞 ∈ ℚ, by definition we have

𝑚
𝑞=
𝑛
for some 𝑚, 𝑛 ∈ ℕ with 𝑛 ≠ 0.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 10

3. Recalling (1.5), we then have

𝑚
= √2 .
𝑛
4. We can square the above equation to get

𝑚2
= 2. (1.6)
𝑛2

5. Withouth loss of generality, we can assume that 𝑚 and 𝑛 have no common factors.
Wait. What does Step 5 mean? You will encounter the sentence withouth loss of gen-
erality many times in mathematics. It is often abbreviated in WLOG. WLOG means that
we can make some extra assumption which does not affect the validity of the proof in
general.
For example in our case we can assume that 𝑚 and 𝑛 have no common factor. This is
because if 𝑚 and 𝑛 had common factors, then it would mean

𝑚 = 𝑎𝑚̃ , 𝑛 = 𝑎𝑛̃

for some factor 𝑎 ∈ ℕ with 𝑎 ≠ 0. Then

𝑚 𝑎𝑚̃ 𝑚̃
= = .
𝑛 𝑎𝑛̃ 𝑛̃
Therefore by (1.6)
𝑚̃ 2
= 2.
𝑛̃ 2
and this time 𝑚̃ and 𝑛̃ have no common factors by construction. The proof can now
proceed in the same way it would have proceeded from Step 4, but in addition we have
the hypothesis that 𝑚̃ and 𝑛̃ have no common factors.

6. Equation (1.6) implies

𝑚2 = 2𝑛2 . (1.7)
Therefore the integer 𝑚2 is an even number.
Why is 𝑚2 even? As you already know, even numbers are

0, 2, 4, 6, 8, 10, 12, …

All these numbers have in common that they can be divided by 2, and so they can be
written as
2𝑝
for some 𝑝 ∈ ℕ. For example 52 is even, because

52 = 2 ⋅ 26 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 11

Instead, odd numbers are

1, 3, 5, 7, 8, 9, 11, …
These can be all written as
2𝑝 + 1
for some 𝑝 ∈ ℕ. For example 53 is odd, because
52 = 2 ⋅ 26 + 1 .

7. Since 𝑚2 is an even number, it follows that also 𝑚 is an even number. Then there exists 𝑝 ∈ ℕ such
that
𝑚 = 2𝑝 . (1.8)
Why is 𝑚 even if 𝑚2 is even? Let us see what happens if we take the square of an even
number 𝑚 = 2𝑝
𝑚2 = (2𝑝)2 = 4𝑝 2 = 2(2𝑝 2 ) = 2𝑞 .
Thus 𝑚2 = 2𝑞 for some 𝑞 ∈ ℕ, and so 𝑚2 is an even number. If instead 𝑚 is odd, then
𝑚 = 2𝑝 + 1 and
𝑚2 = (2𝑝 + 1)2 = 4𝑝 2 + 4𝑝 + 1 = 2(2𝑝 2 + 2𝑝) + 1
showing that also 𝑚2 is odd.
This proves Step 7: Indeed we know that 𝑚2 is an even number from Step 6. If 𝑚 was
odd, then 𝑚2 would be odd. Hence 𝑚 must be even as well.
8. If we substitute (1.8) in (1.7) we get
𝑚2 = 2𝑛2 ⟹ (2𝑝)2 = 2𝑛2 ⟹ 4𝑝 2 = 2𝑛2
Dividing both terms by 2, we obtain
𝑛2 = 2𝑝 2 . (1.9)
9. We now make a series of observations:
• Equation (1.9) says that 𝑛2 is even.
• The same argument in Step 7 guarantees that also 𝑛 is even.
• We have already seen that 𝑚 is even.
• Therefore 𝑛 and 𝑚 are both even.

• Hence 𝑛 and 𝑚 have 2 as common factor.

• But Step 5 says that 𝑛 and 𝑚 have no common factors.
• CONTRADICTION
10. Our reasoning has run into a contradiction, stemming from assumption (1.4). Therefore (1.4) is
FALSE, and so
√2 ∉ ℚ
ending the proof.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 12

Seeing that √2 ∉ ℚ, we might be tempted to just fill in the gap by adding √2 to ℚ. However, with analogous
proof to Theorem 1.6, we can prove that
√𝑝 ∉ ℚ
for each prime number 𝑝. As there are infinite prime numbers, this means that ℚ has infinite gaps. Then we
might attempt to fill in these gaps via the extension

ℚ̃ ∶= ℚ ∪ {√𝑝 ∶ 𝑝 prime} .

However even this is not enough, as we would still have numbers which are not contained in 𝑄,̃ for example

√2 + √3, 𝜋, 𝜋 + √2 ∉ ℚ̃ .

Conclusion: It is now intuitive to think that there is no straightforward way to fill the gaps of ℚ by adding
numbers by hand.

Remark 1.7

Proving that
√2 + √3 ∉ ℚ
is relatively easy, and will be left as an exercise. Instead, proving that

𝜋 ∉ℚ

is way more complicated. There are several proofs of the fact, all requiring mathematics which is more
advanced than the one presented in this course. For some proofs, see this Wikipedia page.

The reality of things is that to complete ℚ and make it into a continuous line we have to add a lot of points.
Indeed, we need to add way more points than the ones already contained in ℚ.

Definition 1.8
Such extension of ℚ will be called ℝ, the set of real numbers.

The inclusions will therefore be

ℕ ⊂ ℤ ⊂ ℚ ⊂ ℝ.
The set ℝ is not at all trivial to construct. In fact, explicit constructions are so involved and technical, that
they are best left for future reading. It is customary instead to proceed as follows:

• We will assume that ℝ exists and satisfies a set of axioms.

• One of the axioms states that ℝ fills all the gaps that ℚ has. Therefore ℝ can be thought as a contin-
uous line.
• We will study the properties of ℝ which descend from such axioms.

For example one of the properties of ℝ will be the following:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 13

Theorem 1.9: We will prove this in the future

ℝ contains all the square roots. This means that for every 𝑥 ∈ ℝ with 𝑥 ≥ 0, we have

√𝑥 ∈ ℝ .

A concrete model for the real numbers ℝ can be constructed using Dedekind cuts. The interested reader
can refer to the Appendix in Chapter 1 of [3], or to the beautifully written Chapter 8.6 in [1]. Such model of
ℝ can be used to prove the following Theorem:

Theorem 1.10: Existence of the Real Numbers

There exists a set ℝ, called the set of Real Numbers, which has the following properties:

• ℝ extends ℚ, that is,

ℚ ⊂ ℝ.
• ℝ satisfies a certain set of axioms.
• ℝ fills all the gaps that ℚ has. In particular ℝ can be represented by a continuous line.

Dr. Silvio Fanzon [email protected]

2 Preliminaries
Before introducing ℝ we want to make sure that we cover all the basics needed for the task.

2.1 Sets
A set is a collection of objects. These objects are called elements of the set. For example in the previous
section we mentioned the following sets:

• ℕ the set of natural numbers

• ℤ the set of integers
• ℚ the set of rational numbers
• ℝ the set of real numbers

Given an arbitrary set 𝐴, we write

𝑥∈𝐴

if the element 𝑥 belongs to the set 𝐴. If an element 𝑥 is not contained in 𝐴, we say that

𝑥 ∉ 𝐴.

Remark 2.1
A set can contain all sorts of elements. For example the students in a classroom can be modelled by a set
𝑆. The elements of the set are the students. For example

𝑆 = {Alice, Olivia, Jake, Sahab}

In this case we have

Alice ∈ 𝑆
but instead

Silvio ∉ 𝑆 .

14
Numbers, Sequences and Series Page 15

2.2 Logic
In this section we introduce some basic logic symbols. Suppose that you are given two statements, say 𝛼 and
𝛽. The formula
𝛼 ⟹ 𝛽
means that 𝛼 implies 𝛽. In other words, if 𝛼 is true then also 𝛽 is true.
The formula
𝛼 ⟸ 𝛽
means that 𝛼 is implied by 𝛽: if 𝛽 is true then also 𝛼 is true.
When we write
𝛼 ⟺ 𝛽 (2.1)
we mean that 𝛼 and 𝛽 are equivalent. Note that (2.1) is equivalent to

𝛼 ⟹ 𝛽 and 𝛽 ⟹ 𝛼 .

Such equivalence is very useful in proofs.

Example 2.2

We have that
𝑥 > 0 ⟹ 𝑥 > −100 ,
and
contradiction ⟸ √2 ∈ ℚ .
Concerning ⟺ we have
𝑥 2 < 2 ⟺ −√2 < 𝑥 < √2 .

We now introduce logic quantifiers. These are

• ∀ which reads for all

• ∃ which reads exists
• ∃! which reads exists unique
• ∄ which reads does not exists

These work in the following way. Suppose that you are given a statement 𝛼(𝑥) which depends on the point
𝑥 ∈ ℝ. Then we say

• 𝛼(𝑥) is satisfied for all 𝑥 ∈ 𝐴 with 𝐴 some collection of numbers. This translates to the symbols

𝛼(𝑥) is true ∀ 𝑥 ∈ 𝐴 ,

• There exists some 𝑥 in ℝ such that 𝛼(𝑥) is satisfied: in symbols

∃ 𝑥 ∈ ℝ such that 𝛼(𝑥) is true,

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 16

• There exists a unique 𝑥0 in ℝ such that 𝛼(𝑥) is satisfied: in symbols

∃! 𝑥0 ∈ ℝ such that 𝛼(𝑥0 ) is true,

• 𝛼(𝑥) is never satisfied:

∄ 𝑥 ∈ ℝ such that 𝛼(𝑥) is true.

Example 2.3

Let us make concrete examples:

• The expression 𝑥 2 is always non-negative. Thus we can say

𝑥 2 ≥ 0 for all 𝑥 ∈ ℝ .

• The equation 𝑥 2 = 1 has two solutions 𝑥 = 1 and 𝑥 = −1. Therefore we can say

∃ 𝑥 ∈ ℝ such that 𝑥 2 = 1 .

• The equation 𝑥 3 = 1 has a unique solution 𝑥 = 1. Thus

∃! 𝑥 ∈ ℝ such that 𝑥 3 = 1 .

• We know that the equation 𝑥 2 = 2 has no solutions in ℚ. Then

∄ 𝑥 ∈ ℚ such that 𝑥 2 = 2 .

2.3 Operations on sets

2.3.1 Union and intersection

For two sets 𝐴 and 𝐵 we define their union as the set

𝐴 ∪ 𝐵 ∶= {𝑥 ∶ 𝑥 ∈ 𝐴 or 𝑥 ∈ 𝐵} .

The intersection of 𝐴 and 𝐵 is defined by

𝐴 ∩ 𝐵 ∶= {𝑥 ∶ 𝑥 ∈ 𝐴 and 𝑥 ∈ 𝐵} .

We denote the empty set by the symbol ∅. Two sets are disjoint if

𝐴 ∩ 𝐵 = ∅.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 17

Example 2.4

Define the subset of rational numbers

5
𝑆 ∶= {𝑥 ∈ ℚ ∶ 0 < 𝑥 < } .
2
Then we have
ℕ ∩ 𝑆 = {1, 2} .
We can also define the sets of even and odd numbers by

𝐸 ∶= {2𝑛 ∶ 𝑛 ∈ ℕ} , (2.2)
𝑂 ∶= {2𝑛 + 1 ∶ 𝑛 ∈ ℕ} . (2.3)

Then we have

ℕ∩𝐸 = 𝐸, ℕ∩𝑂 = 𝑂, (2.4)

𝑂 ∪ 𝐸 = ℕ, 𝑂 ∩ 𝐷 = ∅. (2.5)

2.3.2 Inclusion and equality

Given two sets 𝐴 and 𝐵, we say that 𝐴 is contained in 𝐵 if all the elements of 𝐴 are also contained in 𝐵. This
will be denoted with the inclusion symbol ⊆, that is,
𝐴 ⊆ 𝐵.
In this case we say that
• 𝐴 is a subset of 𝐵,
• 𝐵 is a superset of 𝐴.
The inclusion 𝐴 ⊆ 𝐵 is equivalent to the implication:
𝑥∈𝐴 ⟹ 𝑥∈𝐵
for all 𝑥 ∈ 𝐴. The symbol ⟹ reads implies, and denotes the fact that the first condition implies the
second.

Example 2.5

Given two sets 𝐴 and 𝐵 we always have

(𝐴 ∩ 𝐵) ⊆ 𝐴 , (𝐴 ∩ 𝐵) ⊆ 𝐵 , (2.6)
𝐴 ⊆ (𝐴 ∪ 𝐵) , 𝐵 ⊆ (𝐴 ∪ 𝐵) . (2.7)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 18

We say that two sets 𝐴 and 𝐵 are equal if they contain the same elements. We denote equality by the symbol

𝐴 = 𝐵.

If 𝐴 ⊆ 𝐵 and 𝐴 ≠ 𝐵, we write
𝐴⊂𝐵 or 𝐴 ⊊ 𝐵.

Example 2.6

1. The sets
𝐴 = {1, 2, 3} , 𝐵 = {3, 1, 2}
are equal, that is 𝐴 = 𝐵. This is because they contain exactly the same elements: order does not
matter when talking about sets.

2. Consider the sets

𝐴 = {1, 2} , 𝐵 = {1, 2, 5} .
Then 𝐴 is contained in 𝐵, but 𝐴 is not equal to 𝐵. Therefore we write 𝐴 ⊂ 𝐵 or 𝐴 ⊊ 𝐵.

Proposition 2.7

Let 𝐴 and 𝐵 be sets. Then

𝐴=𝐵
if and only if
𝐴 ⊆ 𝐵 and 𝐵 ⊆ 𝐴 .

Proof
The proof is almost trivial. However it is a good exercise in basic logic, so let us do it.

1. First implication ⟹ :
Suppose that 𝐴 = 𝐵. Let us show that 𝐴 ⊆ 𝐵. Since 𝐴 = 𝐵, this means that all the elements of 𝐴
are also contained in 𝐵. Therefore if we take 𝑥 ∈ 𝐴 we have

𝑥 ∈ 𝐴 ⟹ 𝑥 ∈ 𝐵.

This shows 𝐴 ⊆ 𝐵. The proof of 𝐵 ⊆ 𝐴 is similar.

2. Second implication ⟸ :
Suppose that 𝐴 ⊆ 𝐵 and 𝐵 ⊆ 𝐴. We need to show 𝐴 = 𝐵, that is, 𝐴 and 𝐵 have the same elements.
To this end let 𝑥 ∈ 𝐴. Since 𝐴 ⊆ 𝐵 then we have 𝑥 ∈ 𝐵. Thus 𝐵 contains all the elements of 𝐴.
Since we are also assuming 𝐵 ⊆ 𝐴, this means that 𝐴 contains all the elements of 𝐵. Hence 𝐴 and
𝐵 contain the same elements, and 𝐴 = 𝐵.

The above proposition is very useful when we need to prove that two sets are equal: rather than showing
directly that 𝐴 = 𝐵, we can prove that 𝐴 ⊆ 𝐵 and 𝐵 ⊆ 𝐴.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 19

2.3.3 Infinite unions and intersections

Suppose given a set Ω, and a family of sets 𝐴𝑛 ⊆ Ω, where 𝑛 ∈ ℕ. Then we can define the infinte union
⋃ 𝐴𝑛 ∶= {𝑥 ∈ Ω ∶ 𝑥 ∈ 𝐴𝑛 for at least one 𝑛 ∈ ℕ} .
𝑛∈ℕ

The infinte intersection is defined as

⋂ 𝐴𝑛 ∶= {𝑥 ∈ Ω ∶ 𝑥 ∈ 𝐴𝑛 for all 𝑛 ∈ ℕ} .
𝑛∈ℕ

Example 2.8

Let the ambient set be Ω ∶= ℕ and define the family 𝐴𝑛 by

𝐴1 ∶= {1, 2, 3, 4, …} (2.8)
𝐴2 ∶= {2, 3, 4, 5, …} (2.9)
𝐴3 ∶= {3, 4, 5, 6, …} (2.10)
…… (2.11)
𝐴𝑛 ∶= {𝑛, 𝑛 + 1, 𝑛 + 2, 𝑛 + 3, …} , (2.12)

for arbitrary 𝑛 ∈ ℕ. Then

⋃ 𝐴𝑛 = ℕ . (2.13)
𝑛∈ℕ
The above equality can be easily proven using Proposition 2.7. Indeed, assume that 𝑚 ∈ ∪𝑛 𝐴𝑛 . Then
𝑚 ∈ 𝐴𝑛 for at least one 𝑛 ∈ ℕ. Since 𝐴𝑛 ⊆ ℕ, we conclude that 𝑚 ∈ ℕ. This shows

⋃ 𝐴𝑛 ⊆ ℕ .
𝑛∈ℕ

Conversely, suppose that 𝑚 ∈ ℕ. By definition 𝑚 ∈ 𝐴𝑚 . Hence there exists at least one index 𝑛, 𝑛 = 𝑚 in
this case, such that 𝑚 ∈ 𝐴𝑛 . Then by definition 𝑚 ∈ ∪𝑛∈ℕ 𝐴𝑛 , showing that

ℕ ⊆ ⋃ 𝐴𝑛 .
𝑛∈ℕ

Hence we conclude (2.13) by Proposition 2.7.

We also have that
⋂ 𝐴𝑛 = ∅ . (2.14)
𝑛∈ℕ
We prove the above by contradiction. Indeed, suppose that (2.14) is false, i.e.,

⋂ 𝐴𝑛 ≠ ∅ .
𝑛∈ℕ

This means there exists some 𝑚 ∈ ℕ such that 𝑚 ∈ ∩𝑛∈ℕ 𝐴𝑛 . Hence, by definition, 𝑚 ∈ 𝐴𝑛 for all 𝑛 ∈ ℕ.
However 𝑚 ∉ 𝐴𝑚+1 , yielding a contradiction. Thus (2.14) holds.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 20

2.3.4 Complement

Suppose that 𝐴 and 𝐵 are subsets of a larger set Ω. The complement of 𝐴 with respect to 𝐵 is the set of
elements of 𝐵 which do not belong to 𝐴, that is

𝐵 ∖ 𝐴 ∶= {𝑥 ∈ Ω ∶ 𝑥 ∈ 𝐵 and 𝑥 ∉ 𝐴} .

In particular, the complement of 𝐴 with respect to Ω is denoted by

𝐴𝑐 ∶= Ω ∖ 𝐴 ∶= {𝑥 ∈ Ω ∶ 𝑥 ∉ 𝐴} .

Remark 2.9

Suppose that 𝐴 ⊆ Ω. Then 𝐴 and 𝐴𝑐 form a partition of Ω, in the sense that

𝐴 ∪ 𝐴𝑐 = Ω and 𝐴 ∩ 𝐴𝑐 = ∅ .

Example 2.10

Suppose 𝐴, 𝐵 ⊆ Ω. Then
𝐴 ⊆ 𝐵 ⟺ 𝐵 𝑐 ⊆ 𝐴𝑐 .
Let us prove the above claim:

• First implication ⟹ :
Suppose that 𝐴 ⊆ 𝐵. We need to show that 𝐵𝑐 ⊆ 𝐴𝑐 . Hence, assume 𝑥 ∈ 𝐵𝑐 . By definition this
means that 𝑥 ∉ 𝐵. Now notice that we cannot have that 𝑥 ∈ 𝐴. Indeed, assume 𝑥 ∈ 𝐴. By
assumption we have 𝐴 ⊆ 𝐵, hence 𝑥 ∈ 𝐵. But we had assumed 𝑥 ∈ 𝐵, contradiction. Therefore it
must be that 𝑥 ∉ 𝐴. Thus 𝐵𝑐 ⊆ 𝐴𝑐 .

• Second implication ⟸ :
Essentially the same proof, hence we omit it.

We conclude by stating the De Morgan’s Laws. The proof will be left as an exercise.

Proposition 2.11: De Morgan’s Laws

Suppose 𝐴, 𝐵 ⊆ Ω. Then
(𝐴 ∩ 𝐵)𝑐 = 𝐴𝑐 ∪ 𝐵𝑐
and
(𝐴 ∪ 𝐵)𝑐 = 𝐴𝑐 ∩ 𝐵𝑐 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 21

2.3.5 Power set

Let Ω be an arbitray set. We define the power set of Ω as

𝒫 (Ω) ∶= {𝐴 ∶ 𝐴 ⊆ Ω} ,
that is, the power set of Ω is the set of all subsets of Ω.

Remark 2.12
It holds that:

1. 𝒫 (Ω) is always non-empty, since we have that

∅ ∈ 𝒫 (Ω) , Ω ∈ 𝒫 (Ω) .

2. Given 𝐴, 𝐵 ∈ 𝒫 (Ω), then the sets

𝐴 ∪ 𝐵, 𝐴 ∩ 𝐵, 𝐴𝑐 , 𝐵∖𝐴

are all elements of 𝒫 (Ω).

3. Suppose Ω is discrete and finite, that is,

Ω = {𝑥1 , … , 𝑥𝑚 }

for some 𝑚 ∈ ℕ. Then 𝒫 (Ω) contains 2𝑚 elements.

Indeed, suppose that we want to define a subset 𝐴 ⊂ Ω. Then for each element 𝑥𝑖 ∈ Ω
there are two choices: either we include 𝑥𝑖 ∈ 𝐴, or we do not include 𝑥𝑖 ∈ 𝐴. Therefore
there are
2 ⋅ 2 ⋅ … ⋅ 2 = 2𝑚
⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝑚 times
ways of constructing 𝐴. It follows that Ω possesses exactly 2𝑚 subsets, so that 𝒫 (Ω)
contains 2𝑚 elements.

Example 2.13

Define the set

Ω = {𝑥, 𝑦, 𝑧} .
Then 𝒫 (Ω) has 23 = 8 elements. These are

• ∅
• {𝑥}
• {𝑦}
• {𝑧}
• {𝑥, 𝑦}

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 22

• {𝑥, 𝑧}
• {𝑦, 𝑧}
• {𝑥, 𝑦, 𝑧}

We therefore write

𝒫 (Ω) = {∅, {𝑥}, {𝑦}, {𝑧}, {𝑥, 𝑦} (2.15)

{𝑥, 𝑧}, {𝑦, 𝑧}, {𝑥, 𝑦, 𝑧}} . (2.16)

2.3.6 Product of sets

Suppose 𝐴 and 𝐵 are two sets. The product of 𝐴 and 𝐵 is the set of pairs

𝐴 × 𝐵 ∶= {(𝑎, 𝑏) ∶ 𝑎 ∈ 𝐴, 𝑏 ∈ 𝐵} .

By definition two elements in 𝐴 × 𝐵 are the same, in symbols

(𝑎, 𝑏) = (𝑎,̃ 𝑏)̃

if and only if they are equal component-by-componenent, that is

𝑎 = 𝑎̃ , 𝑏 = 𝑏̃ .

2.4 Equivalence relation

Suppose 𝐴 is a set. A binary relation 𝑅 on 𝐴 is a subset

𝑅 ⊆ 𝐴 × 𝐴.

Definition 2.14: Equivalence relation

A binary relation 𝑅 is called an equivalence relation if it satisfies the following properties:

1. Reflexive: For each 𝑥 ∈ 𝐴 one has

(𝑥, 𝑥) ∈ 𝑅 ,
This is saying that all the elements in 𝐴 must be related to themselves

2. Symmetric: We have
(𝑥, 𝑦) ∈ 𝑅 ⟹ (𝑦, 𝑥) ∈ 𝑅
If 𝑥 is related to 𝑦, then 𝑦 is related to 𝑥

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 23

3. Transitive: We have
(𝑥, 𝑦) ∈ 𝑅 , (𝑦, 𝑧) ∈ 𝑅 ⟹ (𝑥, 𝑧) ∈ 𝑅
If 𝑥 is related to 𝑦, and 𝑦 is related to 𝑧, then 𝑥 must be related to 𝑧

Notation 2.15

If (𝑥, 𝑦) ∈ 𝑅 we write
𝑥∼𝑦
and we say that 𝑥 and 𝑦 are equivalent.

Definition 2.16: Equivalence classes

Suppose 𝑅 is an equivalence relation on 𝐴. The equivalence class of an element 𝑥 ∈ 𝐴 is the set

[𝑥] ∶= {𝑦 ∈ 𝐴 ∶ 𝑦 ∼ 𝑥} .

The set of equivalence classes of elements of 𝐴 with respect to the equivalence relation 𝑅 is denoted by

𝐴/𝑅 ∶= 𝐴/ ∼ ∶= {[𝑥] ∶ 𝑥 ∈ 𝐴} .

In order for the definition of [𝑥] to be well-posed we need to check that:

1. [𝑥] is non-empty.
2. [𝑥] does not depend on the representative 𝑥: we need to check that
𝑥∼𝑦 ⟺ [𝑥] = [𝑦]
This is shown in the following proposition.

Proposition 2.17

Let ∼ be an equivalence relation on 𝐴. Then

1. For each 𝑥 ∈ 𝐴 we have

[𝑥] ≠ ∅

2. For all 𝑥, 𝑦 ∈ 𝐴 it holds

𝑥∼𝑦 ⟺ [𝑥] = [𝑦] .

Proof
Proof of Point 1: By the reflexive property we have 𝑥 ∼ 𝑥. By definition of equivalence class we conclude
that 𝑥 ∈ [𝑥]. This shows
[𝑥] ≠ ∅

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 24

Proof of Point 2: We need to prove a double implication. It is convenient to divide the proof into two
parts.
Part 1: 𝑥 ∼ 𝑦 ⟹ [𝑥] = [𝑦].
Assume 𝑥 ∼ 𝑦. By the definition of an equivalence class
[𝑥] = {𝑧 ∈ 𝐴 ∣ 𝑧 ∼ 𝑥}
is the set of all elements in 𝐴 that are related to 𝑥. Similarly,
[𝑦] = {𝑧 ∈ 𝐴 ∣ 𝑧 ∼ 𝑦} .
We need to show that
[𝑥] = [𝑦] ,
meaning that every element in [𝑥] is also in [𝑦], and vice versa.
1. First, take an arbitrary element 𝑧 ∈ [𝑥].
• By definition, 𝑧 ∼ 𝑥.
• Since ∼ is an equivalence relation, it satisfies the transitive property.
• Therefore, from 𝑧 ∼ 𝑥 and 𝑥 ∼ 𝑦, we can conclude that 𝑧 ∼ 𝑦
• Hence, 𝑧 ∈ [𝑦].
• This shows that [𝑥] ⊆ [𝑦].
2. Now, take an arbitrary element 𝑧 ∈ [𝑦].
• By definition, 𝑧 ∼ 𝑦.
• Since ∼ is an equivalence relation, it satisfies the symmetric property.
• Therefore, from 𝑥 ∼ 𝑦, we also have 𝑦 ∼ 𝑥.
• By the transitive property, from 𝑧 ∼ 𝑦 and 𝑦 ∼ 𝑥, we can conclude that 𝑧 ∼ 𝑥.
• Hence, 𝑧 ∈ [𝑥].
• This shows that [𝑦] ⊆ [𝑥].
Since [𝑥] ⊆ [𝑦] and [𝑦] ⊆ [𝑥], it follows that [𝑥] = [𝑦], as required. Thus, we have shown that 𝑥 ∼ 𝑦 ⟹
[𝑥] = [𝑦].
Part 2: [𝑥] = [𝑦] ⟹ 𝑥 ∼ 𝑦.
Assume [𝑥] = [𝑦]. This means that the equivalence classes of 𝑥 and 𝑦 are the same.
• By point (i) in the Proposition, we have 𝑥 ∈ [𝑥] and 𝑦 ∈ [𝑦].
• Since [𝑥] = [𝑦], we have 𝑥 ∈ [𝑦].
• By the definition of [𝑦], this means 𝑥 ∼ 𝑦.
Thus, we have shown that [𝑥] = [𝑦] ⟹ 𝑥 ∼ 𝑦.
Conclusion: Since we have proven both directions:
• 𝑥 ∼ 𝑦 ⟹ [𝑥] = [𝑦]
• [𝑥] = [𝑦] ⟹ 𝑥 ∼ 𝑦
we conclude that
𝑥 ∼ 𝑦 ⟺ [𝑥] = [𝑦] .
This completes the proof.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 25

The prototypical (and trivial) example of equivalence relation is the equality over ℚ.

Example 2.18: Equality is an equivalence relation

Consider the set of rational numbers ℚ. The equality defines a binary relation on ℚ × ℚ, via

𝑅 ∶= {(𝑥, 𝑦) ∈ ℚ × ℚ ∶ 𝑥 = 𝑦} .

Let us check that 𝑅 is an equivalence relation:

1. Reflexive: It holds, since 𝑥 = 𝑥 for all 𝑥 ∈ ℚ,

2. Symmetric: Again 𝑥 = 𝑦 if and only if 𝑦 = 𝑥,

3. Transitive: If 𝑥 = 𝑦 and 𝑦 = 𝑧 then 𝑥 = 𝑧.

The class of equivalence of 𝑥 ∈ ℚ is given by

[𝑥] = {𝑥} ,

that is, this relation is quite trivial, given that each element of ℚ can only be related to itself. The quotient
space is then
ℚ/𝑅 = {[𝑥] ∶ 𝑥 ∈ ℚ} = {{𝑥} ∶ 𝑥 ∈ ℚ} .

We now give an example of a non-trivial equivalence relation over ℚ.

Example 2.19

Suppose that 𝑅 is a binary relation on the set ℚ of rational numbers defined by

𝑥 ∼ 𝑦 ⟺ 𝑥 − 𝑦 ∈ ℤ.

In other words, we have that

𝑥 ∼ 𝑦 ⟺ 𝑦 = 𝑛 + 𝑥 for some 𝑛 ∈ ℤ ,

which means that 𝑥 and 𝑦 are equivalent iff 𝑦 is a translation of 𝑥 by some integer 𝑛.
We claim that 𝑅 is an equivalence relation on ℚ. Indeed:

1. Reflexive: Let 𝑥 ∈ ℚ. Then 𝑥 − 𝑥 = 0 and 0 ∈ ℤ. Thus 𝑥 ∼ 𝑥.

2. Symmetric: If 𝑥 ∼ 𝑦 then 𝑥 − 𝑦 ∈ ℤ. But then also

−(𝑥 − 𝑦) = 𝑦 − 𝑥 ∈ ℤ
and so 𝑦 ∼ 𝑥.
3. Transitive: Suppose 𝑥 ∼ 𝑦 and 𝑦 ∼ 𝑧. Then

𝑥 − 𝑦 ∈ ℤ and 𝑦 − 𝑧 ∈ ℤ .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 26

Thus we have
𝑥 − 𝑧 = (𝑥 − 𝑦) + (𝑦 − 𝑧) ∈ ℤ
showing that 𝑥 ∼ 𝑧.

We have shown that 𝑅 is an equivalence relation on ℚ.

Question: Can we characterize the equivalence class [𝑥]?
Yes. Note that
𝑦 −𝑥 ∈ℤ
is equivalent to
∃ 𝑛 ∈ ℤ s.t. 𝑦 − 𝑥 = 𝑛
which again is equivalent to
∃ 𝑛 ∈ ℤ s.t. 𝑦 = 𝑥 + 𝑛 .
In summary we have
𝑥∼𝑦 ⟺ ∃ 𝑛 ∈ ℤ s.t. 𝑦 = 𝑥 + 𝑛 .
Therefore the equivalence classes with respect to ∼ are

[𝑥] = {𝑥 + 𝑛 ∶ 𝑛 ∈ ℤ} .

Each equivalence class has exactly one element in [0, 1) ∩ ℚ, meaning that:

∀𝑥 ∈ ℚ , ∃! 𝑞 ∈ ℚ s.t. 0 ≤ 𝑞 < 1 and 𝑞 ∈ [𝑥] . (2.17)

Condition (2.17) is illustrated in Figure 2.1. Indeed: take 𝑥 ∈ ℚ arbitrary. Then 𝑥 ∈ [𝑛, 𝑛 + 1) for some
𝑛 ∈ ℤ. Setting 𝑞 ∶= 𝑥 − 𝑛 we obtain that

𝑥 = 𝑞 +𝑛, 𝑞 ∈ [0, 1) ,

proving (2.17). In particular (2.17) implies that for each 𝑥 ∈ ℚ there exists 𝑞 ∈ [0, 1) ∩ ℚ such that

[𝑥] = [𝑞] .

This allows to conclude that

ℚ/𝑅 = {[𝑥] ∶ 𝑥 ∈ ℚ} = {𝑞 ∈ ℚ ∶ 0 ≤ 𝑞 < 1} .

Figure 2.1: For each 𝑥 ∈ ℚ there exist unique 𝑞 ∈ [0, 1) ∩ ℚ such that 𝑥 = 𝑞 + 𝑛. In particular [𝑥] = [𝑞].

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 27

2.5 Order relation

Order relations are defined similarly to equivalence relations. However notice that symmetry is replaced
by antisymmetry.

Definition 2.20: Partial order

A binary relation 𝑅 on 𝐴 is called a partial order if it satisfies the following properties:

1. Reflexive: For each 𝑥 ∈ 𝐴 one has

(𝑥, 𝑥) ∈ 𝑅 ,

2. Antisymmetric: We have

(𝑥, 𝑦) ∈ 𝑅 and (𝑦, 𝑥) ∈ 𝑅 ⟹ 𝑥 = 𝑦

This is the only new condition with respect to the definition of equivalence relation, and
it replaces symmetry.

3. Transitive: We have
(𝑥, 𝑦) ∈ 𝑅 , (𝑦, 𝑧) ∈ 𝑅 ⟹ (𝑥, 𝑧) ∈ 𝑅

Definition 2.21: Total order

A binary relation 𝑅 on 𝐴 is called a total order relation if it satisfies the following properties:

1. Partial order: 𝑅 is a partial order on 𝐴.

2. Total: For each 𝑥, 𝑦 ∈ 𝐴 we have

(𝑥, 𝑦) ∈ 𝑅 or (𝑦, 𝑥) ∈ 𝑅 .

This is saying that all elements in 𝐴 are related.

The operation of set inclusion is a partial order on 𝑃(Ω) but not a total order.

Example 2.22: Set inclusion is a partial order but not total order

Consider an arbitrary non-empty set Ω and its power set

𝒫 (Ω) = {𝐴 ∶ 𝐴 ⊆ Ω} .
The inclusion defines binary relation on 𝒫 (Ω) × 𝒫 (Ω), via
𝑅 ∶= {(𝐴, 𝐵) ∈ 𝒫 (Ω) × 𝒫 (Ω) ∶ 𝐴 ⊆ 𝐵} .
Let us check that 𝑅 is an order relation:
1. Reflexive: It holds, since 𝐴 ⊆ 𝐴 for all 𝐴 ∈ 𝒫 (Ω),

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 28

2. Transitive: If 𝐴 ⊆ 𝐵 and 𝐵 ⊆ 𝐶, then by definition of inclusion 𝐴 ⊆ 𝐶.

3. Antisymmetric: If 𝐴 ⊆ 𝐵 and 𝐵 ⊆ 𝐴, then 𝐴 = 𝐵 by Proposition 2.7.

Therefore 𝑅 is a partial order on 𝒫 (Ω).

In general 𝑅 is not a total order. For example if we consider

Ω = {𝑥, 𝑦} .

Thus
𝒫 (Ω) = {∅, {𝑥}, {𝑦}, {𝑥, 𝑦}} .
If we pick 𝐴 = {𝑥} and 𝐵 = {𝑦} then 𝐴 ∩ 𝐵 = ∅, meaning that

𝐴 ⊈ 𝐵, 𝐵 ⊈ 𝐴,

showing that 𝑅 is not a total order.

The inequality on ℚ is an example of total order.

Example 2.23: Inequality is a total order

Consider the set of rationals ℚ. The usual inequality defines a binary relation on ℚ × ℚ, via

𝑅 ∶= {(𝑥, 𝑦) ∈ ℚ × ℚ ∶ 𝑥 ≤ 𝑦} .

Let us check that 𝑅 is an order relation:

1. Reflexive: It holds, since 𝑥 ≤ 𝑥 for all 𝑥 ∈ ℚ,

2. Transitive: If 𝑥 ≤ 𝑦 and 𝑦 ≤ 𝑧 then 𝑥 ≤ 𝑧.

3. Antisymmetric: If 𝑥 ≤ 𝑦 and 𝑦 ≤ 𝑥 then 𝑥 = 𝑦.

Finally, we halso have that 𝑅 is a total order on ℚ, since for all 𝑥, 𝑦 ∈ ℚ we have

𝑥 ≤ 𝑦 or 𝑦 ≤ 𝑥 .

Notation 2.24

If 𝐴 is a set and 𝑅 is a total order on 𝐴, we write

(𝑥, 𝑦) ∈ 𝑅 ⟺ 𝑥 ≤ 𝑦 .

Therefore the symbol ≤ will always denote a total order relation.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 29

2.6 Intervals
In this section we assume to have available the set ℝ of real numbers, which we recall is an extension of ℚ.
We now introduce the concept of interval.

Definition 2.25

Let 𝑎, 𝑏 ∈ ℝ with 𝑎 < 𝑏. We define the open interval (𝑎, 𝑏) as the set

(𝑎, 𝑏) ∶= {𝑥 ∈ ℝ ∶ 𝑎 < 𝑥 < 𝑏} .

We define the close interval [𝑎, 𝑏] as the set

[𝑎, 𝑏] ∶= {𝑥 ∈ ℝ ∶ 𝑎 ≤ 𝑥 ≤ 𝑏} .

In general we also define the intervals

[𝑎, 𝑏) ∶= {𝑥 ∈ℝ∶ 𝑎 ≤ 𝑥 < 𝑏} , (2.18)

(𝑎, 𝑏] ∶= {𝑥 ∈ℝ∶ 𝑎 ≤ 𝑥 ≤ 𝑏} , (2.19)
(𝑎, ∞) ∶= {𝑥 ∈ℝ∶ 𝑥 > 𝑎} , (2.20)
[𝑎, ∞) ∶= {𝑥 ∈ℝ∶ 𝑥 ≥ 𝑎} , (2.21)
(−∞, 𝑏) ∶= {𝑥 ∈ℝ∶ 𝑥 < 𝑏} , (2.22)
(−∞, 𝑏] ∶= {𝑥 ∈ℝ∶ 𝑥 ≤ 𝑏} . (2.23)

Some of the above intervals are depicted in Figure 2.2, Figure 2.3, Figure 2.4, Figure 2.5 below.

Figure 2.2: Interval (𝑎, 𝑏)

Figure 2.3: Interval [𝑎, 𝑏]

Figure 2.4: Interval (𝑎, ∞)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 30

Figure 2.5: Interval (−∞, 𝑏]

2.7 Functions

Definition 2.26: Functions

Let 𝐴 and 𝐵 be sets. A function from 𝐴 to 𝐵 is a rule which associates at each element 𝑥 ∈ 𝐴 a single
element 𝑦 ∈ 𝐵. Notations:

• We write
𝑓 ∶𝐴→𝐵
to indicate such rule,
• For 𝑥 ∈ 𝐴, we denote by
𝑦 ∶= 𝑓 (𝑥) ∈ 𝐵
the element associated with 𝑥 by 𝑓 .
• We will often denote the map 𝑓 also by

𝑥 ↦ 𝑓 (𝑥) .

In addition:

• The set 𝐴 is called the domain of 𝑓 ,

• The range or image of 𝑓 is the set

𝑓 (𝐴) ∶= {𝑦 ∈ 𝐵 ∶ 𝑦 = 𝑓 (𝑥) for some 𝑥 ∈ 𝐴} ⊆ 𝐵 .

Warning

We want to stress the importance of the first two sentences in Definition 2.26. Assume that 𝑓 ∶ 𝐴 → 𝐵
is a function. Then:

• To each element 𝑥 ∈ 𝐴 we can only associate one element 𝑓 (𝑥) ∈ 𝐵,

• Every element 𝑥 ∈ 𝐴 has to be associated to an element 𝑓 (𝑥) ∈ 𝐵.

Example 2.27

Assume given the two sets

𝐴 = {𝑎1 , 𝑎2 } , 𝐵 = {𝑏1 , 𝑏2 , 𝑏3 } .
Let us see a few examples:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 31

1. Define 𝑓 ∶ 𝐴 → 𝐵 by setting
𝑓 (𝑎1 ) = 𝑏1 , 𝑓 (𝑎2 ) = 𝑏1 .
In this way 𝑓 is a function, with domain 𝐴 and range

𝑓 (𝐴) = {𝑏1 } ⊆ 𝐵 .

2. Define 𝑔 ∶ 𝐴 → 𝐵 by setting

𝑔(𝑎1 ) = 𝑏2 , 𝑔(𝑎1 ) = 𝑏3 , 𝑔(𝑎2 ) = 𝑏3

Then 𝑔 is NOT a function, since the element 𝑎1 has two elements associated.

3. Define ℎ ∶ 𝐴 → 𝐵 by setting
ℎ(𝑎1 ) = 𝑏1 .
Then 𝑔 is NOT a function, since the element 𝑎2 has no element associated.

Figure 2.6: Schematic picture of the function 𝑓

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 32

Figure 2.7: Schematic picture of the function 𝑔

Figure 2.8: Schematic picture of the function ℎ

Example 2.28

Let us make two examples of functions on ℝ:

1. Define 𝑓 ∶ ℝ → ℝ by
𝑓 (𝑥) = 𝑥 2 .
Note that the domain of 𝑓 is given by ℝ, while the range is
𝑓 (ℝ) = [0, ∞) .

2. Define 𝑔 ∶ ℝ → ℝ as the logarithm:

𝑔(𝑥) = log(𝑥) .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 33

This time the domain is (0, ∞), while the range is 𝑔(ℝ) = ℝ.

Figure 2.9: Plot of function 𝑓 (𝑥) = 𝑥 2

Figure 2.10: Plot of function 𝑔(𝑥) = log(𝑥)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 34

2.8 Absolute value or Modulus

In this section we assume to have available the set ℝ of real numbers, which we recall is an extension of
ℚ.

Definition 2.29: Absolute value

For 𝑥 ∈ ℝ we define its absolute value as the quantity

𝑥 if 𝑥 ≥ 0
|𝑥| = {
−𝑥 if 𝑥 < 0

Example 2.30

By definition one has |𝑥| = 𝑥 if 𝑥 ≥ 0. For example

|𝜋| = 𝜋 , |√2| = √2 , |0| = 0 .

Instead |𝑥| = −𝑥 if 𝑥 < 0. For example

| − 𝜋| = 𝜋 , |−√2| = √2 , | − 10| = 10 .

Let us also make the following basic Remark. The proof will be left as an exercise.

Remark 2.31

For all 𝑥 ∈ ℝ
|𝑥| ≥ 0 .
Moreover
|𝑥| = 0 ⟺ 𝑥 = 0 .

Another basic remark (proof by exercise).

Remark 2.32

For all 𝑥 ∈ ℝ one has

|𝑥| = | − 𝑥| .

We can use the definition of absolute value to define the absolute value function. This is the function

𝑓 ∶ ℝ → ℝ, 𝑓 (𝑥) ∶= |𝑥| .

You might be familiar with the graph associated to 𝑓 , as seen below.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 35

Figure 2.11: Plot of the absolute value function 𝑓 (𝑥) = |𝑥|

It is also useful to understand the absolute value in a geometric way.

Remark 2.33: Geometric interpretation of |𝑥|

A number 𝑥 ∈ ℝ can be represented with a point on the real line ℝ.The non-negative number |𝑥| represents
the distance of 𝑥 from the origin 0. Notice that this works for both positive and negative numbers 𝑥1
and 𝑥2 respectively, as shown in Figure 2.12 below.

Figure 2.12: Geometric interpretation of |𝑥|

Remark 2.34: Geometric interpretation of |𝑥 − 𝑦|

If 𝑥, 𝑦 ∈ ℝ then the number |𝑥 − 𝑦| represents the distance between 𝑥 and 𝑦 on the real line, as shown in
Figure 2.13 below. Note that by Remark 2.32 we have

|𝑥 − 𝑦| = |𝑦 − 𝑥| .

In the next Lemma we show a fundamental equivalence regarding the absolute value.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 36

Figure 2.13: Geometric interpretation of |𝑥 − 𝑦|

Lemma 2.35

Let 𝑥, 𝑦 ∈ ℝ. Then
|𝑥| ≤ 𝑦 ⟺ −𝑦 ≤ 𝑥 ≤ 𝑦 .

The geometric meaning of the above statement is clear: the distance of 𝑥 from the origin is less than 𝑦, in
formulae
|𝑥| ≤ 𝑦 ,
if and only if 𝑥 belongs to the interval [−𝑦, 𝑦], in formulae

−𝑦 ≤ 𝑥 ≤ 𝑦 .

A sketch of this explanation is seen in Figure 2.14 below.

Figure 2.14: Geometric meaning of Lemma 2.35

Proof: Proof of Lemma 2.35

Step 1: First implication.

Suppose first that
|𝑥| ≤ 𝑦 . (2.24)
Recalling that the absolute value is non-negative, from (2.24) we deduce that 0 ≤ |𝑥| ≤ 𝑦. In particular it
holds
𝑦 ≥ 0. (2.25)
We make separate arguments for the cases 𝑥 ≥ 0 and 𝑥 < 0:

• Case 1: 𝑥 ≥ 0. From (2.24), (2.25) and from 𝑥 ≥ 0 we have

−𝑦 ≤ 0 ≤ 𝑥 = |𝑥| ≤ 𝑦

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 37

which shows
−𝑦 ≤ 𝑥 ≤ 𝑦 .
• Case 2: 𝑥 < 0. From (2.24), (2.25) and from 𝑥 < 0 we have

−𝑦 ≤ 0 < −𝑥 = |𝑥| ≤ 𝑦

which shows
−𝑦 ≤ −𝑥 ≤ 𝑦 .
Multiplying the above inequalities by −1 yields

−𝑦 ≤ 𝑥 ≤ 𝑦 .

Step 2: Second implication.

Suppose now that

−𝑦 ≤ 𝑥 ≤ 𝑦 . (2.26)
We make separate arguments for the cases 𝑥 ≥ 0 and 𝑥 < 0:

• Case 1: 𝑥 ≥ 0. Since 𝑥 ≥ 0, from (2.26) we get

|𝑥| = 𝑥 ≤ 𝑦

showing that
|𝑥| ≤ 𝑦 .
• Case 2: 𝑥 < 0. Since 𝑥 < 0, from (2.26) we have

−𝑦 ≤ 𝑥 = −|𝑥| .

Multiplying the above inequality by −1 yields

|𝑥| ≤ 𝑦 .

With the same arguments, just replacing ≤ with <, one can also show the following.

Corollary 2.36

Let 𝑥, 𝑦 ∈ ℝ. Then
|𝑥| < 𝑦 ⟺ −𝑦 < 𝑥 < 𝑦 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 38

2.9 Triangle inequality

The triangle inequality relates the absolute value to the sum operation. It is a very important inequality,
which we will use a lot in the future.

Theorem 2.37: Triangle inequality

For every 𝑥, 𝑦 ∈ ℝ we have

||𝑥| − |𝑦|| ≤ |𝑥 + 𝑦| ≤ |𝑥| + |𝑦| . (2.27)

Before proceeding with the proof, let us discuss the geometric meaning of the triangle inequality.

Remark 2.38: Geometric meaning of triangle inequality

The notion of absolute value can be extended also to vectors in the plane. Suppose that 𝑥 and 𝑦 are two
vectors in the plane, as in Figure 2.15 below. Then |𝑥| and |𝑦| can be interpreted as the lengths of these
vectors.
Using the rule of sum of vectors, we can draw 𝑥 + 𝑦, as shown in Figure 2.16 below. From the picture it
is evident that
|𝑥 + 𝑦| ≤ |𝑥| + |𝑦| , (2.28)
that is, the length of each side of a triangle does not exceed the sum of the lengths of the two remaining sides.
Note that (2.28) is exactly the second inequality in (2.27). This is why (2.27) is called triangle inequality.

Figure 2.15: Vectors 𝑥 and 𝑦

Proof: Proof of Theorem 2.37

Assume that 𝑥, 𝑦 ∈ ℝ. We prove the two inequalities in (2.27) individually.

Step 1. Proof of the second inequality in (2.27).
Trivially we have
|𝑥| ≤ |𝑥| .
Therefore we can apply Lemma 2.35 and infer

−|𝑥| ≤ 𝑥 ≤ |𝑥| . (2.29)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 39

Figure 2.16: Summing the vectors 𝑥 and 𝑦. The triangle inequality relates the length of 𝑥 + 𝑦 to the length of
𝑥 and 𝑦

Similarly we have that |𝑦| ≤ |𝑦|, and so Lemma 2.35 implies

−|𝑦| ≤ 𝑦 ≤ |𝑦| . (2.30)

Summing (2.29) and (2.30) we get

−(|𝑥| + |𝑦|) ≤ 𝑥 + 𝑦 ≤ |𝑥| + |𝑦| .

We can now again apply Lemma 2.35 to get

|𝑥 + 𝑦| ≤ |𝑥| + |𝑦| , (2.31)

which is the second inequality in (2.27).

Step 2. Proof of the first inequality in (2.27).
Note that the trivial identity
𝑥 =𝑥 +𝑦 −𝑦
always holds. We then have

|𝑥| = |𝑥 + 𝑦 − 𝑦| (2.32)
= |(𝑥 + 𝑦) + (−𝑦)| (2.33)
= |𝑎 + 𝑏| (2.34)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 40

with 𝑎 = 𝑥 + 𝑦 and 𝑏 = −𝑦. We can now apply (2.31) to 𝑎 and 𝑏 to obtain

|𝑥| = |𝑎 + 𝑏| (2.35)
≤ |𝑎| + |𝑏| (2.36)
= |𝑥 + 𝑦| + | − 𝑦| (2.37)
= |𝑥 + 𝑦| + |𝑦| (2.38)

Therefore
|𝑥| − |𝑦| ≤ |𝑥 + 𝑦| . (2.39)
We can now swap 𝑥 and 𝑦 in (2.39) to get

|𝑦| − |𝑥| ≤ |𝑥 + 𝑦| .

By rearranging the above inequality we obtain

−|𝑥 + 𝑦| ≤ |𝑥| − |𝑦| . (2.40)

Putting together (2.39) and (2.40) yields

−|𝑥 + 𝑦| ≤ |𝑥| − |𝑦| ≤ |𝑥 + 𝑦| .

By Lemma 2.35 the above is equivalent to

||𝑥| − |𝑦|| ≤ |𝑥 + 𝑦| ,

which is the first inequality in (2.27).

An immediate consequence of the triangle inequality are the following inequalities, which are left as an
exercise.

Remark 2.39

For any 𝑥, 𝑦 ∈ ℝ it holds

||𝑥| − |𝑦|| ≤ |𝑥 − 𝑦| ≤ |𝑥| + |𝑦| . (2.41)
Moreover for any 𝑥, 𝑦, 𝑧 ∈ ℝ it holds

|𝑥 − 𝑦| ≤ |𝑥 − 𝑧| + |𝑧 − 𝑦| .

Notice that the inequality in (2.41) differs from the triangle inequality (2.27) by a sign. Indeed it can be shown
tha (2.27) and (2.41) are equivalent.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 41

2.10 Proofs in Mathematics

In a mathematical proof one needs to show that
𝛼 ⟹ 𝛽 (2.42)
where
• 𝛼 is a given set of assumptions, or Hypothesis
• 𝛽 is a conclusion, or Thesis
Proving (2.42) means convincing ourselves that 𝛽 follows from 𝛼. Common strategies to prove (2.42) are:
1. Contradiction: Assume that the thesis is false, and hope to reach a contradiction: that is, prove that
¬𝛽 ⟹ contradiction
where ¬𝛽 is the negation of 𝛽.
For example we already proved by contradiction that
Definition of ℚ ⟹ √2 ∉ ℚ ,
In the above statement
𝛼 = (Definition of ℚ) .
𝛽 = (√2 ∉ ℚ) .
Therefore
¬𝛽 = (√2 ∈ ℚ) .

2. Direct: Sometimes proofs will also need direct arguments, meaning that one need to show directly
that (2.42) holds.
3. Contrapositive: The statement (2.42) is equivalent to
¬𝛽 ⟹ ¬𝛼 . (2.43)
Thus, instead of proving (2.42), one could show (2.43). The statement (2.43) is called the contrapositive
of (2.42).
Let us make an example.

Proposition 2.40

Two real numbers 𝑎, 𝑏 are equal if and only if for every real number 𝜀 > 0 it follows that |𝑎 − 𝑏| < 𝜀.

Before proceeding with the proof, note that the above stetement is just saying that:
Two numbers are equal if and only if they are arbitrarily close
By arbitrarily close we mean that they are as close as you want the to be.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 42

Proof: of Proposition 2.40

Let us first rephrase the statement using mathematical symbols:

Let 𝑎, 𝑏 ∈ ℝ. Then it holds:
𝑎 = 𝑏 ⟺ |𝑎 − 𝑏| < 𝜀 , ∀ 𝜀 > 0 .

Setting
𝛼 = (𝑎 = 𝑏) (2.44)
𝛽 = (|𝑎 − 𝑏| < 𝜀 , ∀ 𝜀 > 0) (2.45)
the statement is equivalent to
𝛼 ⟺ 𝛽.
To show the above, it is sufficient to show that
𝛼 ⟹ 𝛽 and 𝛽 ⟹ 𝛼.
Step 1. Proof that 𝛼 ⟹ 𝛽.
This proof can be carried out by a direct argument. Since we are assuming 𝛼, this means
𝑎 = 𝑏.
We want to see that 𝛽 holds. Therefore fix an arbitrary 𝜀 > 0. This means that 𝜀 can be any positive
number, as long as you fix it. Clearly
|𝑎 − 𝑏| = |0| = 0 < 𝜀
since 𝑎 = 𝑏, |0| = 0, and 𝜀 > 0. The above shows that
|𝑎 − 𝑏| < 𝜀 .
As 𝜀 > 0 was arbitrary, we have just proven that
|𝑎 − 𝑏| < 𝜀 , ∀ 𝜀 > 0 ,
meaning that 𝛽 holds and the proof is concluded.
Step 2. Proof that 𝛽 ⟹ 𝛼.
Let us prove this implication by showing the contrapositive
¬𝛼 ⟹ ¬𝛽 .
So let us assume ¬𝛼 is true. This means that
𝑎 ≠ 𝑏.
We have to see that ¬𝛽 holds. But ¬𝛽 means that
∃ 𝜀0 > 0 s.t. |𝑎 − 𝑏| ≥ 𝜀0 .
The above is satisfied by choosing
𝜀0 ∶= |𝑎 − 𝑏| ,
since 𝜀0 > 0 given that 𝑎 ≠ 𝑏.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 43

2.11 Induction
Another technique for carrying out proofs is induction, which we take as an axiom.

Axiom 2.41: Principle of Induction

Let 𝑆 ⊆ ℕ. Suppose that

1. We have 1 ∈ 𝑆, and
2. Whenever 𝑛 ∈ 𝑆, then (𝑛 + 1) ∈ 𝑆.

Then we have
𝑆 = ℕ.

Important

The above is an axiom, meaning that we do not prove it, but rather we just assume it holds.

Remark 2.42

It would be possible to prove the Principle of Induction starting from elementary axioms for ℕ, called
the Peano Axioms, see the Wikipedia page.
However, in justifying basic principles of mathematics, one at some point needs to draw a line. This
means that something which looks elementary needs to be assumed to hold, in order to have a starting
point for proving deeper statements.
In the case of the Principle of Induction, the intuition is clear:

The Principle of Induction is just describing the domino effect: If one tile falls, then the next
one will fall as well. Therefore if the first tile falls, all the tiles will fall.

It seems reasonable to assume such evident principle.

The Principle of Induction can be used to prove statements which depend on some index 𝑛 ∈ ℕ. Precisely,
the following statement holds.

Corollary 2.43: Principle of Inducion - Alternative formulation

Let 𝛼(𝑛) be a statement which depends on 𝑛 ∈ ℕ. Suppose that

1. 𝛼(1) is true, and

2. Whenever 𝛼(𝑛) is true, then 𝛼(𝑛 + 1) is true.

Then 𝛼(𝑛) is true for all 𝑛 ∈ ℕ.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 44

Proof
Define the set
𝑆 ∶= {𝑛 ∈ ℕ s.t. 𝛼(𝑛) is true} .
Then

1. We have 1 ∈ 𝑆, since 𝛼(1) is true.

2. If 𝑛 ∈ 𝑆 then 𝛼(𝑛) is true. By assumption this implies that 𝛼(𝑛 + 1) is true. Therefore (𝑛 + 1) ∈ 𝑆.

Therefore 𝑆 satisfies the assumptions of the Induction Principle and we conclude that

𝑆 = ℕ.

By definition this means that 𝛼(𝑛) is true for all 𝑛 ∈ ℕ.

Example 2.44: Formula for summing first 𝑛 natural numbers

Using the Principle of Induction we can prove that

𝑛(𝑛 + 1)
1 + 2 + 3 + … + (𝑛 − 1) + 𝑛 = (2.46)
2
holds for all 𝑛 ∈ ℕ.
Proof. To be really precise, consider the statement
𝛼(𝑛) ∶= the above formula is true for 𝑛 .
In order to apply induction, we need to show that
1. 𝛼(1) is true,
2. If 𝛼(𝑛) is true then 𝛼(𝑛 + 1) is true.
Let us proceed:
1. It is immediate to check that (2.46) holds for 𝑛 = 1.
2. Suppose (2.46) holds for 𝑛. Then
𝑛(𝑛 + 1)
1 + … + 𝑛 + (𝑛 + 1) = + (𝑛 + 1) (2.47)
2
𝑛(𝑛 + 1) + 2(𝑛 + 1)
= (2.48)
2
(𝑛 + 1)(𝑛 + 2)
= (2.49)
𝑛
where in the first equality we used that (2.46) holds for 𝑛. We then have
(𝑛 + 1)(𝑛 + 2)
1 + … + 𝑛 + (𝑛 + 1) = ,
𝑛
which shows that (2.46) holds for 𝑛 + 1.
By the Principle of Induction we then conclude that 𝛼(𝑛) is true for all 𝑛 ∈ ℕ, which means that (2.46)
holds for all 𝑛 ∈ ℕ.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 45

Example 2.45: Statements about sequences of numbers

Suppose you are given a collection of numbers

{𝑥𝑛 s.t. 𝑛 ∈ ℕ} .

Such collection of numbers is called sequence. Assume that

𝑥1 ∶= 1 (2.50)
𝑥
𝑥𝑛+1 ∶= 𝑛 + 1 . (2.51)
2
A sequence defined as above is called recurrence sequence. Using the above rule we can compute all
the terms of 𝑥𝑛 .

For example
𝑥1 1 3
𝑥2 = +1= +1= (2.52)
2 2 2
𝑥 3 7
𝑥3 = 2 + 1 = +1= . (2.53)
2 4 4

By computing these terms, we suspect that the sequence might be increasing, meaning that

𝑥𝑛+1 ≥ 𝑥𝑛 (2.54)

for all 𝑛 ∈ ℕ.
Claim. (2.54) holds for all 𝑛 ∈ ℕ.
Proof of Claim.
We argue by induction:

1. We have seen that 𝑥1 = 1 and 𝑥2 = 3/2. Thus

𝑥2 ≥ 𝑥1 .

2. Suppose now that

𝑥𝑛+1 ≥ 𝑥𝑛 . (2.55)
We need to prove that
𝑥𝑛+2 ≥ 𝑥𝑛+1 . (2.56)
Indeed, we can multiply the inequality (2.55) by 1/2 and add 1 to get
𝑥𝑛+1 𝑥
+ 1 ≥ 𝑛 + 1.
2 2
The above is equivalent, by definition, to (2.56).

Therefore the assumptions of the Induction Principle are satisfied, and (2.54) follows.

Dr. Silvio Fanzon [email protected]

3 Real Numbers
In this chapter we introduce the system of Real Numbers ℝ and study some of its properties.

3.1 Fields
In order to introduce ℝ, we need the concepts of binary operation and field. We proceed in a general setting,
starting from a set 𝐾 .

Definition 3.1: Binary operation

A binary operation on a set 𝐾 is a function

∘ ∶ 𝐾 ×𝐾 →𝐾

which maps the ordered pair (𝑥, 𝑦) into 𝑥 ∘ 𝑦.

Notation 3.2

There are two main binary operations we are interested in:

• Addition: denoted by +. The addition, or sum of 𝑥, 𝑦 ∈ 𝐾 is denoted by

𝑥 +𝑦.

• Multiplication: denoted by ⋅. The multiplication, or product of 𝑥, 𝑦 ∈ 𝐾 is denoted by

𝑥 ⋅ 𝑦 or 𝑥𝑦 .

Example 3.3: of binary operation

Let 𝐾 = {0, 1}. We can for example define operations of sum and product on 𝐾 according to the tables
+ 0 1 ⋅ 0 1
0 0 1 0 0 0
1 1 0 1 0 1
The above mean that
0 + 0 = 1 + 1 = 0, 0 + 1 = 1 + 0 = 1,

46
Numbers, Sequences and Series Page 47

0 ⋅ 0 = 0 ⋅ 1 = 1 ⋅ 0 = 0, 1 ⋅ 1 = 1.
This is just one option. Note that we could not have defined

1 + 1 = 2,

since 2 ∉ 𝐾 .

Binary operations take ordered pairs of elements of 𝐾 as input. Therefore the operation

𝑥 ∘𝑦 ∘𝑧

does not make sense, since we do not know which one between

𝑥 ∘𝑦 or 𝑦 ∘𝑧

has to be performed first. Moreover the outcome of an operation depends on order:

𝑥 ∘𝑦 ≠𝑦 ∘𝑥.

This motivates the following definition.

Definition 3.4

Let 𝐾 be a set and ∘ ∶ 𝐾 × 𝐾 → 𝐾 be a binary operation on 𝐾 . We say that:

1. ∘ is commutative if
𝑥 ∘𝑦 =𝑦 ∘𝑥, ∀ 𝑥, 𝑦 ∈ 𝐾
2. ∘ is associative if
(𝑥 ∘ 𝑦) ∘ 𝑧 = 𝑥 ∘ (𝑦 ∘ 𝑧) , ∀ 𝑥, 𝑦, 𝑧 ∈ 𝐾
3. An element 𝑒 ∈ 𝐾 is called neutral element of ∘ if

𝑥 ∘𝑒 =𝑒∘𝑥 =𝑥, ∀𝑥 ∈ 𝐾

4. Let 𝑒 be a neutral element of ∘ and let 𝑥 ∈ 𝐾 . An element 𝑦 ∈ 𝐾 is called an inverse of 𝑥 with

respect to ∘ if
𝑥 ∘𝑦 = 𝑦 ∘𝑥 = 𝑒.

Example 3.5

Let 𝐾 with + and ⋅ be as in Example 0.55. The sum satisfies:

• + is commutative, since
0 + 1 = 1 + 0 = 0.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 48

• + is associative, since for example

(0 + 1) + 1 = 1 + 1 = 0 , 0 + (1 + 1) = 0 + 0 = 0 ,

and therefore
(0 + 1) + 1 = 0 + (1 + 1) .
In general one can show that + is associative by checking all the other permutations.
• The neutral element of + is 0, since

0 + 0 = 0, 1 + 0 = 0 + 1 = 1.

• Every element has an inverse. Indeed, the inverse of 0 is 0, since

0 + 0 = 0,

while the inverse of 1 is 1, since

1 + 1 = 1 + 1 = 0.

The multiplication satisfies:

• ⋅ is commutative, since
1 ⋅ 0 = 0 ⋅ 1 = 0.
• ⋅ is associative, since for example

(0 ⋅ 1) ⋅ 1 = 0 ⋅ 1 = 0 , 0 ⋅ (1 ⋅ 1) = 0 ⋅ 1 = 0 ,

and therefore
(0 ⋅ 1) ⋅ 1 = 0 ⋅ (1 ⋅ 1) .
By checking all the other permutations one can show that ⋅ is associative.
• The neutral element of ⋅ is 1, since

0 ⋅ 1 = 1 ⋅ 0 = 0, 1 ⋅ 1 = 1.

• The element 0 has no inverse, since

0 ⋅ 0 = 0 ⋅ 1 = 1 ⋅ 0 = 0,
and thus we never obtain the neutral element 1. The inverse of 1 is given by 1, since

1 ⋅ 1 = 1.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 49

Example 3.6

Let 𝐾 = {0, 1} be a set with binary relation ∘ defined by the table

∘ 0 1
0 1 1
1 0 0

In this case ∘ is not commutative since

0 ∘ 1 = 1, 1∘0=0

and therefore
0 ∘ 1 ≠ 1 ∘ 0.
Moreover ∘ is not associative, since
(0 ∘ 1) ∘ 1 = 1 ∘ 1 = 0 ,
while
0 ∘ (1 ∘ 1) = 0 ∘ 0 = 1 ,
so that
(0 ∘ 1) ∘ 1 ≠ 0 ∘ (1 ∘ 1) .

We are ready to define fields.

Definition 3.7: Field

Let 𝐾 be a set with binary operations of addition

+ ∶ 𝐾 ×𝐾 →𝐾, (𝑥, 𝑦) ↦ 𝑥 + 𝑦
and multiplication
⋅ ∶ 𝐾 ×𝐾 →𝐾, (𝑥, 𝑦) ↦ 𝑥 ⋅ 𝑦 = 𝑥𝑦 .
We call the triple (𝐾 , +, ⋅) a field if:
1. The addition + satisfies: ∀ 𝑥, 𝑦, 𝑧 ∈ 𝐾
• (A1) Commutativity and Associativity:
𝑥 +𝑦 =𝑦 +𝑥
(𝑥 + 𝑦) + 𝑧 = 𝑥 + (𝑦 + 𝑧)
• (A2) Additive Identity: There exists a neutral element in 𝐾 for +, which we call 0. It holds:
𝑥 +0=0+𝑥 =𝑥
• (A3) Additive Inverse: There exists an inverse of 𝑥 with respect to +. We call this element
the additive inverse of 𝑥 and denote it by −𝑥. It holds
𝑥 + (−𝑥) = (−𝑥) + 𝑥 = 0

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 50

2. The multiplication ⋅ satisifes: ∀ 𝑥, 𝑦, 𝑧 ∈ 𝐾

• (M1) Commutativity and Associativity:

𝑥 ⋅𝑦 =𝑦 ⋅𝑥

(𝑥 ⋅ 𝑦) ⋅ 𝑧 = 𝑥 ⋅ (𝑦 ⋅ 𝑧)
• (M2) Multiplicative Identity: There exists a neutral element in 𝐾 for ⋅, which we call 1.
It holds:
𝑥 ⋅1=1⋅𝑥 =𝑥
• (M3) Multiplicative Inverse: If 𝑥 ≠ 0 there exists an inverse of 𝑥 with respect to ⋅. We call
this element the multiplicative inverse of 𝑥 and denote it by 𝑥 −1 . It holds

𝑥 ⋅ 𝑥 −1 = 𝑥 −1 ⋅ 𝑥 = 1

3. The operations + and ⋅ are related by

• (AM) Distributive Property: ∀ 𝑥, 𝑦, 𝑧 ∈ 𝐾

𝑥 ⋅ (𝑦 + 𝑧) = (𝑥 ⋅ 𝑦) + (𝑦 ⋅ 𝑧) .

Example 3.8

Let 𝐾 with + and ⋅ be as in Example 0.55. We can show that (𝐾 , +, ⋅) is a field. Indeed we have already
shown in Example 3.5 that:
• (A1) and (M1) hold,
• (A2) holds with neutral element 0,
• (M2) holds with neutral element 1,
• (A3) every element has an additive inverse, with
−0 = 0 , −1 = 1 ,
• (M3) every element which is not 0 a multiplicative inverse, with
1−1 = 1 .
We are left to show the Distributive Property (AM). Indeed:
• (AM) For all 𝑦, 𝑧 ∈ 𝐾 we have
0 ⋅ (𝑦 + 𝑧) = 0 , (0 ⋅ 𝑦) + (0 ⋅ 𝑧) = 0 + 0 = 0 ,
and also
1 ⋅ (𝑦 + 𝑧) = 𝑦 + 𝑧 , (1 ⋅ 𝑦) + (1 ⋅ 𝑧) = 𝑦 + 𝑧 .
Thus (AM) holds.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 51

Definition 3.9: Subtraction and division

Let (𝐾 , +, ⋅) be a field. We define:

• Subtraction as the operation − defined by

𝑥 − 𝑦 ∶= 𝑥 + (−𝑦) , ∀ 𝑥, 𝑦 ∈ 𝐾 ,

where −𝑦 is the additive inverse of 𝑦.

• Division as the operation / defined by

𝑥/𝑦 ∶= 𝑥 ⋅ 𝑦 −1 , ∀ 𝑥, 𝑦 ∈ 𝐾 , 𝑦 ≠ 0 ,

where 𝑦 −1 is the multiplicative inverse of 𝑦.

Proposition 3.10: Uniqueness of neutral elements and inverses

Let (𝐾 , +, ⋅) be a field. Then

1. There is a unique element in 𝐾 with the property of 0,

2. There is a unique element in 𝐾 with the property of 1,
3. For all 𝑥 ∈ 𝐾 there is a unique additive inverse −𝑥,
4. For all 𝑥 ∈ 𝐾 , 𝑥 ≠ 0, there is a unique multiplicative inverse 𝑥 −1 .

Proof

1. Suppose that 0 ∈ 𝐾 and 0̃ ∈ 𝐾 are both neutral element of +, that is, they both satisfy (A2). Then

0 + 0̃ = 0

since 0̃ is a neutral element for +. Moreover

0̃ + 0 = 0̃

since 0 is a neutral element for +. By commutativity of +, see property (A1), we have

0 = 0 + 0̃ = 0̃ + 0 = 0̃ ,

showing that 0 = 0.̃ Hence the neutral element for + is unique.

2. Exercise.
3. Let 𝑥 ∈ 𝐾 and suppose that 𝑦, 𝑦̃ ∈ 𝐾 are both additive inverses of 𝑥, that is, they both satisfy (A3).
Therefore
𝑥 +𝑦 =0
since 𝑦 is an additive inverse of 𝑥 and
𝑥 + 𝑦̃ = 0

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 52

since 𝑦̃ is an additive inverse of 𝑥. Therefore we can use commutativity and associativity and of +,
see property (A1), and the fact that 0 is the neutral element of +, to infer

𝑦 = 𝑦 + 0 = 𝑦 + (𝑥 + 𝑦)̃
= (𝑦 + 𝑥) + 𝑦̃ = (𝑥 + 𝑦) + 𝑦̃
= 0 + 𝑦̃ = 𝑦̃ ,

concluding that 𝑦 = 𝑦.̃ Thus there is a unique additive inverse of 𝑥, and

𝑦 = 𝑦̃ = −𝑥 ,

with −𝑥 the element from property (A3).

4. Exercise.

Using the properties of field we can also show that the usual properties of sum, subtraction, multiplication
and division still hold in any field. We list such properties in the following proposition.

Proposition 3.11: Properties of field operations

Let (𝐾 , +, ⋅) be a field. Then for all 𝑥, 𝑦, 𝑧 ∈ 𝐾 ,

• 𝑥 +𝑦 =𝑥 +𝑧 ⟹ 𝑦 =𝑧
• 𝑥 ⋅ 𝑦 = 𝑥 ⋅ 𝑧 and 𝑥 ≠ 0 ⟹ 𝑦 = 𝑧
• −0 = 0
• 1−1 = 1
• 𝑥 ⋅0=0
• −1 ⋅ 𝑥 = −𝑥
• −(−𝑥) = 𝑥
• (𝑥 −1 )−1 = 𝑥 if 𝑥 ≠ 0
• (𝑥 ⋅ 𝑦)−1 = 𝑥 −1 ⋅ 𝑦 −1

The above properties can be all proven with elementary use of the field properties (A1)-(A3), (M1)-(M3) and
(AM). This is an exercise in patience, and is left to the reader.
Let us conclude with examining the sets of numbers introduced in Chapter 1.

Theorem 3.12

Consider the sets ℕ, ℤ, ℚ with the usual operations + and ⋅. We have:

• (ℕ, +, ⋅) is not a field:

It satisfies properties (A1), (A2), (M1), (M2), (AM) of fields. It is missing properties (A3) and (M3),
the additive and multiplicative inverse properties, respectively.
• (ℤ, +, ⋅) is not a field:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 53

It satisfies properties (A1), (A2), (A3), (M1), (M2), (AM) of fields. Thus it is only missing (M3), the
multiplicative inverse property.
• (ℚ, +, ⋅) is a field.

The proof is omitted.

3.2 Ordered fields

Definition 3.13

Let 𝐾 be a set with binary operations + and ⋅, and with an order relation ≤. We call (𝐾 , +, ⋅, ≤) an ordered
field if:

1. (𝐾 , +, ⋅) is a field

2. There ≤ is of total order on 𝐾 : ∀ 𝑥, 𝑦, 𝑧 ∈ 𝐾

• (O1) Reflexivity:
𝑥≤𝑥
• (O2) Antisymmetry:
𝑥 ≤ 𝑦 and 𝑦 ≤ 𝑥 ⟹ 𝑥 = 𝑦
• (O3) Transitivity:
𝑥 ≤ 𝑦 and 𝑦 ≤ 𝑧 ⟹ 𝑥 = 𝑧
• (O4) Total order:

𝑥 ≤ 𝑦 or 𝑦 ≤ 𝑥

3. The operations + and ⋅, and the total order ≤, are related by the following properties: ∀𝑥, 𝑦, 𝑧 ∈ 𝐾
• (AM) Distributive: Relates addition and multiplication via

𝑥 ⋅ (𝑦 + 𝑧) = 𝑥 ⋅ 𝑦 + 𝑥 ⋅ 𝑧

• (AO) Relates addition and order with the requirement:

𝑥 ≤𝑦 ⟹ 𝑥 +𝑧 ≤𝑦 +𝑧

• (MO) Relates multiplication and order with the requirement:

𝑥 ≥ 0, 𝑦 ≥ 0 ⟹ 𝑥 ⋅ 𝑦 ≥ 0

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 54

Example 3.14

(ℚ, +, ⋅, ≤) is an ordered field.

3.3 Cut Property

We have just introduced the notion of field, and noted that the set of rational numbers with the usual opera-
tions
(ℚ, +, ⋅)
is a field.
We now need to address the key issue we proved in Chapter 1, that is, that

√2 ∉ ℚ .

This means that ℚ has gaps, and cannot be represented as a continuous line. The rigorous definition of lack
of gaps needs the concept of cut of a set.

Definition 3.15: Partition of a set

Let 𝑆 be a non-empty set. The pair (𝐴, 𝐵) is a partition of 𝑆 if

𝐴, 𝐵 ⊆ 𝑆 , 𝐴 ≠ ∅, 𝐵 ≠ ∅,

and
𝑆 = 𝐴∪𝐵, 𝐴 ∩ 𝐵 = ∅.

Figure 3.1: Schematic picture of a partition (𝐴, 𝐵) of the set 𝐾 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 55

Definition 3.16: Cut of a set

Let 𝑆 be a non-empty set with a total order relation ≤. The pair (𝐴, 𝐵) is a cut of 𝑆 if

1. (𝐴, 𝐵) is a partition of 𝑆,
2. We have
𝑎 ≤ 𝑏, ∀𝑎 ∈ 𝐴, ∀𝑏 ∈ 𝐵.

The cut of a set is often called Dedekind cut, named after Richard Dedekind, who used cuts to give an
explicit construction of the real numbers ℝ, see Wikipedia page.

Definition 3.17: Cut property

Let 𝑆 be a non-empty set with a total order relation ≤. We say that 𝑆 has the cut property if for every
cut (𝐴, 𝐵) of 𝑆 there exists some 𝑠 ∈ 𝑆 such that

𝑎 ≤ 𝑠 ≤ 𝑏, ∀𝑎 ∈ 𝐴, ∀𝑏 ∈ 𝐵.

We call 𝑠 the separator of the cut (𝐴, 𝐵).

Example 3.18

Let 𝑆 = ℚ and consider the sets

𝐴 = (−∞, 𝑠] ∩ ℚ , 𝐵 = (𝑠, ∞) ∩ ℚ .

for some 𝑠 ∈ ℚ. Then the pair (𝐴, 𝐵) is a cut of ℚ, and 𝑠 is the separator.

Figure 3.2: (𝐴, 𝐵) is a cut of ℚ with separator 𝑠.

Question 3.19

Do all ordered fields have the Cut Property? Does ℚ have the Cut Property?

The answer to the above question is NO. For example the pair
𝐴 = (−∞, √2) ∩ ℚ , 𝐵 = (√2, ∞) ∩ ℚ . (3.1)
is a cut of ℚ, since √2 ∉ ℚ. However what is the separator? It should be 𝑠 = √2, given that clearly
𝑎 ≤ √2 ≤ 𝑏 , ∀𝑎 ∈ 𝐴, ∀𝑏 ∈ 𝐵.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 56

However √2 ∉ ℚ, so we are NOT ALLOWED to take it as separator. Indeed, we can show that (𝐴, 𝐵) defined
as in (3.1) has no separator.

Figure 3.3: (𝐴, 𝐵) is a cut of ℚ which has no separator

Theorem 3.20: ℚ does not have the cut property.

ℚ does not have the cut property. More explicitly, there exist a cut (𝐴, 𝐵) of ℚ which has no separator.

Remark 3.21: Ideas for the proof of Theorem 0.72

Before proceeding with the proof, let us summarize the ideas behind it:
We will consider the cut (𝐴, 𝐵) in (3.1). We then assume by contradiction that (𝐴, 𝐵) admits a separator
𝐿 ∈ ℚ, so that
𝑎 ≤ 𝐿 ≤ 𝑏, ∀𝑎 ∈ 𝐴, ∀𝑏 ∈ 𝐵. (3.2)
Since (𝐴, 𝐵) is a partition of ℚ, then either 𝐿 ∈ 𝐴 or 𝐿 ∈ 𝐵. These will both lead to a contradiction:
• If 𝐿 ∈ 𝐴, by definition of 𝐴 we have
𝐿 < √2 .
We want to contradict the fact that 𝐿 is a separator for the cut (𝐴, 𝐵). The idea is that √2 ∉ ℚ, and
therefore it is possible to find a rational number 𝐿̃ such that
𝐿 < 𝐿̃ < √2 .
How do we find such 𝐿̃ in practice? We look for a number 𝐿̃ 𝑛 of the form
1
𝐿̃ 𝑛 = 1 +
𝑛
for some 𝑛 ∈ ℕ to be suitably chosen later. Clearly 𝐿̃ 𝑛 ∈ ℚ and
𝐿 < 𝐿̃ 𝑛
for all 𝑛 ∈ ℕ. We need to prove that we can find 𝑛0 ∈ ℕ such that
𝐿 < 𝐿̃ 𝑛0 < √2 . (3.3)
This is indeed possible: There exists 𝑛0 ∈ ℕ such that (3.3) holds. From (3.3) we see that 𝐿̃ 𝑛0 ∈ 𝐴.
Since 𝐿 is a separator, from (3.2) we obtain
𝐿̃ 𝑛0 ≤ 𝐿 ,
which contradicts (3.3).

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 57

• If 𝐿 ∈ 𝐵, by definition of 𝐵 we have
√2 < 𝐿 .
The idea is the same as above: Since √2 ∉ ℚ, we can find 𝐿̃ ∈ ℚ such that

√2 < 𝐿̃ < 𝐿 .

Since we want 𝐿̃ to be a rational number smaller than 𝐿, we look for 𝐿̃ of the form
1
𝐿̃ 𝑛 ∶= 𝐿 − ,
𝑛
for a suitable 𝑛 ∈ ℕ. This satisfies 𝐿̃ 𝑛 ∈ ℚ and

𝐿̃ 𝑛 < 𝐿 ,

for all 𝑛 ∈ ℕ. We then find 𝑛0 ∈ ℕ such that

√2 < 𝐿̃ 𝑛0 < 𝐿 . (3.4)

The above shows that 𝐿̃ 𝑛0 ∈ 𝐵. As 𝐿 is a separator, we find that

𝐿 ≤ 𝐿̃ 𝑛0 ,

which contradicts (3.4).

Therefore, both cases 𝐿 ∈ 𝐴 or 𝐿 ∈ 𝐵 lead to a contradiction. Since these are all the possibilities, we
conclude that the cut (𝐴, 𝐵) has no separator in ℚ.

Time to make the ideas in the above remark rigorous.

Proof: Proof of Theorem 0.72

Let 𝐴 and 𝐵 be the sets defined in (3.1). It is useful to rewrite 𝐴 and 𝐵 in the form

𝐴 = 𝐴 1 ∪ 𝐴2 ,

where
𝐴1 = {𝑞 ∈ ℚ ∶ 𝑞 < 0} ,
𝐴2 = {𝑞 ∈ ℚ ∶ 𝑞 ≥ 0 , 𝑞 2 < 2} ,
and
𝐵 = {𝑞 ∈ ℚ ∶ 𝑞 > 0, 𝑞 2 > 2} .
Step 1. (𝐴, 𝐵) is a cut of ℚ:
We need to prove the following:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 58

1. (𝐴, 𝐵) is a partition of ℚ. This is because 𝐴, 𝐵 ⊆ ℚ with 𝐴 ≠ ∅ and 𝐵 ≠ ∅. Moreover 𝐴 ∩ 𝐵 = ∅ and

𝐴 ∪ 𝐵 = ℚ,

given that √2 ∉ ℚ, and so there is no element 𝑞 ∈ ℚ such that 𝑞 2 = 2.

2. It holds
𝑎 ≤ 𝑏 , ∀𝑎 ∈ 𝐴 , ∀ 𝑏 ∈ 𝐵 .
Indeed, suppose that 𝑎 ∈ 𝐴 and 𝑏 ∈ 𝐵. We have two cases:
• 𝑎 ∈ 𝐴1 : Therefore 𝑎 < 0. In particular

𝑎 < 0 < 𝑏,

given that 𝑏 > 0 for all 𝑏 ∈ 𝐵. Thus 𝑎 < 𝑏.

• 𝑎 ∈ 𝐴2 : Therefore 𝑎 ≥ 0 and 𝑎2 < 2. In particular

𝑎2 < 2 < 𝑏 2 ,

since 𝑏 2 > 2 for all 𝑏 ∈ 𝐵. In particular

𝑎2 < 𝑏 2 .

Since 𝑏 > 0 for all 𝑏 ∈ 𝐵, from the above inequality we infer 𝑎 < 𝑏, concluding.

Step 2. (𝐴, 𝐵) has no separator:

Suppose by contradiction that (𝐴, 𝐵) admits a separator

𝐿 ∈ ℚ.

By definition this means

𝑎 ≤ 𝐿 ≤ 𝑏, ∀𝑎 ∈ 𝐴 , ∀ 𝑏 ∈ 𝐵 . (3.5)
Since
𝐿 ∈ ℚ, ℚ = 𝐴∪𝐵, 𝐴 ∩ 𝐵 = ∅,
then either 𝐿 ∈ 𝐴 or 𝐿 ∈ 𝐵. We will see that both these possibilities lead to a contradiction:
Case 1: 𝐿 ∈ 𝐴.
By (3.5) we know that
𝑎 ≤ 𝐿, ∀𝑎 ∈ 𝐴. (3.6)
In particular the above implies
𝐿≥0 (3.7)
since 0 ∈ 𝐴. Therefore we must have 𝐿 ∈ 𝐴2 , that is,

𝐿 ≥ 0 and 𝐿2 < 2 . (3.8)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 59

Set
1
𝐿̃ ∶= 𝐿 +
𝑛
for 𝑛 ∈ ℕ, 𝑛 ≠ 0 to be chosen later. Clearly we have

𝐿̃ ∈ ℚ and 𝐿 < 𝐿̃ . (3.9)

From (3.7) and (3.9) we have also

𝐿̃ > 0 . (3.10)
We now want to show that there is a choice of 𝑛 such that 𝐿̃ 2 < 2, which will lead to a contradiction.
Indeed, we can estimate

1 2
𝐿̃ 2 = (𝐿 + )
𝑛
1 𝐿
= 𝐿2 + 2 + 2
𝑛 𝑛
1 𝐿 1 1
< 𝐿2 + + 2 (using < 2)
𝑛 𝑛 𝑛 𝑛
2𝐿 + 1
= 𝐿2 + .
𝑛
If we now impose that
2𝐿 + 1
𝐿2 + < 2,
𝑛
we can rearrange the above and obtain

𝑛(2 − 𝐿2 ) > 2𝐿 + 1 .

Now note that 𝐿2 < 2 by assumption (3.8). Thus we can divived by (2 − 𝐿2 ) and obtain
2𝐿 + 1
𝑛> .
2 − 𝐿2
Therefore we have just shown that
2𝐿 + 1
𝑛> ⟹ 𝐿̃ 2 < 2 .
2 − 𝐿2
Together with (3.10) this implies 𝐿̃ ∈ 𝐴. Therefore we have

𝐿̃ ≤ 𝐿

by (3.6). On the other hand it also holds

𝐿̃ > 𝐿
by (3.9), and therefore we have a contradiction. Thus 𝐿 ∉ 𝐴.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 60

Case 2: 𝐿 ∈ 𝐵.
As 𝐿 ∈ 𝐵, we have by definition
𝐿 > 0, 𝐿2 > 2 . (3.11)
Moreover since 𝐿 is a separator, see (3.5), in particular
𝐿 ≤ 𝑏, ∀𝑏 ∈ 𝐵. (3.12)
Define now
1
𝐿̃ ∶= 𝐿 −
𝑛
with 𝑛 ∈ ℕ, 𝑛 ≠ 0 to be chosen later. Clearly we have
𝐿̃ ∈ ℚ , 𝐿̃ < 𝐿 . (3.13)
We now show that 𝑛 can be chosen so that 𝐿̃ ∈ 𝐵. Indeed
1 2
𝐿̃ 2 = (𝐿 − )
𝑛
1 𝐿
= 𝐿2 + 2 − 2
𝑛 𝑛
1 𝐿 1 1
> 𝐿2 − 2 − 2 (using 2
> − 2)
𝑛 𝑛 𝑛 𝑛
1 𝐿 1 1
> 𝐿2 − − 2 (using − 2 > − )
𝑛 𝑛 𝑛 𝑛
1 + 2𝐿
= 𝐿2 − .
𝑛
Now we impose
1 + 2𝐿
𝐿2 − >2
𝑛
which is equivalent to
𝑛(𝐿2 − 2) > 1 + 2𝐿 .
Since we are assuming 𝐿 ∈ 𝐵, then 𝐿2 > 2, see (3.11). Therefore we can divide by (𝐿2 − 2) and get
1 + 2𝐿
𝑛> .
𝐿2 − 2
In total, we have just shown that
1 + 2𝐿
𝑛> ⟹ 𝐿̃ 2 > 2 ,
𝐿2 − 2
proving that 𝐿̃ ∈ 𝐵. Therefore by (3.12) we get
𝐿 ≤ 𝐿̃ .
This contradicts (3.13).
Conclusion:
We have seen that assuming that (𝐴, 𝐵) has a separator 𝐿 ∈ ℚ leads to a contradiction. Thus the cut
(𝐴, 𝐵) has no separator.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 61

Remark 3.22

The above proof can be summarized by saying that the set

𝐴 = (−∞, √2) ∩ ℚ

does not admit a largest element in ℚ, and that the set

𝐵 = (√2, ∞) ∩ ℚ

does not admit a lowest element in ℚ. We will clarify this remark in the next section.

3.4 Supremum and infimum

A crucial definition in Analysis is the one of supremum or infimum of a set. This is also another way of
studying the gaps of ℚ.

Example 3.23: Intuition about supremum and infimum

Consider the set

𝐴 = [0, 1) ∩ ℚ .
Intuitively, we understand that 𝐴 is bounded, i.e. not infinite. We also see that

• 0 is the lowest element of 𝐴

• 1 is the highest element of 𝐴

However we see that 0 ∈ 𝐴 while 1 ∉ 𝐴. We will see that

• 0 can be defined as the infimum and minimum of 𝐴.

• 1 can be defined as the supremum, but not maximum, of 𝐴.

3.4.1 Upper bound, supremum, maximum

We start by defining the supremum. First we need the notion of upper bound of a set.

Definition 3.24: Upper bound and bounded above

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 :

1. We say that 𝑏 ∈ 𝐾 is an upper bound for 𝐴 if
𝑎 ≤ 𝑏, ∀𝑎 ∈ 𝐴.
2. We say that 𝐴 is bounded above if there exists and upper bound 𝑏 ∈ 𝐾 for 𝐴.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 62

Definition 3.25: Supremum

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 . A number 𝑠 ∈ 𝐾 is called least upper bound or supremum
of 𝐴 if:

i. 𝑠 is an upper bound for 𝐴,

ii. 𝑠 is the smallest upper bound of 𝐴, that is,

If 𝑏 ∈ 𝐾 is upper bound for 𝐴 then 𝑠 ≤ 𝑏 .

Notation 3.26

We will almost always prefer the name supremum to least upper bound. For 𝐴 ⊆ 𝐾 the supremum is
denoted by
𝑠 ∶= sup 𝐴 .

Remark 3.27

Note that if a set 𝐴 ⊆ 𝐾 in NOT bounded above, then the supremum does not exist, as there are no upper
bounds of 𝐴.

Proposition 3.28

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 . If

sup 𝐴

exists, then it is unique.

Proof
Suppose there exist 𝑠1 , 𝑠2 ∈ 𝐾 such that

𝑠1 = sup 𝐴, 𝑠2 = sup 𝐴 .

Then:

• Since 𝑠2 = sup 𝐴, in particular 𝑠2 is an upper bound for 𝐴. Since 𝑠1 = sup 𝐴 then 𝑠1 is the lowest
upper bound. Thus we get
𝑠1 ≤ 𝑠 2 .
• Exchanging the roles 𝑠1 and 𝑠2 in the above reasoning we also get

𝑠2 ≤ 𝑠 1 .

This shows 𝑠1 = 𝑠2 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 63

Warning

In general:

• A set can have infinite upper bounds,

• The supremum does not belong to the set.

For example
𝐴 = [0, 1) ∩ ℚ
has for upper bounds all the numbers 𝑏 ∈ ℚ with 𝑏 > 1. Moreover one can show that

sup 𝐴 = 1 ,

and so
sup 𝐴 ∉ 𝐴 .

Warning

The supremum does not exist in general. For example let

𝐴 = [0, √2) ∩ ℚ .

We will show that sup 𝐴 does not exist in ℚ. Indeed we will have that

sup 𝐴 = √2 ∈ ℝ .

Figure 3.4: Supremum and upper bounds of a set 𝐴 in the field 𝐾

Definition 3.29: Maximum

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 . A number 𝑀 ∈ 𝐾 is called the maximum of 𝐴 if:

𝑀 ∈ 𝐴 and 𝑎 ≤ 𝑀 , ∀𝑎 ∈ 𝐴 .

We denote the maximum by

𝑀 = max 𝐴 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 64

Proposition 3.30

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 . If the maximum of 𝐴 exists, then also the supremum exists,
and
sup 𝐴 = max 𝐴 .

Proof
Let
𝑀 = max 𝐴 .
Then:

• By definition we have 𝑀 ∈ 𝐴 and

𝑎≤𝑀, ∀𝑎 ∈ 𝐴.
In particular the above tells us that 𝑀 is an upper bound of 𝐴.
• We claim that 𝑀 is the least upper bound. Indeed, suppose 𝑏 is an upper bound of 𝐴, that is,

𝑎 ≤ 𝑏, ∀𝑎 ∈ 𝐴.
In particular, since 𝑀 ∈ 𝐴, by the above condition we have

𝑀 ≤ 𝑏.

Therefore 𝑀 is the least upper bound of 𝐴, meaning that 𝑀 = sup 𝐴.

Warning

The converse of the above statement is not true: In general the sup might exist while the max does not.
For example
𝐴 = [0, 1) ∩ ℚ
is such that
sup 𝐴 = 1
but max 𝐴 does not exist. Instead for the set

𝐵 = [0, 1] ∩ ℚ

we have that
max 𝐴 = sup 𝐴 = 1 .

3.4.2 Lower bound, infimum, minimum

We now introduce the definitions of lower bound, infimum, minimum. These are the counterpart of upper
bound, supremum and maximum, respectively.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 65

Definition 3.31: Upper bound, bounded below, infimum, minimum

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 :

1. We say that 𝑙 ∈ 𝐾 is a lower bound for 𝐴 if

𝑙 ≤ 𝑎, ∀𝑎 ∈ 𝐴.

2. We say that 𝐴 is bounded below if there exists a lower bound 𝑙 ∈ 𝐾 for 𝐴.

3. We say that 𝑖 ∈ 𝐾 is the greatest lower bound or infimum of 𝐴 if:

• 𝑖 is a lower bound for 𝐴,
• 𝑖 is the largest lower bound of 𝐴, that is,

If 𝑙 ∈ 𝐾 is a lower bound for 𝐴 then 𝑙 ≤ 𝑖 .

If it exists, the infimum is denoted by

𝑖 = inf 𝐴 .

4. We say that 𝑚 ∈ 𝐾 is the minimum of 𝐴 if:

𝑚 ∈ 𝐴 and 𝑚 ≤ 𝑎 , ∀𝑎 ∈ 𝐴 .

If it exists, we denote the minimum by

𝑚 = min 𝐴 .

Figure 3.5: Infimum and lower bounds of a set 𝐴 in the field 𝐾

Proposition 3.32

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 :

1. If inf 𝐴 exists, then it is unique.

2. If the minimum of 𝐴 exists, then also the infimum exists, and

inf 𝐴 = min 𝐴 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 66

The proof uses similar arguments to the one employed in the previous section, and is left to the reader as an
exercise.

Warning

We have

• A set can have infinite lower bounds,

• The infimum does not belong to the set.

For example
𝐴 = (0, 1) ∩ ℚ
has for lower bounds all the numbers 𝑏 ∈ ℚ with 𝑏 < 1. Moreover we will show that

inf 𝐴 = 0 ,

and so
inf 𝐴 ∉ 𝐴 .

Warning

The infimum does not exist in general. For example let

𝐴 = (√2, 5] ∩ ℚ .

We will show that inf 𝐴 does not exist in ℚ. Indeed we will have that

inf 𝐴 = √2 ∈ ℝ .

Warning

In general the inf might exist while the min does not. For example

𝐴 = (0, 1) ∩ ℚ

is such that
inf 𝐴 = 0
but min 𝐴 does not exist. Instead for the set

𝐵 = [0, 1] ∩ ℚ

we have that
inf 𝐴 = min 𝐴 = 0 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 67

Proposition 3.33

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 . If inf 𝐴 and sup 𝐴 exist, then

inf 𝐴 ≤ 𝑎 ≤ sup 𝐴 , ∀𝑎 ∈ 𝐴 .

The proof is simple, and is left as an exercise. We now have a complete picture about supremum and infimum,
see figure below.

Figure 3.6: Supremum, upper bounds, infimum and lower bounds of a set 𝐴 in 𝐾

We conclude with another simple proposition. The proof is again left to the reader.

Proposition 3.34: Relationship between sup and inf

Let (𝐾 , +, ⋅, ≤) be an ordered field and 𝐴 ⊆ 𝐾 . Define

−𝐴 ∶= {−𝑎 ∶ 𝑎 ∈ 𝐴} .

It holds:

• If sup 𝐴 exists, then inf 𝐴 exists and

inf(−𝐴) = − sup 𝐴 .

• If inf 𝐴 exists, then sup 𝐴 exists and

sup(−𝐴) = − inf 𝐴 .

3.5 Completeness
We have introduced the concepts of supremum and infimum on an ordered field 𝐾 .

Question 3.35

Suppose (𝐾 , +, ⋅, ≤) is an ordered field, and that 𝐴 ⊆ 𝐾 is non-empty and bounded above. Does
sup 𝐴

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 68

always exist?

The answer to the above question is NO. Like we did with the Cut Property, the counterexample can be
found in the set of rational numbers ℚ. A set bounded above for which the supremum does nor exist is, for
example,
𝐴 = [0, √2) ∩ ℚ . (3.14)

Theorem 3.36

There exists a set 𝐴 ⊆ ℚ such that

• 𝐴 is non-empty,
• 𝐴 is bounded above,
• sup 𝐴 does not exist in ℚ.

The proof uses the same ideas we used for showing that ℚ does not have the Cut Property.

Proof
Define the set 𝐴 as in (3.14). Equivalently, this can be written as

𝐴 = {𝑞 ∈ ℚ ∶ 𝑞 ≥ 0 , 𝑞 2 < 2} .

Step 1. 𝐴 is bounded above.

Take 𝑏 ∶= 9. Then 𝑏 is an upper bound for 𝐴. Indeed by definition

𝑞 2 < 2 , 𝑞 ≥ 0 , ∀𝑞 ∈ 𝐴 .

Therefore
𝑞2 < 2 < 9 ⟹ 𝑞2 < 9 ⟹ 𝑞 < 3 = 𝑏 .
Step 2. sup 𝐴 does not exist.
Assume by contradiction that
𝑠 = sup 𝐴 ∈ ℚ
exists. By definition it holds
𝑠 ≥ 𝑞, ∀𝑞 ∈ 𝐴 (3.15)
𝑏 ≥ 𝑞, ∀𝑞 ∈ 𝐴 ⟹ 𝑠 ≤ 𝑏 (3.16)
There are two possibilities: 𝑠 ∈ 𝐴 or 𝑠 ∉ 𝐴:

• Case 1. 𝑠 ∈ 𝐴.
If 𝑠 ∈ 𝐴 by definition
𝑠 ≥ 0, 𝑠2 < 2 . (3.17)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 69

Define
1
𝑠 ̃ ∶= 𝑠 +
𝑛
with 𝑛 ∈ ℕ, 𝑛 ≠ 0 to be chosen later. Then

1 2
𝑠 ̃2 = (𝑠 + )
𝑛
1 𝑠
= 𝑠2 + 2 + 2
𝑛 𝑛
1 𝑠 1 1
< 𝑠2 + + 2 (using < 2)
𝑛 𝑛 𝑛 𝑛
2𝑠 + 1
= 𝑠2 + .
𝑛
If we now impose that
2𝑠 + 1
𝑠2 + < 2,
𝑛
we can rearrange the above and obtain

𝑛(2 − 𝑠 2 ) > 2𝑠 + 1 .

Now note that 𝑠 2 < 2 by assumption (3.17). Thus we can divide by (2 − 𝑠 2 ) and obtain
2𝑠 + 1
𝑛> .
2 − 𝑠2
To summarize, we have just shown that
2𝑠 + 1
𝑛> 2
⟹ 𝑠 ̃2 < 2 .
2−𝑠
Moreover 𝑠 ̃ ∶= (𝑠 + 1/𝑛) ∈ ℚ. Therefore
𝑠̃ ∈ 𝐴 .
Since 𝑠 = sup 𝐴, we then have
𝑠̃ ≤ 𝑠 .
However
1
𝑠 ̃ ∶= 𝑠 + > 𝑠,
𝑛
yielding a contradiction. Thus 𝑠 ∈ 𝐴 is not possible.

• Case 2. 𝑠 ∉ 𝐴.
If 𝑠 ∉ 𝐴, by the fact that 𝑠 = sup 𝐴 and by definition of 𝐴 we get

𝑠 > 0, 𝑠2 > 2 . (3.18)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 70

Define
1
𝑠 ̃ ∶= 𝑠 − .
𝑛
We have
1 2
𝑠 ̃2 = (𝑠 − )
𝑛
2 1 𝑠
=𝑠 + 2 −2
𝑛 𝑛
1 𝑠 1 1
> 𝑠2 − 2 − 2 (using 2
> − 2)
𝑛 𝑛 𝑛 𝑛
1 𝑠 1 1
> 𝑠2 − − 2 (using − 2 > − )
𝑛 𝑛 𝑛 𝑛
1 + 2𝑠
= 𝑠2 − .
𝑛
Now we impose
1 + 2𝑠
𝑠2 − >2
𝑛
which is equivalent to
𝑛(𝑠 2 − 2) > 1 + 2𝑠 .
By (3.18) we have 𝑠 2 > 2. Therefore we can divide by (𝑠 2 − 2) and get
1 + 2𝑠
𝑛> .
𝑠2 − 2
In total, we have just shown that
1 + 2𝑠
𝑛> ⟹ 𝑠 ̃2 > 2 .
𝑠2 − 2
Therefore 𝑠 ̃ ∉ 𝐴, and by definition of 𝐴 we have

𝑠̃ ≥ 𝑞 , ∀𝑞 ∈ 𝐴 .

Moreover 𝑠 ̃ ∶= (𝑠 − 1/𝑛) ∈ ℚ. Therefore 𝑠 ̃ is an upper bound of 𝐴 in ℚ. Since 𝑠 = sup 𝐴 is the

smallest upper bound, see (3.16), it follows

𝑠 ≤ 𝑠̃ .

However
1
𝑠 ̃ ∶= 𝑠 − < 𝑠,
𝑛
obtaining a contradiction. Then 𝑠 ∉ 𝐴.

Conclusion.
We have assumed by contradiction that 𝑠 = sup 𝐴 exists in ℚ. In this case either 𝑠 ∈ 𝐴 or 𝑠 ∉ 𝐴. In both
cases we found a contradiction. Therefore sup 𝐴 does not exist.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 71

The above theorem shows that the supremum does not necessarily exist. What about the infimum?

Question 3.37

Suppose (𝐾 , +, ⋅, ≤) is an ordered field, and that 𝐴 ⊆ 𝐾 is non-empty and bounded below. Does

inf 𝐴

always exist?

The answer to the above question is again NO. A set bounded below for which the infimum does nor exist is,
for example,
𝐴 = (√2, 10] ∩ ℚ .
The proof of this fact is, of course, very similar to the one of Theorem 3.36, and is therefore omitted.
Thus infimum and supremum do not exist in general. The fields for which all the bounded sets admit supre-
mum or infimum are called complete.

Definition 3.38: Completeness

Let (𝐾 , +, ⋅, ≤) be an ordered field. We say that 𝐾 is complete if it holds the property:

• (AC) For every 𝐴 ⊆ 𝐾 non-empty and bounded above

sup 𝐴 ∈ 𝐾 .

Notation 3.39

We have that:

• Property (AC) is called Axiom of Completeness

• If 𝐾 is an ordered field in which (AC) holds, then 𝐾 is called a complete ordered field

Notice that if the Axiom of Completeness holds, then also the infimum exists. This is shown in the follow-
ing proposition.

Proposition 3.40

Let (𝐾 , +, ⋅, ≤) be a complete ordered field. Suppose that 𝐴 ⊆ 𝐾 is non-empty and bounded below. Then

inf 𝐴 ∈ 𝐾 .

Proof
Suppose that 𝐴 ⊆ 𝐾 is non-empty and bounded below. Then

−𝐴 ∶= {−𝑎 ∶ 𝑎 ∈ 𝐴}

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 72

is non-empty and bounded above. By completeness we have that sup(−𝐴) exists in 𝐾 . But then Proposi-
tion 0.86 implies that inf 𝐴 exists in 𝐾 , with

inf 𝐴 = − sup(−𝐴) .

3.6 Equivalence of Completeness and Cut Property

We can show that Completeness is equivalent to the Cut Property. Such result is not essential, but its proof
is very instructive.

Theorem 3.41: Equivalence of Cut Property and Completeness

Let (𝐾 , +, ⋅, ≤) be an ordered field. Then they are equivalent:

1. 𝐾 has the Cut Property

2. 𝐾 is Complete

Remark 3.42: Ideas for proving Theorem 0.93

The proof of Theorem 0.93 is rather long, but the ideas are simple:
Step 1. Cut Property ⟹ Completeness. Suppose 𝐾 has the Cut Property. To prove that 𝐾 is Complete,
we need to:

• Consider an arbitrary set 𝐴 ⊆ 𝐾 such that 𝐴 ≠ ∅ and 𝐴 is bounded above.

• Show that 𝐴 has a supremum.

To achieve this, consider the set

𝐵 ∶= {𝑏 ∈ 𝐾 ∶ 𝑏 ≥ 𝑎 , ∀𝑎 ∈ 𝐴} ,

which is the set of Upper Bounds of 𝐴. We can show that the pair

(𝐵𝑐 , 𝐵)

is a Cut of 𝐾 . As 𝐾 has the Cut Property, then there exists 𝑠 ∈ 𝐾 separator of (𝐵𝑐 , 𝐵). We will show that
the separator 𝑠 is the supremum of 𝐴
𝑠 = sup 𝐴 .
Thus 𝐾 is complete. See Figure 3.7 for a schematic picture of the above construction.
Step 2. Completeness ⟹ Cut Property. Conversely, suppose that 𝐾 is Complete. To prove that 𝐾 has
the Cut Property, we need to:

• Consider a cut (𝐴, 𝐵) of 𝐾 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 73

• Show that (𝐴, 𝐵) has a separator 𝑠 ∈ 𝐾 .

This implication is easier. Indeed, since 𝐴 is non-empty and bounded above, by Completeness there exists

sup 𝐴 ∈ 𝐾 .

We will show that

𝑠 ∶= sup 𝐴
is a separator for the cut (𝐴, 𝐵). See Figure 3.8 for a schematic picture of the above construction.

Figure 3.7: Let 𝑠 be the separator of the cut (𝐵𝑐 , 𝐵), with 𝐵 the set of upper bounds of 𝐴. Then 𝑠 = sup 𝐴.

Figure 3.8: Let (𝐴, 𝐵) be a cut of 𝐾 and let 𝑠 = sup 𝐴. Then 𝑠 is the separator of the cut (𝐴, 𝐵).

Keeping the above ideas in mind, let us proceed with the proof.

Proof: Proof of Theorem 0.93

Step 1. Cut Property ⟹ Completeness.

We need to prove that 𝐾 is complete. To this end, consider 𝐴 ⊆ 𝐾 non-empty and bounded above. Define
the set of upper bounds of 𝐴:
𝐵 ∶= {𝑏 ∈ 𝐾 ∶ 𝑏 ≥ 𝑎 , ∀𝑎 ∈ 𝐴} .
Claim. The pair (𝐵𝑐 , 𝐵) is a cut of 𝐾 .
Proof of Claim. We have to prove two points:

• (𝐵𝑐 , 𝐵) forms a partition of 𝐾 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 74

Indeed, we have 𝐵 ≠ ∅, since 𝐴 is bounded above. Further, we have 𝐵𝑐 ≠ ∅, since 𝐴 is

non-empty. Thus
𝐾 = 𝐵𝑐 ∪ 𝐵 , 𝐵𝑐 ∩ 𝐵 = ∅ .
Then (𝐵𝑐 , 𝐵) is a partition of 𝐾 .

• We have
𝑥 ≤𝑦, ∀ 𝑥 ∈ 𝐵𝑐 , ∀ 𝑦 ∈ 𝐵 . (3.19)

To show the above, let 𝑥 ∈ 𝐵𝑐 and 𝑦 ∈ 𝐵. By definition of 𝐵 we have that elements of 𝐵𝑐 are
not upper bounds of 𝐴. Therefore 𝑥 is not an upper bound. This means there exists 𝑎̃ ∈ 𝐴
which is larger than 𝑥, that is,
𝑥 ≤ 𝑎̃ .
Since 𝑦 ∈ 𝐵, then 𝑦 is an upper bound for 𝐴, so that

𝑎 ≤ 𝑦 , ∀𝑎 ∈ 𝐴 .

Therefore
𝑥 ≤ 𝑎̃ ≤ 𝑦 ,
concluding (3.19).

Thus (𝐵𝑐 , 𝐵) is a cut of 𝐾 and the claim is proven.

Since (𝐵𝑐 , 𝐵) is a cut of 𝐾 , by the Cut Property there exists a separator 𝑠 ∈ 𝐾 such that

𝑥 ≤𝑠≤𝑦, ∀ 𝑥 ∈ 𝐵𝑐 , ∀ 𝑦 ∈ 𝐵 . (3.20)

Claim. 𝑠 is an upper bound for 𝐴.

Proof of Claim.
Suppose by contradiction that 𝑠 is not an upper bound for 𝐴. Therefore by definition of upper bound,
there exists 𝑎̃ ∈ 𝐴 such that
𝑠 < 𝑎̃ .
Consider the mid-point between 𝑠 and 𝑎,̃ that is,
𝑠 + 𝑎̃
𝑚 ∶= ∈𝐾.
2
Since 𝑚 is the mid-point between 𝑠 and 𝑎,̃ and 𝑠 < 𝑎,̃ it holds

𝑠 < 𝑚 < 𝑎̃ .

Indeed, since 𝑠 < 𝑎̃ then

2𝑠 𝑠 + 𝑎̃ 2𝑎̃
𝑠= < < = 𝑎̃ .
2 2 2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 75

In particular the above tells us that 𝑚 is not an upper bound for 𝐴, given that 𝑎̃ ∈ 𝐴 and 𝑚 < 𝑎.̃ Therefore
𝑚 ∈ 𝐵𝑐 , by definition of 𝐵𝑐 . Therefore(3.20) implies

𝑚 ≤ 𝑠,

which contradicts 𝑠 < 𝑚. Hence 𝑠 is an upper bound of 𝐴, concluding the proof of Claim.
Conclusion. We have shown that 𝑠 is an upper bound of 𝐴. Condition
(3.20) tells us that
𝑠 ≤ 𝑦 , ∀𝑦 ∈ 𝐵 .
Recalling that 𝐵 is the set of upper bounds of 𝐴, this means that 𝑠 is the smallest upper bound of 𝐴, that
is,
𝑠 = sup 𝐴 ∈ 𝐾 .
Step 2. Completeness ⟹ Cut Property.
Suppose 𝐾 is complete. We need to show that 𝐾 has the Cut Property. Therefore assume (𝐴, 𝐵) is a cut
of 𝐾 , that is,
𝐴 ≠ ∅, 𝐵 ≠ ∅,
𝐾 = 𝐴∪𝐵, 𝐴 ∩ 𝐵 = ∅,
𝑎 ≤ 𝑏, ∀𝑎 ∈ 𝐴 , ∀ 𝑏 ∈ 𝐵 . (3.21)
Since 𝐵 ≠ ∅, from (3.21) it follows that 𝐴 is bounded above: indeed, every element of 𝐵 is an upper bound
for 𝐴, thanks to (3.21). Since 𝐴 ≠ ∅, by the Axiom of Completeness we have

𝑠 = sup 𝐴 ∈ 𝐾 .

In particular, by definition of supremum, we have

𝑎 ≤ 𝑠, ∀𝑎 ∈ 𝐴.

Let now 𝑏 ∈ 𝐵 be arbitrary. From (3.21) we have that

𝑎 ≤ 𝑏, ∀𝑎 ∈ 𝐴. (3.22)

Therefore 𝑏 is an upper bound of 𝐴. Since 𝑠 = sup 𝐴, we have that 𝑠 is the smallest upper bound, and so

𝑠 ≤ 𝑏.

Given that 𝑏 ∈ 𝐵 was arbitrary, it actually holds

𝑠 ≤ 𝑏, ∀𝑏 ∈ 𝐵. (3.23)

From (3.22) and (3.23) we therefore have

𝑎 ≤ 𝑠 ≤ 𝑏 , ∀𝑎 ∈ 𝐴 , ∀ 𝑏 ∈ 𝐵 ,

showing that 𝑠 is a separator of (𝐴, 𝐵). Thus 𝐾 has the Cut Property.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 76

3.7 Axioms of Real Numbers

We now have all the key elements to introduce the Real Numbers ℝ. These ingredients are:
• Definition of ordered field,
• The Cut Property or Axiom of Completeness.
The definition of ℝ is given in an axiomatic way.

Definition 3.43: System of Real Numbers ℝ

A system of Real Numbers is a set ℝ satisfying the following properties:

1. There is an operation + of addition on ℝ

+ ∶ ℝ × ℝ → ℝ, (𝑥, 𝑦) ↦ 𝑥 + 𝑦

The addition satisifes: ∀ 𝑥, 𝑦, 𝑧 ∈ ℝ

• (A1) Commutativity and Associativity:

𝑥 +𝑦 =𝑦 +𝑥

(𝑥 + 𝑦) + 𝑧 = 𝑥 + (𝑦 + 𝑧)
• (A2) Additive Identity: ∃ 0 ∈ ℝ s.t.

𝑥 +0=0+𝑥 =𝑥

• (A3) Additive Inverse: ∃ (−𝑥) ∈ ℝ s.t.

𝑥 + (−𝑥) = (−𝑥) + 𝑥 = 0

2. There is an operation ⋅ of multiplication on ℝ

⋅ ∶ ℝ × ℝ → ℝ, (𝑥, 𝑦) ↦ 𝑥 ⋅ 𝑦 = 𝑥𝑦

The multiplication satisifes: ∀ 𝑥, 𝑦, 𝑧 ∈ ℝ

• (M1) Commutativity and Associativity:

𝑥 ⋅𝑦 =𝑦 ⋅𝑥

(𝑥 ⋅ 𝑦) ⋅ 𝑧 = 𝑥 ⋅ (𝑦 ⋅ 𝑧)
• (M2) Multiplicative Identity: ∃ 1 ∈ ℝ s.t.

𝑥 ⋅1=1⋅𝑥 =𝑥

• (M3) Multiplicative Inverse: If 𝑥 ≠ 0 , ∃ 𝑥 −1 ∈ ℝ s.t.

𝑥 ⋅ 𝑥 −1 = 𝑥 −1 ⋅ 𝑥 = 1

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 77

3. There is a relation ≤ of total order on ℝ. The order satisfies: ∀ 𝑥, 𝑦, 𝑧 ∈ ℝ

• (O1) Reflexivity:
𝑥≤𝑥
• (O2) Antisymmetry:
𝑥 ≤ 𝑦 and 𝑦 ≤ 𝑥 ⟹ 𝑥 = 𝑦
• (O3) Transitivity:
𝑥 ≤ 𝑦 and 𝑦 ≤ 𝑧 ⟹ 𝑥 = 𝑧
• (O4) Total order:

𝑥 ≤ 𝑦 or 𝑦 ≤ 𝑥
4. The operations + and ⋅, and the total order ≤, are related by the following properties: ∀𝑥, 𝑦, 𝑧 ∈ ℝ
• (AM) Distributive: Relates addition and multiplication via

𝑥 ⋅ (𝑦 + 𝑧) = 𝑥 ⋅ 𝑦 + 𝑥 ⋅ 𝑧

• (AO) Relates addition and order with the requirement:

𝑥 ≤𝑦 ⟹ 𝑥 +𝑧 ≤𝑦 +𝑧

• (MO) Relates multiplication and order with the requirement:

𝑥 ≥ 0, 𝑦 ≥ 0 ⟹ 𝑥 ⋅ 𝑦 ≥ 0

5. Cut Property holds:

• (CP) Every cut (𝐴, 𝐵) of ℝ admits a separator 𝑠 ∈ ℝ s.t.

𝑎 ≤ 𝑠 ≤ 𝑏, ∀𝑎 ∈ 𝐴, ∀𝑏 ∈ 𝐵

Remark 3.44

Since Cut Property and Axiom of Completeness are equivalent by Theorem 0.93, one can replace the
Cut Property in Definition 0.95 Point 5 with:

5. Axiom of Completeness holds:

• (AC) For every 𝐴 ⊆ ℝ non-empty and bounded above

sup 𝐴 ∈ ℝ

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 78

Notation 3.45

For 𝑥 ∈ ℝ, 𝑥 ≠ 0, the multiplicative inverse is also denoted by

1
𝑥 −1 = .
𝑥

Remark 3.46

Recall that

• (𝐾 , +, ⋅) satisfying
(A1)-(A3), (M1)-(M3), (AM)
is a field
• (𝐾 , +, ⋅, ≥) satisfying

(A1)-(A3), (M1)-(M3), (O1)-(O4) ,(AM), (AO), (MO)

is an ordered field

In particular we have that

(ℝ, +, ⋅, ≤)
is a complete ordered field: that is, an ordered field in which the Cut Property (CP) or Axiom of
Completeness (AC) hold

Important

It can be shown that (ℝ, +, ⋅, ≤) is the only complete ordered field.

The above has to be intended in the following sense: if (𝐾 , +, ⋅, ≥) is another complete ordered
field, then 𝐾 looks like ℝ. Mathematically this means that there exists an invertible map
Ψ ∶ ℝ → 𝐾 , called isomorphism of fields, which preserves the operations +, ⋅ and the order
≤.

Question 3.47

We have only postulated the existence of ℝ. Does such complete ordered field actually exist?

The answer is YES. There are several equivalent models for the system ℝ. If time allows, we will look into
one of these models at the end of the module.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 79

3.8 Special subsets of ℝ

In Definition 0.95 we have introduced ℝ as a complete ordered field. This was done axiomatically and in a non
constructive way. What happens now to the sets ℕ, ℤ, ℚ? Are they well defined? Does it still hold that

ℕ,ℤ,ℚ ⊆ ℝ?

The definitions that we gave in Chapter 1 for ℕ, ℤ and ℚ are not related to the system of real numbers ℝ we
just introduced. To overcome this problem, we will have to define new sets

ℕℝ , ℤℝ , ℚℝ

from scracth, starting from the axioms of ℝ. Note that we are using the subscript ℝ to distinguish these new
sets from the old ones.

3.8.1 Natural numbers

Let us start with the definition of ℕℝ . We would like ℕℝ to be

ℕℝ = {1, 2, 3, …} .

Note that we are denoting the above numbers with bold symbols in order to distinguish them from the ele-
ments of ℝ. The key property that we would like ℕℝ to have is the following:

Every n ∈ ℕℝ has a successor (n + 1) ∈ ℕℝ .

How do we ensure this property? We could start by defining

1 ∶= 1 ,

with 1 the neutral element of the multiplication in ℝ, which exists by the field axiom (M2) in Defintion 0.95.
We could then define 2 by setting
2 ∶= 1 + 1 .
We need a formal definition to capture this idea. This is the concept of inductive set.

Definition 3.48: Inductive set

Let 𝑆 ⊆ ℝ. We say that 𝑆 is an inductive set if they are satisfied:

• 1 ∈ 𝑆,
• If 𝑥 ∈ 𝑆, then (𝑥 + 1) ∈ 𝑆.

Note that in the above definition we just used:

• The existence of the neutral element 1, given by axiom (M2).

• The operation of sum in ℝ, which is again given as an axiom.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 80

Example 3.49

We have that

• ℝ is an inductive set.

Indeed we have 1 ∈ ℝ by axiom (M2). Moreover (𝑥 + 1) ∈ ℝ for every 𝑥 ∈ ℝ, by definition of

sum +.

• The set 𝐴 = {0, 1} is not an inductive set.

This is because 1 ∈ 𝐴, but (1 + 1) ∉ 𝐴, since 1 + 1 ≠ 0.

Therefore ℝ is an inductive set, showing that the definition of inductive set is not sufficient to fully describe
our intuitive idea of ℕℝ . The right way to define ℕℝ is as follows:

ℕℝ is the smallest inductive subset of ℝ .

To make the above definition formal we need a few observations.

Proposition 3.50

Let ℳ be a collection of inductive subsets of ℝ. Then

𝑆 ∶= ⋂ 𝑀
𝑀∈ℳ

is an inductive subset of ℝ.

Proof
We have to show that the two properties of inductive sets hold for 𝑆:

• We have 1 ∈ 𝑀 for every 𝑀 ∈ ℳ, since these are inductive sets. Thus

1∈ ⋂ 𝑀 =𝑆.
𝑀∈ℳ

• Suppose that 𝑥 ∈ 𝑆. By definition of 𝑆 this implies that 𝑥 ∈ 𝑀 for all 𝑀 ∈ ℳ. Since 𝑀 is an

inductive set, then (𝑥 + 1) ∈ 𝑀. Therefore (𝑥 + 1) ∈ 𝑀 for all 𝑀 ∈ ℳ, showing that (𝑥 + 1) ∈ 𝑆.

Therefore 𝑆 is an inductive set.

We are now ready to define the natural numbers ℕℝ .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 81

Definition 3.51: Set of Natural Numbers

Let ℳ be the collection of all inductive subsets of ℝ. We define the set of natural numbers in ℝ as

ℕℝ ∶= ⋂ 𝑀 .
𝑀∈ℳ

Therefore ℕℝ is the intersection of all the inductive subsets of ℝ. From this definition it follows that ℕℝ is
the smallest inductive subset of ℝ, as shown in the following proposition.

Proposition 3.52: ℕℝ is the smallest inductive subset of ℝ

Let 𝐶 ⊆ ℝ be an inductive subset. Then

ℕℝ ⊆ 𝐶 .
In other words, ℕℝ is the smallest inductive set in ℝ.

Proof
Let ℳ be the collection of all inductive subsets of ℝ. By definition

ℕℝ = ⋂ 𝑀 .
𝑀∈ℳ

Let 𝑥 ∈ ℕℝ , then 𝑥 ∈ 𝑀 for all 𝑀 ∈ ℳ. Since 𝐶 ∈ ℳ then 𝑥 ∈ 𝐶. This shows ℕℝ ⊆ 𝐶.

The definition of ℕℝ guarantees that all numbers in ℕℝ are larger than 1.

Theorem 3.53

Let 𝑥 ∈ ℕℝ . Then
𝑥 ≥ 1.

Proof
Define the set
𝐶 ∶= {𝑥 ∈ ℝ ∶ 𝑥 ≥ 1} .
We have that 𝐶 is an inductive subset of ℝ.
By definition 1 ∈ 𝐶. Suppose now that 𝑥 ∈ 𝐶, so that 𝑥 ≥ 1. Since 1 ≥ 0 as a consequence of
the field axioms, we deduce that
𝑥 + 1 ≥ 𝑥 + 0 = 𝑥 ≥ 1,
showing that 𝑥 + 1 ≥ 1. Thus (𝑥 + 1) ∈ 𝐶.
By Proposition 0.104 we conclude that
ℕℝ ⊆ 𝐶 ,
showing that 𝑥 ≥ 1 for all 𝑥 ∈ ℕℝ .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 82

Notation 3.54

We have just shown that all the numebers 𝑥 ∈ ℕℝ satisfy

𝑥 ≥ 1.

Moreover by the fact that ℕℝ is an inductive set, we know that

1 + 1 ∈ ℕℝ ,

since 1 ∈ ℕℝ . We denote
2 ∶= 1 + 1 .
Similarly, we will have that
2 + 1 ∈ ℕℝ ,
since 2 ∈ ℕℝ . We denote
3 ∶= 2 + 1 .
In this way we give a name to all the numbers in ℕℝ .

3.8.2 Principle of induction

The Principle of Induction is a consequence of the definition of ℕℝ , see Definition 0.103, and of the field
axioms of ℝ in Definition 0.95.

Theorem 3.55: Principle of Induction

Let 𝛼(𝑛) be a statement depending on 𝑛 ∈ ℕℝ . Assume that

1. 𝛼(1) is true.
2. If 𝛼(𝑛) is true then also 𝛼(𝑛 + 1) is true.

Then 𝛼(𝑛) is true for all 𝑛 ∈ ℕℝ .

Proof
Define the set
𝐶 ∶= {𝑥 ∈ ℕℝ ∶ 𝛼(𝑛) is true} .
We have that 𝐶 is an inductive subset of ℝ.

Indeed:
• 1 ∈ 𝐶 since 𝛼(1) is true by assumption.
• If 𝑛 ∈ 𝐶 then 𝛼(𝑛) is true. By assumption 𝛼(𝑛 + 1) is true. Therefore (𝑛 + 1) ∈ 𝐶.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 83

By Proposition 0.104 we conclude that

ℕℝ ⊆ 𝐶 .
As by definition 𝐶 ⊆ ℕℝ , we have proven that

ℕℝ = 𝐶 ,

showing that 𝛼(𝑛) is true for all 𝑛 ∈ ℕℝ .

As a consequence of the principle of induction, we can prove that ℕℝ is closed under the field operations of
sum and multiplication.

Theorem 3.56

For all 𝑛, 𝑚 ∈ ℕℝ we have:

1. ℕℝ is closed under addition, that is,

𝑚 + 𝑛 ∈ ℕℝ .

2. ℕℝ is closed under multiplication, that is,

𝑚 ⋅ 𝑛 ∈ ℕℝ ,

3. If 𝑚 > 𝑛 there exists 𝑘 ∈ ℕℝ such that

𝑚 =𝑛+𝑘.

Proof
We only prove the first point, the other statements are left as an exercise. Fix 𝑚 ∈ ℕℝ . We prove that

𝑚 + 𝑛 ∈ ℕℝ , ∀ 𝑛 ∈ ℕℝ , (3.24)

by using induction.

• Induction base: We have 𝑚 + 1 ∈ ℕℝ , since 𝑚 ∈ ℕℝ and ℕℝ is an inductive set.

• Inductive hypothesis: Suppose 𝑚 + 𝑛 ∈ ℕℝ . Since ℕℝ is an inductive set, we have (𝑚 + 𝑛) + 1 ∈ ℕℝ .
By associativity of the sum, see axiom (A1), we get

𝑚 + (𝑛 + 1) = (𝑚 + 𝑛) + 1 ∈ ℕℝ ,

which is the desired theis.

By the Induction Principle of Theorem 0.107 we conclude (3.24).

As a consequence of the above theorem, we see that the restriction of the operations of sum and multiplication

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 84

to ℕℝ are still binary operations:

+ ∶ ℕℝ × ℕ ℝ → ℕ ℝ , ⋅ ∶ ℕ ℝ × ℕℝ → ℕ ℝ .

Equipped with the above operations, ℕℝ satisfies the following properties.

Theorem 3.57

(ℕℝ , +, ⋅, ≤) satisfies the following axioms from Definition 0.95:

• (A1).
• (M1), (M2).
• (O1)-(O4).
• (AM), (AO), (MO).

The proof is trivial, as it follows immediately from the inclusion of ℕℝ in ℝ.

3.8.3 Integers

We have seen in Theorem 3.56 that ℕℝ is closed under addition. However ℕℝ is not closed under subtraction.
We therefore define the set of integers ℤℝ in a way that we can perform subtraction of any two natural
numbers.

Definition 3.58: Set of Integers

The set of integers in ℝ is defined by

ℤℝ ∶= {𝑚 − 𝑛 ∶ 𝑛, 𝑚 ∈ ℕℝ } .

In the definition of ℤℝ we denote by −𝑛 the inverse of 𝑛 in ℝ, which exists by the field axiom (A3) in Definition
0.95. The following characterization explains the relationship between ℤℝ and ℕℝ .

Theorem 3.59

It holds
ℤℝ = {−𝑛 ∶ 𝑛 ∈ ℕℝ } ∪ {0} ∪ ℕℝ .

Proof
Define the set
𝑀 ∶= {−𝑛 ∶ 𝑛 ∈ ℕℝ } ∪ {0} ∪ ℕℝ .

• 𝑀 ⊆ ℤℝ : Suppose 𝑚 ∈ 𝑀. We have 3 cases:

– If 𝑚 ∈ {−𝑛 ∶ 𝑛 ∈ ℕℝ } then there exists 𝑛 ∈ ℕℝ such that 𝑚 = −𝑛. Thus

𝑚 = −𝑛 = 1 − (𝑛 + 1) ∈ ℤℝ ,

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 85

since 1 ∈ ℕℝ and 𝑛 + 1 ∈ ℕℝ because 𝑛 ∈ ℕℝ .

– If 𝑚 = 0 then
𝑚 = 0 = 1 − 1 ∈ ℤℝ ,
as 1 ∈ ℕℝ .
– If 𝑚 ∈ ℕℝ then
𝑚 = (𝑚 + 1) − 1 ∈ ℤℝ ,
since 1 ∈ ℕℝ and 𝑚 + 1 ∈ ℕℝ , given that 𝑚 ∈ ℕℝ .
In all 3 cases we have shown that 𝑚 ∈ ℤℝ , proving that 𝑀 ⊆ ℤℝ .

• ℤℝ ⊆ 𝑀: Let 𝑧 ∈ ℤℝ . Then 𝑧 = 𝑚 − 𝑛 for some 𝑛, 𝑚 ∈ ℕℝ . We have 3 cases:

– If 𝑚 = 𝑛 then
(𝐴3)
𝑧 =𝑚−𝑛 =𝑚−𝑚 = 0∈𝑀.
– If 𝑚 > 𝑛, by Theorem 3.56 there exists 𝑘 ∈ ℕℝ such that 𝑚 = 𝑘 + 𝑛. Therefore

𝑧 = 𝑚 − 𝑛 = (𝑘 + 𝑛) − 𝑛
(𝐴1) (𝐴3)
= 𝑘 + (𝑛 − 𝑛) = 𝑘 + 0
(𝐴2)
= 𝑘∈𝑀,

since 𝑘 ∈ ℕℝ .
– If 𝑚 < 𝑛, by Theorem 3.56 there exists 𝑘 ∈ ℕℝ such that 𝑛 = 𝑘 + 𝑚. Therefore

𝑧 = 𝑚 − 𝑛 = −𝑘 ∈ 𝑀 ,

since 𝑘 ∈ ℕℝ , where again we have used (implicitly) the field axioms (A1), (A2) and (A3).

Therefore ℤℝ = 𝑀.

Like we did with ℕℝ , we can also show that ℤℝ is closed under the operations of sum and multiplication.

Theorem 3.60

For all 𝑛, 𝑚 ∈ ℤℝ we have:

1. ℤℝ is closed under addition, that is,

𝑚 + 𝑛 ∈ ℤℝ .

2. ℤℝ is closed under multiplication, that is,

𝑚 ⋅ 𝑛 ∈ ℤℝ ,

The proof is left as an exercise. As a consequence of Theorem 3.60 we have that the restriction of the opera-

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 86

tions of sum and multiplication to ℤℝ are still binary operations:

+ ∶ ℤℝ × ℤ ℝ → ℤ ℝ , ⋅ ∶ ℤ ℝ × ℤℝ → ℤ ℝ .

Equipped with the above operations, ℤℝ satisfies the following properties.

Theorem 3.61

(ℤℝ , +, ⋅, ≤) satisfies the following axioms from Definition 0.95:

• (A1), (A2), (A3).

• (M1), (M2).
• (O1)-(O4).
• (AM), (AO), (MO).

Proof
The fact that
(A1), (A2), (M1), (M2), (O1)-(O4), (AM), (AO), (MO)
are satisfied descends immediately from the inclusion

ℤℝ ⊆ ℝ .

We are left to prove (A3). This is non-trivial because a priori the additive inverse −𝑧 of some 𝑧 ∈ ℤℝ
belongs to ℝ. We need to check that −𝑧 ∈ ℤℝ . Indeed, since 𝑧 ∈ ℤℝ , there exist 𝑛, 𝑚 ∈ ℕℝ such that
𝑧 = 𝑚 − 𝑛. Define 𝑦 ∶= 𝑛 − 𝑚. We have that 𝑦 ∈ ℤℝ and

𝑧 + 𝑦 = (𝑚 − 𝑛) + (𝑛 − 𝑚) = (𝑚 − 𝑚) + (𝑛 − 𝑛) = 0 .

Therefore 𝑦 is the inverse of 𝑧 and 𝑦 ∈ ℤℝ , proving that the sum in ℤℝ satisfies (A3).

Remark 3.62

ℤℝ does not satisfy (M3).

For example, let us show that 2 ∈ ℤℝ has no inverse in ℤℝ . Indeed, let 𝑚 ∈ ℤℝ . By Theorem
3.59 we have 3 cases:
• 𝑚 ∈ ℕℝ : Since 2 > 1 we have
2⋅𝑚 >1⋅𝑚 ≥1
where in the last inequality we used that 𝑚 ≥ 1 for all 𝑚 ∈ ℕℝ , as shown in Theorem
3.53. The above shows that
2 ⋅ 𝑚 > 1,
and therefore 𝑚 cannot be the inverse of 2.
• 𝑚 = 0: Then 2 ⋅ 𝑚 = 0, so that 𝑚 cannot be the inverse of 2.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 87

• 𝑚 = −𝑛 with 𝑛 ∈ ℕℝ . Then
2 ⋅ 𝑚 = 2 ⋅ (−𝑛) < 0 ,
so that 𝑚 cannot be the inverse of 2.
As we have exhausted all the possibilities, we conclude that 2 does not have a multiplicative
inverse in ℕℝ .

3.8.4 Rational numbers

In Theorem 3.61 and 3.62 we have seen that ℤℝ satisfy all the field axiom, except for (M3). We therefore
extend ℤℝ in a way that the extension contains multiplicative inverses. The extension is the set of rational
numbers ℚℝ .

Definition 3.63: Set of Rational Numbers

The set of rational numbers in ℝ is

𝑚
ℚℝ ∶= { ∶ 𝑚 ∈ ℤℝ , 𝑛 ∈ ℕℝ } .
𝑛

Notice that in the above definition we are just using the field axiom (M3), with
𝑚
∶= 𝑚 ⋅ 𝑛−1 .
𝑛
The inverse of 𝑛 exists because we are assuming 𝑛 ∈ ℕℝ , and therefore 𝑛 cannot be 0, as a consequence of
Theorem 3.53.
The set ℚℝ is closed under addition and multiplication (exercise). Therefore they are well defined the opera-
tions:
+ ∶ ℚℝ × ℚ ℝ → ℚ ℝ , ⋅ ∶ ℚ ℝ × ℚℝ → ℚ ℝ .

Theorem 3.64

(ℚℝ , +, ⋅, ≤) is an ordered field.

Proof
All the field properties, except for (M3), follow from the inclusion

ℚℝ ⊆ ℝ

and from the field properties of ℝ. To check (M3), let 𝑞 ∈ ℚℝ with 𝑞 ≠ 0. Therefore 𝑞 = 𝑚/𝑛 for 𝑚 ∈ ℤℝ ,
𝑛 ∈ ℕℝ . As 𝑞 ≠ 0 and 𝑛 ≠ 0, see Theorem 3.53, we deduce that 𝑚 ≠ 0. We have two cases:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 88

• 𝑚 > 0: In this case 𝑚 ∈ ℕℝ by Theorem 3.59. Therefore

𝑛
𝑝= ∈ ℚℝ
𝑚
by definition, since 𝑛, 𝑚 ∈ ℕℝ . By commutativity we have
𝑚 𝑛
𝑞⋅𝑝 = ⋅ = 1.
𝑛 𝑚
• 𝑚 < 0: Then 𝑚 = −𝑥 with 𝑥 ∈ ℕℝ by Theorem 3.59. Therefore
−𝑛
𝑝= ∈ ℚℝ
𝑥
by definition, since −𝑛 ∈ ℤℝ and 𝑥 ∈ ℕℝ . By commutativity we have
𝑚 −𝑛 𝑚 −𝑛
𝑞⋅𝑝 = ⋅ = ⋅ = 1.
𝑛 𝑥 𝑛 −𝑚

Therefore 𝑞 always admits a multiplicative inverse 𝑞 −1 belonging to ℚℝ , proving (M3).

The set ℚℝ does not have the Cut Property or the Axiom of Completeness.

Theorem 3.65

ℚℝ is not complete.

The proof of the above Theorem is a one to one copy of the proof of Theorem 3.36: indeed the proof of
Theorem 3.36 only makes use of field axioms, and thus it applies to ℚℝ .

Notation 3.66

From now on we denote

ℕ ∶= ℕℝ , ℤ ∶= ℤℝ , ℚ ∶= ℚℝ ,
dropping the subscript ℝ.

Dr. Silvio Fanzon [email protected]

4 Properties of ℝ
Now that we established the axiomatic definition of the Real Numbers ℝ as a complete ordered field, let
us investigate some of the properties of ℝ. These will be consequence of the axioms of the real numbers,
particulaly of the Axiom of Completeness.

4.1 Archimedean Property

The Archimedean property is one of the most useful properties of ℝ, and it essentially states that the set of
natural numbers ℕ is not bounded above in ℝ.
More precisley, the Archimedean Property says two things:

1. For any 𝑥 ∈ ℝ we can always find a natural number 𝑛 ∈ ℕ such that

𝑛>𝑥.

2. For any 𝑥 ∈ ℝ with 𝑥 > 0, we can always find a natural number 𝑚 ∈ ℕ such that
1
0< <𝑥.
𝑚

The situation is depicted in Figure 4.2.

Figure 4.1: For any 𝑥 > 0 we can find 𝑛, 𝑚 ∈ ℕ such that 1/𝑚 < 𝑥 < 𝑛.

Remark 4.1

The Archimedean property might sound trivial. However there are examples of ordered fields 𝐾 that
satisfy:
1. ℕ ⊆ 𝐾 .
2. 𝐾 does not have the Archimedean property.
3. In particular, ℕ is bounded above in 𝐾 .
Of course such fields 𝐾 cannot be complete.

89
Numbers, Sequences and Series Page 90

If 𝐾 is complete, then 𝐾 is essentially ℝ, and we are going to prove the Archimedean Property
holds in ℝ.

Let us proceed with the precise statement of the Archimedean property in ℝ.

Theorem 4.2: Archimedean Property

Let 𝑥 ∈ ℝ be given. Then:

1. There exists 𝑛 ∈ ℕ such that

𝑛>𝑥.

2. Suppose in addition that 𝑥 > 0. There exists 𝑛 ∈ ℕ such that

1
<𝑥.
𝑛

Proof
Part 1. Let 𝑥 ∈ ℝ. Suppose by contradiction that there is no 𝑛 ∈ ℕ such that

𝑛>𝑥.

This means that

𝑛≤𝑥 ∀𝑛 ∈ ℕ . (4.1)
The above is saying that the set ℕ is bounded above. Since ℕ is not empty, by the Axiom of Completeness
there exists
𝛼 = sup ℕ .
Claim: (𝛼 − 1) is not an upper bound for ℕ.
Proof of Claim. Indeed, we have
(𝛼 − 1) < 𝛼 . (4.2)
Therefore 𝛼 − 1 cannot be an upper bound for ℕ. Indeed, if by contradiction 𝛼 − 1 was an upper bound
for ℕ, then we would have
𝛼 ≤ (𝛼 − 1) ,
since 𝛼 is the smallest upper bound for ℕ. This contradicts (4.2). Therefore 𝛼 − 1 is not an upper bound
for ℕ.
Conclusion. Since 𝛼 − 1 is not an upper bound for ℕ, there exists 𝑛0 ∈ ℕ such that

𝛼 − 1 < 𝑛0 .

The above implies

𝛼 < 𝑛0 + 1 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 91

Since
(𝑛0 + 1) ∈ ℕ ,
we have obtained a contradiction, given that 𝛼 was the supremum of ℕ. Thus (4.1) is false, meaning that
there exists 𝑛 ∈ ℕ such that
𝑛>𝑥.
Part 2. Suppose 𝑥 ∈ ℝ with 𝑥 > 0. We can define
1
𝑦 ∶= .
𝑥
By Part 1 there exists 𝑛 ∈ ℕ such that
1
𝑛>𝑦 = .
𝑥
Using that 𝑥 > 0, we can rearrange the above inequlaity to obtain
1
<𝑥,
𝑛
which is the desired thesis.

There is another formulation of the Archimedean Property which, depending on the situation, might be more
useful. This formulation says the following: If 𝑥, 𝑦 ∈ ℝ are such that

0<𝑥 <𝑦,

then there exists 𝑛 ∈ ℕ such that

𝑛𝑥 > 𝑦 .
In other words, if one does 𝑛 steps of size 𝑥 in the positive numbers direction, then the resulting number 𝑛𝑥
will be larger than 𝑦. The situation is depicted in Figure 4.2.

Figure 4.2: For 0 < 𝑥 < 𝑦 there exists 𝑛 ∈ ℕ such that that 𝑛𝑥 > 𝑦. In the picture 𝑛 = 3.

Theorem 4.3: Archimedean Property (Alternative formulation)

Let 𝑥, 𝑦 ∈ ℝ, with 0 < 𝑥 < 𝑦. There exists 𝑛 ∈ ℕ such that

𝑛𝑥 > 𝑦 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 92

Proof
Suppose by contradiction that there does not exist some 𝑛 ∈ ℕ such that

𝑛𝑥 > 𝑦 .

This means that

𝑛𝑥 ≤ 𝑦 , ∀𝑛 ∈ ℕ. (4.3)
Define the set
𝐴 ∶= {𝑛𝑥 ∶ 𝑛 ∈ ℕ} .
Condition (4.3) is saying that 𝐴 is bounded above by 𝑦. Morever 𝐴 is trivially non-empty. By the Axiom
of Completeness there exists
𝛼 = sup 𝐴 .
Since 𝛼 is the supremum of 𝐴, by definition of supremum and of the set 𝐴, we have

𝑛𝑥 ≤ 𝛼 , ∀𝑛 ∈ ℕ. (4.4)

As (4.4) holds for every 𝑛 ∈ ℕ, then it also holds for (𝑛 + 1), meaning that

(𝑛 + 1)𝑥 ≤ 𝛼 .

The above implies

𝑛𝑥 ≤ 𝛼 − 𝑥 .
As 𝑛 was arbitrary, we conclude that

𝑛𝑥 ≤ 𝛼 − 𝑥 , ∀𝑛 ∈ ℕ.

The above is saying that (𝛼 − 𝑥) is an upper bound for 𝐴. Since 𝛼 is the supremum of 𝐴, in particular 𝛼
is the smallest upper bound. Thus it must hold

𝛼 ≤𝛼 −𝑥.

The above is equivalent to

𝑥 ≤ 0,
which contradicts our assumption of 𝑥 > 0. Therefore (4.3) is false, and there exists 𝑛 ∈ ℕ such that

𝑛𝑥 > 𝑦 ,

concluding the proof.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 93

4.2 Nested Interval Property

Another consequence of the axiom of completeness is the Nested Interval Property. This is yet another
way of saying the same thing: ℝ does not have gaps.
Let us look at a construction. Suppose given some closed intervals

𝐼𝑛 ∶= [𝑎𝑛 , 𝑏𝑛 ] = {𝑥 ∈ ℝ ∶ 𝑎𝑛 ≤ 𝑥 ≤ 𝑏𝑛 } ,

where the end points are ordered in the following way:

𝑎1 ≤ 𝑎 2 ≤ … ≤ 𝑎 𝑛 ≤ ⋯ ≤ 𝑏 𝑛 ≤ … 𝑏 𝑛 ≤ 𝑏 1 ,

as shown in Figure 4.3.

Figure 4.3: Nested intervals 𝐼𝑛 = [𝑎𝑛 , 𝑏𝑛 ].

The intervals 𝐼𝑛 are nested, meaning that

𝐼1 ⊃ 𝐼2 ⊃ 𝐼3 ⊃ … 𝐼𝑛 ⊃ …

For finite intersections we clearly have

𝑘
⋂ 𝐼𝑛 = 𝐼𝑘 ,
𝑛=1
that is, intersecting the first 𝑘 intervals yields 𝐼𝑘 , the smallest interval in the sequence.

Question 4.4

Consider the infinite intersection

∞
⋂ 𝐼𝑛 ∶= {𝑥 ∈ ℝ ∶ 𝑥 ∈ 𝐼𝑛 , ∀ 𝑛 ∈ ℕ} .
𝑛=1
What can we say about it? Is it empty? Is it not empty?

The answer is that the infinite intersection is not empty, because ℝ was constructed in a way that it does not
have gaps.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 94

Theorem 4.5: Nested Interval Property

For each 𝑛 ∈ ℕ assume given a closed interval

𝐼𝑛 ∶= [𝑎𝑛 , 𝑏𝑛 ] = {𝑥 ∈ ℝ ∶ 𝑎𝑛 ≤ 𝑥 ≤ 𝑏𝑛 } .

Suppose that the intervals are nested, that is,

𝐼𝑛 ⊃ 𝐼𝑛+1 , ∀𝑛 ∈ ℕ.

Then
∞
⋂ 𝐼𝑛 ≠ ∅ . (4.5)
𝑛=1

Proof
By definition we have
∞
⋂ 𝐼𝑛 ∶= {𝑥 ∈ ℝ ∶ 𝑥 ∈ 𝐼𝑛 , ∀ 𝑛 ∈ ℕ} .
𝑛=1
We want to prove (4.5). This means we need to find a real number 𝑥 such that

𝑥 ∈ 𝐼𝑛 , ∀𝑛 ∈ ℕ. (4.6)

Idea of the Proof: Condition (4.6) is saying that it should hold

𝑥 ≥ 𝑎𝑛 , ∀𝑛 ∈ ℕ.

We might be tempted to choose 𝑥 to be any of the 𝑏𝑛 . This choice would indeed satisy the
above. However (4.6) also implies that

𝑥 ≤ 𝑏𝑛 , ∀𝑛 ∈ ℕ.

Therefore 𝑥 has to be larger than all the 𝑎𝑛 , but not too large. This suggests that 𝑥 should be
defined as a supremum.

Define the set

𝐴 ∶= {𝑎𝑛 ∶ 𝑛 ∈ ℕ} .
The set 𝐴 is non-empty and is bounded above by any of the 𝑏𝑛 . Therefore there exists

𝑥 = sup 𝐴 .

By definition of supremum and definition of the set 𝐴, we have

𝑎𝑛 ≤ 𝑥 , ∀𝑛 ∈ ℕ.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 95

On the other hand, consider an arbitrary number 𝑏𝑛 . By construction we have

𝑎𝑖 ≤ 𝑏 𝑛 , ∀𝑖 ∈ ℕ.

Therefore 𝑏𝑛 is an upper bound for 𝐴. Since the supremum is the smallest upper bound, we conclude that

𝑥 ≤ 𝑏𝑛 .

The index 𝑛 was chosen arbitrarily, and therefore

𝑥 ≤ 𝑏𝑛 , ∀𝑛 ∈ ℕ.

In total we have
𝑎𝑛 ≤ 𝑥 ≤ 𝑏 𝑛 , ∀𝑛 ∈ ℕ,
showing that 𝑥 satisfies (4.6). Therefore (4.5) holds and the proof is concluded.

Important

The assumption that 𝐼𝑛 is closed is crucial in Theorem 0.123. Without such assumption the thesis of
Theorem 0.123 does not hold in general, as seen in Example 4.6 below.

Example 4.6

Consider the open intervals

1
𝐼𝑛 ∶= (0, ) .
𝑛
These are clearly nested
𝐼𝑛 ⊃ 𝐼𝑛+1 , ∀𝑛 ∈ ℕ.
For this choice of 𝐼𝑛 we have
∞
⋂ 𝐼𝑛 = ∅ . (4.7)
𝑛=1
Indeed, suppose by contradiction that the intersection is non-empty. Then there exists 𝑥 ∈ ℕ such that
𝑥 ∈ 𝐼𝑛 , ∀𝑛 ∈ ℕ.
By definition of 𝐼𝑛 the above reads
1
0<𝑥< , ∀𝑛 ∈ ℕ. (4.8)
𝑛
Since 𝑥 > 0, by the Archimedean Property in Theorem 0.120 Point 2, there exists 𝑛0 ∈ ℕ such that
1
0< <𝑥.
𝑛0
The above contradicts (4.8). Therefore (4.7) holds.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 96

4.3 Revisiting Sup and inf

We now investigate some of the properties of supremum and infimum in ℝ. The first property is an alternative
characterization of the supremum, which we will often use. A sketch of such characterization is in Figure 4.4
below.

Proposition 4.7: Characterization of Supremum

Let 𝐴 ⊆ ℝ be a non-empty set. Suppose that 𝑠 ∈ ℝ is an upper bound for 𝐴. They are equivalent:

1. 𝑠 = sup 𝐴
2. For every 𝜀 > 0 there exists 𝑥 ∈ 𝐴 such that

𝑠−𝜀 <𝑥.

Figure 4.4: Let 𝑠 = sup 𝐴. Then for every 𝜀 > 0 there exist 𝑥 ∈ 𝐴 such that 𝑠 − 𝜀 < 𝑥.

Proof: Proof of Proposition 0.125

Step 1. Assume that

𝑠 = sup 𝐴 .
Let 𝜀 > 0 be arbitrary. We clearly have that

𝑠−𝜀 < 𝑠. (4.9)

Therefore (𝑠 − 𝜀) cannot be an upper bound of 𝐴. Indeed, if by contradiction (𝑠 − 𝜀) was an upper bound,

then we would have
𝑠 ≤ (𝑠 − 𝜀) ,
since 𝑠 is the smallest upper bound. The above contradicts (4.9), and therefore (𝑠 − 𝜀) is not an upper
bound for 𝐴. Hence there exists some 𝑥 ∈ 𝐴 such that

𝑠−𝜀 <𝑥,

concluding.
Step 2. Assume that Point 2 in the statement of Proposition 0.125 holds. By assumption we have that 𝑠 is
an upper bound for 𝐴. Suppose by contradiction that

𝑠 ≠ sup 𝐴 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 97

This is equivalent to the statement

𝑠 is not the smallest upper bound of 𝐴. (4.10)

Hence there exists an upper bound 𝑏 of 𝐴 such that

𝑏 < 𝑠.

Let
𝜀 ∶= 𝑠 − 𝑏 .
By assumption there exists 𝑥 ∈ 𝐴 such that

𝑠−𝜀 <𝑥.

Substituting the definition of 𝜀 we get

𝑠−𝑠+𝑏 <𝑥 ⟹ 𝑏<𝑥.

Since 𝑏 is an upper bound for 𝐴 and 𝑥 ∈ 𝐴, the above is a contradiction. Therefore (4.10) is false, and 𝑠 is
the smallest upper bound of 𝐴. Thus 𝑠 = sup 𝐴.

The analogue of Proposition 0.125 is as follows. The proof is left as an exercise.

Proposition 4.8: Characterization of Infimum

Let 𝐴 ⊆ ℝ be a non-empty set. Suppose that 𝑖 ∈ ℝ is a lower bound for 𝐴. They are equivalent:

1. 𝑖 = inf 𝐴
2. For every 𝜀 ∈ ℝ, with 𝜀 > 0, there exists 𝑥 ∈ 𝐴 such that

𝑥 <𝑖+𝜀.

A sketch of the characterization in Proposition 0.126 can be found in Figure 4.5 below.

Figure 4.5: Let 𝑖 = inf 𝐴. Then for every 𝜀 > 0 there exist 𝑥 ∈ 𝐴 such that 𝑥 < 𝑖 + 𝜀.

With the above characterizations of supremum and infimum, it is now easier to prove that some candidate
number is the supremum or infimum of some set. As an example, let us characterize supremum and infimum
of an open interval of ℝ.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 98

Proposition 4.9

Let 𝑎, 𝑏 ∈ ℝ with 𝑎 < 𝑏. Let

𝐴 ∶= (𝑎, 𝑏) = {𝑥 ∈ ℝ ∶ 𝑎 < 𝑥 < 𝑏} .
Then
inf 𝐴 = 𝑎 , sup 𝐴 = 𝑏 .

Proof
We will only prove that
inf 𝐴 = 𝑎 ,
since the proof of
sup 𝐴 = 𝑏
is similar.
By definition of 𝐴, we have that
𝑎<𝑥, ∀𝑥 ∈ 𝐴.
The above says that 𝑎 is a lower bound for 𝐴.
Claim. 𝑎 is the largest lower bound of 𝐴.
Proof of Claim. Let 𝐿 be a lower bound for 𝐴, that is,

𝐿≤𝑥, ∀𝑥 ∈ 𝐴.

We have to prove that

𝐿 ≤ 𝑎. (4.11)
Indeed suppose by contradiction that (4.11) does not hold, namely that

𝑎 < 𝐿.

Since 𝑎 < 𝑏 we have

2𝑎 𝑎 + 𝑏 2𝑏
𝑎= < < = 𝑏,
2 2 𝑏
showing that the midpoint between 𝑎 and 𝑏 satisfies

𝑎+𝑏
∈ 𝐴.
2
Since 𝐿 is a lower bound for 𝐴 we have
𝑎+𝑏
𝐿≤ < 𝑏. (4.12)
2
Consider the midpoint
𝑎+𝐿
𝑀 ∶= .
2
We have that
𝑀 ∈ 𝐴.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 99

Indeed, recalling that 𝑎 < 𝐿, we have

2𝑎 𝑎 + 𝐿
𝑎= < =𝑀.
2 2
Moreover by (4.12) we have 𝐿 < 𝑏. Thus

𝑎+𝐿 𝑎+𝑏
𝑀= ≤ < 𝑏.
2 2
This shows 𝑀 ∈ 𝐴.

Moreover
𝑀 < 𝐿.

This is because 𝑎 < 𝐿, and therefore

𝑎 + 𝐿 2𝐿
𝑀= < = 𝐿.
2 2

This is a contradiction, since by assumption 𝐿 is a lower bound for 𝐴, and thus we should have

𝐿≤𝑀.

Therefore (4.11) holds, showing that 𝑎 is the largest lower bound of 𝐴. Thus 𝑎 = inf 𝐴.

As a corollary of the above we have that the maximum and minimum of an open interval do not exist.

Corollary 4.10

Let 𝑎, 𝑏 ∈ ℝ with 𝑎 < 𝑏. Let

𝐴 ∶= (𝑎, 𝑏) = {𝑥 ∈ ℝ ∶ 𝑎 < 𝑥 < 𝑏} .
Then min 𝐴 and max 𝐴 do not exist.

Proof
Suppose by contradiction that min 𝐴 exists. We have shown that if the minimum of a set exists, then it
must be
min 𝐴 = inf 𝐴 .
Since
inf 𝐴 = 𝑎 ,
by Proposition 4.9, we would obtain that
min 𝐴 = 𝑎 .
By definition min 𝐴 ∈ 𝐴, so that 𝑎 ∈ 𝐴. This is contradiction. Then min 𝐴 does not exist.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 100

The proof that max 𝐴 does not exist is similar, and is left as an exercise.

We can also consider intervals for which one or both of the sides are closed.

Corollary 4.11

Let 𝑎, 𝑏 ∈ ℝ with 𝑎 < 𝑏. Let

𝐴 ∶= [𝑎, 𝑏) = {𝑥 ∈ ℝ ∶ 𝑎 ≤ 𝑥 < 𝑏} .
Then
min 𝐴 = inf 𝐴 = 𝑎 , sup 𝐴 = 𝑏 ,
max 𝐴 does not exist.

The proof is very similar to the ones above, and is left to the reader for exercise. Let us now compute supre-
mum and infimum of a set which is not an interval.

Proposition 4.12

Define the set

1
𝐴 ∶= { ∶ 𝑛 ∈ ℕ} .
𝑛
Then
inf 𝐴 = 0 , sup 𝐴 = max 𝐴 = 1 .

Proof
Part 1. We have
1
≤ 1, ∀𝑛 ∈ ℕ.
𝑛
Therefore 1 is an upper bound for 𝐴. Let us prove it is the least upper bound: let 𝑏 be an upper bound
for 𝐴. Since 1 ∈ 𝐴 and 𝑏 is an upper bound, we have 1 ≤ 𝑏. Hence 1 is the least upper bound, and

sup 𝐴 = max 𝐴 = 1 .

Part 2. We have
1
> 0, ∀𝑛 ∈ ℕ,
𝑛
showing that 0 is a lower bound for 𝐴. Suppose by contradiction that 0 is not the infimum. Therefore 0
is not the largest lower bound. Then there exists 𝜀 ∈ ℝ such that:

• 𝜀 is a lower bound for 𝐴, that is,

1
𝜀≤ , ∀𝑛 ∈ ℕ, (4.13)
𝑛
• 𝜀 is larger than 0:
0<𝜀.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 101

As 𝜀 > 0, by the Archimedean Property there exists 𝑛0 ∈ ℕ such that

1
0< <𝜀.
𝑛0

This contradicts (4.13). Thus 0 is the largest lower bound of 𝐴, that is, 0 = inf 𝐴.
Part 3. We have that min 𝐴 does not exist. Indeed suppose by contradiction that min 𝐴 exists. Then

min 𝐴 = inf 𝐴 .

As inf 𝐴 = 0 by Part 2, we conclude min 𝐴 = 0. As min 𝐴 ∈ 𝐴, we obtain 0 ∈ 𝐴, which is a contradiction.

4.4 Density of ℚ in ℝ
A set 𝑆 is dense in ℝ if the elements of 𝑆 are arbitrarily close to the elements of ℝ.

Definition 4.13: Dense set

Let 𝑆 ⊆ ℝ. We say that 𝑆 is dense in ℝ if for every 𝑥 ∈ ℝ and 𝜀 ∈ ℝ with 𝜀 > 0, there exist 𝑞 ∈ ℚ such that

|𝑥 − 𝑞| < 𝜀 .

In other words, the above definition is saying that 𝑆 and ℝ are tightly knitted together. An equivalent defini-
tion of dense set is given below.

Remark 4.14

Let 𝑆 ⊆ ℝ. They are equivalent:

• 𝑆 is dense in ℝ.
• For every pair of numbers 𝑥, 𝑦 ∈ ℝ with 𝑥 < 𝑦, there exists 𝑠 ∈ 𝑆 such that

𝑥 <𝑠<𝑦.

We now prove that ℚ is dense in ℝ.

Theorem 4.15: Density of ℚ in ℝ

Let 𝑥, 𝑦 ∈ ℝ, with 𝑥 < 𝑦. There exists 𝑞 ∈ ℚ such that

𝑥 <𝑞<𝑦.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 102

Figure 4.6: Let 𝑛 ∈ ℕ be such that 1/𝑛 < 𝑦 − 𝑥. Then take 𝑚 so that 𝑚/𝑛 ∈ (𝑥, 𝑦).

Proof
We need to find 𝑞 ∈ ℚ such that
𝑥 <𝑞<𝑦. (4.14)
By definition of ℚ, we have that 𝑞 has to be 𝑞 = 𝑚/𝑛 for 𝑚 ∈ ℤ and 𝑛 ∈ ℕ. Therefore (4.14) is equivalent
to finding 𝑚 ∈ ℤ and 𝑛 ∈ ℕ such that
𝑚
𝑥 < <𝑦. (4.15)
𝑛
The idea is to proceed as in Figure 4.6: We take 𝑛 such that 1/𝑛 is small enough so that we can make 𝑚
jumps of size 1/𝑛 and end up between 𝑥 and 𝑦.
To this end, let 𝑛 ∈ ℕ be such that
1
<𝑦 −𝑥. (4.16)
𝑛
Such 𝑛 exists thanks to the Archimedean Property in Theorem 0.120 Point 2. Inequality (4.15) is equivalent
to
𝑛𝑥 < 𝑚 < 𝑛𝑦 .
Take 𝑚 ∈ ℤ such that
𝑚 − 1 ≤ 𝑛𝑥 < 𝑚 . (4.17)

Why does such 𝑚 exist? Because by Archimedean Property in Theorem 0.120 Point 1, there
exists 𝑚′ ∈ ℤ such that
𝑚′ > 𝑛𝑥
We can then choose 𝑚 to be the smallest element in ℤ such that 𝑚 > 𝑛𝑥. Such 𝑚 satisfies
(4.17).

The second inequality in (4.17) implies

𝑚
𝑥< ,
𝑛
which is the first inequality in (4.15). Now note that (4.16) is equivalent to 𝑥 < 𝑦 − 1/𝑛. We can use the
latter and the first inequality in (4.17) to estimate

𝑚 ≤ 1 + 𝑛𝑥
1
< 1 + 𝑛 (𝑦 − )
𝑛
= 𝑛𝑦 ,

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 103

which yields
𝑚
<𝑦.
𝑛
Therefore the second inequality in (4.15) is proven, concluding the proof.

We have constructed the real numbers ℝ so that they would fill the gaps of ℚ. Formally, these gaps are the
numbers in ℝ ∖ ℚ. Let us give a name to this set.

Definition 4.16: Irrational numbers

The set of irrational numbers in ℝ is

ℐ ∶= ℝ ∖ ℚ .

Question 4.17

How many gaps does ℚ have? In other words, how many irrational numbers are out there?

The answer is quite surprising, and is a corollary of the density result of Theorem 0.133: The irrational
numbers are dense in ℝ.

Corollary 4.18

Let 𝑥, 𝑦 ∈ ℝ, with 𝑥 < 𝑦. There exists 𝑡 ∈ ℐ such that

𝑥 <𝑡 <𝑦.

Proof
Consider
𝑥̃ ∶= 𝑥 − √2 , 𝑦̃ ∶= 𝑦 − √2 .
Since 𝑥 < 𝑦, we have
𝑥̃ < 𝑦̃ .
By Theorem 0.133 there exists 𝑞 ∈ ℚ such that
𝑥̃ < 𝑞 < 𝑦̃ .
Adding √2 to the above inequalities we get
𝑥 <𝑡 <𝑦, 𝑡 ∶= 𝑞 + √2 . (4.18)
We claim that 𝑡 ∈ ℐ . Indeed, suppose by contradiction 𝑡 ∈ ℚ. Then
√2 = 𝑡 − 𝑞 ∈ ℚ ,
since 𝑡, 𝑞 ∈ ℚ, and ℚ is closed under summation. Since √2 ∈ ℐ , we obtain a contradiction. Thus 𝑡 ∈ ℐ
and (4.18) is our thesis.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 104

4.5 Existence of 𝑘-th Roots

We have started our discussion by proving that

√2 ∉ ℚ . (4.19)

We have shown that (4.19) implies that the set

𝐴 ∶= {𝑞 ∈ ℚ ∶ 𝑞 2 < 2}

does not have a supremum in ℚ. We then introduced the Real Numbers ℝ so that each non-empty and bounded
above set would have a supremum. As the set 𝐴 is non-empty and bounded above, there exists 𝛼 ∈ ℝ such
that
𝛼 = sup 𝐴 .
We are going to prove that
𝛼2 = 2 ,
which means that in ℝ we can take the square root of 2. More in general, with the same fudamental idea, we
can prove that for each 𝑥 ∈ ℝ with 𝑥 ≥ 0 and 𝑘 ∈ ℕ, there exists 𝛼 ∈ ℝ such that

𝛼𝑘 = 𝑥 .

Theorem 4.19: Existence of 𝑘-th roots

Let 𝑥 ∈ ℝ with 𝑥 ≥ 0 and 𝑘 ∈ ℕ. There exists a unique 𝛼 ∈ ℝ such that

𝛼𝑘 = 𝑥 .

The proof of Theorem 0.137 rests on similar ideas to the ones used to prove that ℚ does not have the cut
property.

Proof: Proof of Theorem 0.137

Part 1: Uniqueness.
Suppose 𝛼1 , 𝛼2 ∈ ℝ are such that
𝛼1𝑘 = 𝛼2𝑘 = 𝑥 .
If 𝛼1 ≠ 𝛼2 , then
𝛼1𝑘 ≠ 𝛼2𝑘 ,
obtaining a contradiction. Therefore 𝛼1 = 𝛼2 .
Part 2: Existence.
Let 𝑥 ∈ ℝ with 𝑥 ≥ 0. If 𝑥 = 0 there is nothing to prove, as
0𝑘 = 0 ,
so that 𝛼 = 0. Therefore we can assume 𝑥 > 0. Define the subset of ℝ
𝐴 ∶= {𝑡 ∈ ℝ ∶ 𝑡 𝑘 < 𝑥} .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 105

Clearly 𝐴 is non-empty and bounded above.

An upper bound is given, for example, by 𝑏 ∶= 𝑥 + 1. Indeed, since we are assuming 𝑥 ≥ 0,

then 𝑥 + 1 ≥ 1. In particular we have

(𝑥 + 1)𝑘 ≥ 𝑥 + 1 .

Let 𝑡 ∈ 𝐴. Then
𝑡 𝑘 < 𝑥 < 𝑥 + 1 < (𝑥 + 1)𝑘 ,
showing that 𝑡 < 𝑥 + 1.

By the Axiom of Completeness of ℝ, there exists 𝛼 ∈ ℝ such that

𝛼 = sup 𝐴 .

We claim that
𝛼𝑘 = 𝑥 . (4.20)
Suppose by contradiction that (4.20) does not hold. We will need the formula: For all 𝑎, 𝑏 ∈ ℝ it holds

𝑎𝑘 − 𝑏 𝑘 = (𝑎 − 𝑏)(𝑎𝑘−1 + 𝑎𝑘−2 𝑏 + 𝑎𝑘−3 𝑏 𝑘 + … + 𝑎𝑏 𝑘−2 + 𝑏 𝑘−1 ) . (4.21)

Formula (4.21) can be easily proven by induction on 𝑘. Since we are assuming that (4.20) does not hold,
we have two cases:

• 𝛼 𝑘 < 𝑥: We know that 𝛼 is the supremum of 𝐴. We would like to violate this, by finding a number
𝐿 which is larger than 𝛼, but still belongs to 𝐴. This means 𝐿 has to satisfy

𝛼 < 𝐿, 𝐿𝑘 < 𝑥 .

We look for 𝐿 of the form

1
𝐿𝑛 ∶= 𝛼 +
𝑛
for 𝑛 ∈ ℕ to be chosen later. Clearly
𝛼 < 𝐿𝑛 , (4.22)
for all 𝑛 ∈ ℕ. We now search for 𝑛0 ∈ ℕ such that

𝐿𝑘𝑛0 < 𝑥 .

Using formula (4.21) with 𝑎 = 𝛼 and 𝑏 = 𝐿𝑛0 we obtain

1 𝑘−1
𝐿𝑘𝑛0 − 𝛼 𝑘 = (𝐿 + 𝐿𝑘−2
𝑛0 𝛼 + … + 𝐿𝑛0 𝛼
𝑘−2 + 𝛼 𝑘−1 ) . (4.23)
𝑛0 𝑛 0
Now notice that (4.22) implies
𝑗
𝛼 𝑗 < 𝐿𝑛0

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 106

for all 𝑗 ∈ ℕ. Using this estimate on all the terms 𝛼 𝑗 appearing in the RHS of (4.23) we obtain
1 𝑘−1
𝐿𝑘𝑛0 − 𝛼 𝑘 = (𝐿 + 𝐿𝑘−2
𝑛0 𝛼 + … + 𝐿𝑛0 𝛼
𝑘−2 + 𝛼 𝑘−1 )
𝑛0 𝑛 0
1 𝑘−1
< (𝐿 + 𝐿𝑘−2 𝑘−2 𝑘−1
𝑛0 𝐿𝑛0 + … + 𝐿𝑛0 𝐿𝑛0 + 𝐿𝑛0 )
𝑛0 𝑛 0
𝑘 𝑘−1
= 𝐿
𝑛0 𝑛 0
Rearranging the above we get
𝑘 𝑘−1
𝐿𝑘𝑛0 < 𝐿𝑛0 + 𝛼 𝑘 . (4.24)
𝑛0
Now note that
1
𝐿𝑛0 = 𝛼 + < 𝛼 + 1.
𝑛0
Therefore
𝐿𝑘−1
𝑛0 < (𝛼 + 1)
𝑘−1 ,

and from (4.24) we obtain

𝑘
𝐿𝑘𝑛0 < (𝛼 + 1)𝑘−1 + 𝛼 𝑘 .
𝑛0
We wanted to find 𝑛0 ∈ ℕ so that 𝐿𝑘𝑛0 < 𝑥. Therefore we impose
𝑘
(𝛼 + 1)𝑘−1 + 𝛼 𝑘 < 𝑥 ,
𝑛0
and find that the above is satisfied for
𝑘(𝛼 + 1)𝑘−1
𝑛0 > . (4.25)
𝑥 − 𝛼𝑘
Notice that the RHS in (4.25) is a positive real number, since 𝛼 𝑘 < 𝑥 by assumption. Therefore, by
the Archimedean Property of Theorem 0.120 Point 1, there exists 𝑛0 ∈ ℕ satisfying (4.25).
We have therefore shown the existence of 𝑛0 ∈ ℕ such that

𝛼 < 𝐿𝑛0 , 𝐿𝑘𝑛0 < 𝑥 .

The above says that 𝐿𝑛0 ∈ 𝐴 and that

𝐿𝑛0 > 𝛼 = sup 𝐴 ,

which is a contradiction, as sup 𝐴 is an upper bound for 𝐴.

• 𝛼 𝑘 > 𝑥: We know that 𝛼 is the supremum of 𝐴. We would like to find a contradiction, by finding
an upper bound 𝐿 for 𝐴 which is smaller than 𝛼. This means 𝐿 has to satisfy

𝐿<𝛼, 𝐿𝑘 > 𝑥 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 107

Such 𝐿 is an upper bound for 𝐴: If 𝑡 ∈ 𝐴 then

𝑡 𝑘 < 𝑥 < 𝐿𝑘 ⟹ 𝑡 < 𝐿.

We therefore look for 𝐿 of the form

1
𝐿𝑛 ∶= 𝛼 −
𝑛
for 𝑛 ∈ ℕ to be chosen later. In this way

𝐿𝑛 < 𝛼 , (4.26)

for all 𝑛 ∈ ℕ. We now search for 𝑛0 ∈ ℕ such that

𝐿𝑘𝑛0 > 𝑥 .

Using formula (4.21) with 𝑎 = 𝐿𝑛0 and 𝑏 = 𝛼 we obtain

1 𝑘−1
𝛼 𝑘 − 𝐿𝑘𝑛0 = (𝛼 + 𝛼 𝑘−2 𝐿𝑛0 + … + 𝛼𝐿𝑘−2 𝑘−1
𝑛0 + 𝐿𝑛0 ) . (4.27)
𝑛0

Now notice that (4.26) implies

𝑗
𝐿𝑛0 < 𝛼 𝑗
𝑗
for all 𝑗 ∈ ℕ. Using this estimate on all the terms 𝐿𝑛0 appearing in the RHS of (4.27) we obtain
1 𝑘−1
𝛼 𝑘 − 𝐿𝑘𝑛0 = (𝛼 + 𝛼 𝑘−2 𝐿𝑛0 + … + 𝛼𝐿𝑘−2 𝑘−1
𝑛0 + 𝐿 𝑛0 )
𝑛0
1 𝑘−1
< (𝛼 + 𝛼 𝑘−2 𝛼 + … + 𝛼𝛼 𝑘−2 + 𝛼 𝑘−1 )
𝑛0
𝑘 𝑘−1
= 𝛼
𝑛0
Rearranging the above we get
𝑘 𝑘−1
𝐿𝑘𝑛0 > 𝛼 𝑘 − 𝛼 .
𝑛0
We wanted to find 𝑛0 ∈ ℕ so that 𝐿𝑘𝑛0 > 𝑥. Therefore we impose

𝑘 𝑘−1
𝛼𝑘 − 𝛼 >𝑥,
𝑛0
and find that the above is satisfied for

𝑘𝛼 𝑘−1
𝑛0 > . (4.28)
𝛼𝑘 − 𝑥

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 108

Notice that the RHS in (4.28) is a positive real number, since 𝛼 𝑘 > 𝑥 by assumption. Therefore, by
the Archimedean Property of Theorem 0.120 Point 1, there exists 𝑛0 ∈ ℕ satisfying (4.28).
We have therefore shown the existence of 𝑛0 ∈ ℕ such that

𝐿𝑛0 < 𝛼 , 𝐿𝑘𝑛0 > 𝑥 .

Condition 𝐿𝑘𝑛0 > 𝑥 says that 𝐿𝑛0 is an upper bound for 𝐴. At the same time it holds

𝐿𝑛0 < 𝛼 = sup 𝐴 ,

which is a contradiction, as sup 𝐴 is the smallest upper bound for 𝐴.

Therefore, both cases 𝛼 𝑘 > 𝑥 and 𝛼 𝑘 < 𝑥 lead to a contradiction. Hence 𝛼 𝑘 = 𝑥, concluding.

Definition 4.20: 𝑘-th root of a number

Let 𝑥 ∈ ℝ with 𝑥 ≥ 0 and 𝑘 ∈ ℕ. The real number 𝛼 such that

𝛼𝑘 = 𝑥

is called the 𝑘-th root of 𝑥, and is denoted by

𝑘
√ 𝑥 ∶= 𝛼 .

4.6 Cardinality
We have proven that the sets or rational numbers ℚ and irrational numbers ℐ are both dense in ℝ, with

ℝ=ℚ∪ℐ .

From this result we might think that ℝ is obtained by mixing ℚ and ℐ in equal proportions. This is however
false. We will see that ℝ has much more elements than ℚ. Therefore also the set of irrational numbers ℐ is
much larger than ℚ.
To make the above discussion precise, we need to define what we mean by size of a set. For this, we need the
concept of bijective function.

Definition 4.21: Bijective function

Let 𝑋 , 𝑌 be sets and 𝑓 ∶ 𝑋 → 𝑌 be a function. We say that:

• 𝑓 is injective if it holds:
𝑓 (𝑥) = 𝑓 (𝑦) ⟹ 𝑥 =𝑦.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 109

• 𝑓 is surjective if it holds:
∀𝑦 ∈ 𝑌 , ∃ 𝑥 ∈ 𝑋 s.t. 𝑓 (𝑥) = 𝑦 .
• 𝑓 is bijective if it is both injective and surjective.

In other words: A function 𝑓 ∶ 𝑋 → 𝑌 is

• injective if any two different elements in 𝑋 are mapped into two different elements in 𝑌 .
• surjective if every element in 𝑌 has at least one element in 𝑋 associated via 𝑓 .
• bijective if to each element in 𝑋 we associate one and only one element in 𝑌 via 𝑓 .

Example 4.22: Injectivity

Consider the sets

𝑋 = {1, 2, 3} , 𝑌 = {𝑎, 𝑏, 𝑐, 𝑑, 𝑒} .

• The function 𝑓 ∶ 𝑋 → 𝑌 defined by

𝑓 (1) = 𝑐 , 𝑓 (2) = 𝑎 , 𝑓 (3) = 𝑒 ,

is injective.

• The function 𝑔 ∶ 𝑋 → 𝑌 defined by

𝑔(1) = 𝑐 , 𝑔(2) = 𝑎 , 𝑔(3) = 𝑐 ,

is not injective, since

𝑔(1) = 𝑔(3) = 𝑐 , 1 ≠ 3.

• The function ℎ ∶ ℝ → ℝ defined by

ℎ(𝑥) = 𝑥 2
is not injective, since
ℎ(1) = ℎ(−1) = 1 , 1 ≠ −1 .

• The function 𝑙 ∶ ℝ → ℝ defined by

𝑙(𝑥) = 2𝑥
is injective, since
𝑙(𝑥) = 𝑙(𝑦) ⟹ 2𝑥 = 2𝑦 ⟹ 𝑥 =𝑦.

Example 4.23: Surjectivity

Consider the sets

𝑋 = {1, 2, 3, 4} , 𝑌 = {𝑎, 𝑏, 𝑐} .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 110

• The function 𝑓 ∶ 𝑋 → 𝑌 defined by

𝑓 (1) = 𝑐 , 𝑓 (2) = 𝑎 , 𝑓 (3) = 𝑎 , 𝑓 (4) = 𝑏 ,

is surjective.

• The function 𝑔 ∶ 𝑋 → 𝑌 defined by

𝑔(1) = 𝑎 , 𝑔(2) = 𝑎 , 𝑔(3) = 𝑐 , 𝑔(4) = 𝑎 ,

is not surjective, since there is no element 𝑥 ∈ 𝑋 such that

𝑔(𝑥) = 𝑏 .

• The function ℎ ∶ ℝ → ℝ defined by

ℎ(𝑥) = 𝑥 2
is not surjective, since there is no 𝑥 ∈ ℝ such that

ℎ(𝑥) = 𝑥 2 = −1 .

• The function 𝑙 ∶ ℝ → [0, ∞) defined by

𝑙(𝑥) = 𝑥 2
is surjective, since for every 𝑦 ≥ 0 there exists 𝑥 ∈ ℝ such that

𝑙(𝑥) = 𝑥 2 = 𝑦 .

This is true by Theorem 0.137.

Example 4.24: Bijectivity

• Let 𝑋 = {1, 2, 3}, 𝑌 = {𝑎, 𝑏, 𝑐}. The function 𝑓 ∶ 𝑋 → 𝑌 defined by

𝑓 (1) = 𝑐 , 𝑓 (2) = 𝑎 , 𝑓 (3) = 𝑏 ,

is bijective, since it is both injective and surjective.

• Let 𝑋 = {1, 2, 3}, 𝑌 = {𝑎, 𝑏}. The function 𝑔 ∶ 𝑋 → 𝑌 defined by

𝑔(1) = 𝑎 , 𝑔(2) = 𝑏 , 𝑔(3) = 𝑏 ,

is not bijective, since it is not injective: we have

𝑔(2) = 𝑔(3) = 𝑏 2 ≠ 3.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 111

• Let 𝑋 = {1, 2}, 𝑌 = {𝑎, 𝑏, 𝑐}. The function ℎ ∶ ℝ → ℝ defined by

ℎ(1) = 𝑎 , ℎ(2) = 𝑐 ,

is not bijective, since it is not surjective: there is no 𝑥 ∈ 𝑋 such that

ℎ(𝑥) = 𝑏 .

• Let 𝑋 = {1, 2, 3}, 𝑌 = {𝑎, 𝑏, 𝑐}. The function 𝑙 ∶ ℝ → [0, ∞) defined by

𝑙(1) = 𝑎 , 𝑙(2) = 𝑎 , 𝑙(3) = 𝑏 ,

is not bijective, as it is neither injective nor surjective.

We are ready to define the size of a set.

Definition 4.25: Cardinality, finite, countable, uncountable

Let 𝑋 be a set. The cardinality of 𝑋 is the number of elements in 𝑋 . We denote the cardinality of 𝑋 by

|𝑋 | ∶= # of elements in 𝑋 .

Further, we say that:

• 𝑋 is finite if there exists a natural number 𝑛 ∈ ℕ and a bijection

𝑓 ∶ 𝑋 → {1, 2, … , 𝑛} .

In particular
|𝑋 | = 𝑛 .

• 𝑋 is countable if there exists a bijection

𝑓 ∶ 𝑋 → ℕ.

In this case we denote the cardinality of 𝑋 by

|𝑋 | = |ℕ| .

• 𝑋 is uncountable if 𝑋 is neither finite, nor countable.

In other words: A set 𝑋 is

• finite, if 𝑋 can be listed as

𝑋 = {𝑥1 , … , 𝑥𝑛 }

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 112

for some 𝑛 ∈ ℕ.
• countable, if 𝑋 can be listed as
𝑋 = {𝑥𝑛 ∶ 𝑛 ∈ ℕ} .

• uncountable, if 𝑋 cannot be listed.

Question 4.26

Is there an intermediate cardinality between finite and countable?

The answer is no, as shown in the next proposition.

Proposition 4.27

Let 𝑋 be a countable set and 𝐴 ⊆ 𝑋 . Then either 𝐴 is finite or countable.

Proof
If 𝐴 is finite we are done. Therefore suppose 𝐴 is not finite. Since 𝑋 is countable there exists a bijection
𝑓 ∶ ℕ → 𝑋 . Let 𝑛1 ∈ ℕ be such that

𝑛1 = min{𝑛 ∈ ℕ s.t. 𝑓 (𝑛) ∈ 𝐴} .

Note that 𝑛1 exists since 𝑓 is surjective. Define

𝑔(1) ∶= 𝑓 (𝑛1 ) .

Now let
𝑛2 = min{𝑛 ∈ ℕ s.t. 𝑛 > 𝑛1 , 𝑓 (𝑛) ∈ 𝐴} .
Notice that 𝑛2 exists, since 𝑓 is surjective and 𝐴 is not finite. Set

𝑔(2) ∶= 𝑓 (𝑛2 ) .

Iterating, we define
𝑛𝑘 = min{𝑛 ∈ ℕ s.t. 𝑛 > 𝑛𝑘−1 , 𝑓 (𝑛) ∈ 𝐴}
and
𝑔(𝑘) ∶= 𝑓 (𝑛𝑘 ) .
In this way we have defined a function 𝑔 ∶ ℕ → 𝐴. We have:

• 𝑔 is injective: This is because 𝑔 was defined thourgh 𝑓 , and 𝑓 is injective.

• 𝑔 is surjective: If 𝑥 ∈ 𝐴, by surjectivity of 𝑓 there exists 𝑛̃ ∈ ℕ such that 𝑓 (𝑛)̃ = 𝑥. Therefore

𝑛̃ ∈ {𝑛 ∈ ℕ s.t. 𝑓 (𝑛) ∈ 𝐴} ,

so that 𝑛𝑘 = 𝑛,̃ for some 𝑘 ≥ 𝑛̃ − 1.

Hence 𝑔 is bijective, showing that 𝐴 is countable.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 113

Example 4.28

1. Let 𝑋 = {𝑎, 𝑏, 𝑐} and 𝑌 = {1, 2, 3}. The function 𝑓 ∶ 𝑋 → 𝑌 defined by

𝑓 (1) = 𝑎 , 𝑓 (2) = 𝑏 , 𝑓 (3) = 𝑐 ,
is bijective. Therefore 𝑋 is finite, with |𝑋 | = 3.
2. Let 𝑋 = ℕ. The function 𝑓 ∶ 𝑋 → ℕ defined by
𝑓 (𝑛) ∶= 𝑛 ,
is bijective. Therefore 𝑋 = ℕ is countable.
3. Let 𝑋 be the set of even numbers
𝑋 = {2𝑛 ∶ 𝑛 ∈ ℕ} .
Define the map 𝑓 ∶ 𝑋 → ℕ by
𝑚
𝑓 (𝑚) ∶= .
2
We have that:
• 𝑓 is injective: 𝑓 (𝑚) = 𝑓 (𝑘) implies that 𝑚/2 = 𝑘/2 which implies 𝑚 = 𝑘.
• 𝑓 is surjective: If 𝑛 ∈ ℕ, then 𝑓 (2𝑛) = 𝑛.
Therefore 𝑓 is bijective, showing that 𝑋 is countable and |𝑋 | = |ℕ|.
4. Let 𝑋 = ℤ the set of integers. Define 𝑓 ∶ ℕ → ℤ by
𝑛
if 𝑛 even
2
𝑓 (𝑛) ∶= { 𝑛 + 1
− if 𝑛 odd
2
For example
𝑓 (0) = 0 , 𝑓 (1) = −1 , 𝑓 (2) = 1 , 𝑓 (3) = −2 ,
𝑓 (4) = 2 , 𝑓 (5) = −3 , 𝑓 (6) = 3 , 𝑓 (7) = −4 .
We have:
• 𝑓 is injective: Indeed, suppose that 𝑚 ≠ 𝑛. If 𝑛 and 𝑚 are both even or both odd we have,
respectively
𝑚 𝑛
𝑓 (𝑚) = ≠ = 𝑓 (𝑛)
2 2
𝑚+1 𝑛+1
𝑓 (𝑚) = − ≠− = 𝑓 (𝑛) .
2 2
If instead 𝑚 is even and 𝑛 is odd, we get
𝑚 𝑛+1
𝑓 (𝑚) =
≠− = 𝑓 (𝑛) ,
2 2
since the LHS is positive and the RHS is negative. The case when 𝑚 is odd and 𝑛 even is
similar.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 114

• 𝑓 is surjective: Let 𝑧 ∈ ℤ. If 𝑧 ≥ 0, then 𝑚 ∶= 2𝑧 belongs to ℕ, is even, and

𝑓 (𝑚) = 𝑓 (2𝑧) = 𝑧 .

If instead 𝑧 < 0, then 𝑚 ∶= −2𝑧 − 1 belongs to ℕ, is odd, and

𝑓 (𝑚) = 𝑓 (−2𝑧 − 1) = 𝑧 .

Therefore 𝑓 is bijective, showing that ℤ is countable and

|ℤ| = |ℕ| .

We have seen that the sets ℕ and ℤ are countable. What about ℚ? To study this case, we need the following
result.

Proposition 4.29

Let the set 𝐴𝑛 be countable for all 𝑛 ∈ ℕ. Define

𝐴 = ⋃ 𝐴𝑛 .
𝑛∈ℕ

Then 𝐴 is countable.

Proof
Since each 𝐴𝑖 is countable, we can list their elements as

𝐴𝑖 = {𝑎𝑘𝑖 ∶ 𝑘 ∈ ℕ} = {𝑎1𝑖 , 𝑎2𝑖 , 𝑎3𝑖 , 𝑎4𝑖 , …} .

The proof that 𝐴 is countable is based on a diagonal argument by Georg Cantor, see Wikipedia page.
The idea of th is that we can list the elements of the sets 𝐴𝑖 in an infinite square: In the first row we put
the elements of 𝐴1 , in the second row the elements of 𝐴2 , and so on. Therefore the 𝑖-th row contains
the elements of 𝐴𝑖 . This procedure is illustrated in Figure 4.7. Therefore this infinite square contains all
the elements of 𝐴. We then list all the elements of the square by looking at the diagonals, as shown in
Figure 4.7. This procedure defines a function 𝑓 ∶ ℕ → 𝐴. For example the first few terms of 𝑓 are

𝑓 (1) = 𝑎11 , 𝑓 (2) = 𝑎21 , 𝑓 (3) = 𝑎12 , 𝑓 (4) = 𝑎13 .

Since 𝑓 is injective and surjective, we have that 𝐴 is countable.

Theorem 4.30: ℚ is countable

The set of rational numbers ℚ is countable.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 115

Figure 4.7: The i-th row contains all the elements 𝑎1𝑖 , 𝑎2𝑖 , 𝑎3𝑖 , … of the countable set 𝐴𝑖 . We define the function
𝑓 ∶ ℕ → 𝐴 by going throgh the square diagonally.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 116

Proof
For 𝑖 ∈ ℕ define the sets
𝑚
𝐿𝑖 ∶= { ∶ 𝑚 ∈ ℤ} .
𝑖
We have that 𝑓 ∶ 𝐿𝑖 → ℤ defined by
𝑚
𝑓 ( ) ∶= 𝑚
𝑖
is a bijection. As ℤ is countable, we deduce that 𝐿𝑖 is countable. Therefore the set 𝐿 defined by

𝐿 ∶= ⋃ 𝐿𝑖
𝑖∈ℕ

is countable by Proposition 4.29. Clearly we have

ℚ ⊆ 𝐿.

Since ℚ is not finite, by Proposition 4.27 we conclude that ℚ is countable.

We have proven that the sets

ℕ, ℤ, ℚ,
are all countable. What about ℝ?

Theorem 4.31: ℝ is uncountable

The set of Real Numbers ℝ is uncountable.

Proof
Suppose by contradiction ℝ is countable. Then there exists a bijection 𝑓 ∶ ℕ → ℝ, meaning that we can
list the elements of ℝ as
ℝ = {𝑥1 , 𝑥2 , 𝑥3 , 𝑥4 , 𝑥5 , …} .
Let 𝐼1 be a closed interval such that
𝑥1 ∉ 𝐼1 .
Let 𝐼2 be another closed interval, contained in 𝐼1 , and such that 𝑥2 ∉ 𝐼2 . Such interval exists, because
𝐼1 contains two disjoint closed intervals: hence 𝑥2 can be at most in one of these two intervals. To
summarize, we have
𝑥1 ∉ 𝐼1 , 𝑥2 ∉ 𝐼2 , 𝐼2 ⊆ 𝐼1 .
We can iterate this procedure, and construct a sequence of nested intervals 𝐼𝑛 such that
𝐼𝑛+1 ⊆ 𝐼𝑛 , 𝑥𝑛 ∉ 𝐼𝑛 ,
for all 𝑛 ∈ ℕ, see Figure 4.8. Since 𝑥𝑘 ∉ 𝐼𝑘 , we conclude that
∞
𝑥𝑘 ∉ ⋂ 𝐼𝑛 , ∀𝑘 ∈ ℕ.
𝑛=1

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 117

As the points 𝑥𝑘 are all the elements of ℝ, we conclude that

∞
⋂ 𝐼𝑛 = ∅ .
𝑛=1

However the Nested Interval Property of Theorem 0.123 implies that

∞
⋂ 𝐼𝑛 ≠ ∅ ,
𝑛=1

yielding a contradiction. Therefore ℝ is uncountable.

Figure 4.8: The intervals 𝐼𝑛 are nested, and can be chosen so that 𝑥𝑛 ∉ 𝐼𝑛 .

As a corollary we obtain that also the irrational numbers are uncountable.

Theorem 4.32

The set of irrational numbers

ℐ ∶= ℝ ∖ ℚ
is uncountable.

Proof
In Theorems 0.148, 0.149 we have shown that ℝ in uncountable and ℚ is countable. Suppose by contra-
diction that ℐ is countable. Then
ℚ∪ℐ
is countable by Proposition 4.29, being union of countable sets. Since by definition

ℝ=ℚ∪ℐ ,

we conclude that ℝ is countable. Contradiction.

Dr. Silvio Fanzon [email protected]

5 Complex Numbers
We have seen that √𝑥 exists in ℝ for all for 𝑥 ≥ 0. We defined

√𝑥 ∶= 𝛼 , 𝛼 ∶= sup{𝑡 ∈ ℝ ∶ 𝑡 2 < 𝑥} ,

and proved that

𝛼2 = 𝑥 .
This procedure was possible for 𝑥 ≥ 0.

Question 5.1

Is there a number 𝛼 ∈ ℝ such that

𝛼 2 = −1 ? (5.1)

The answer to the above question is no. This is because ℝ is an ordered field, and from axiom (MO) it follows
that:
𝑥2 ≥ 0 , ∀ 𝑥 ∈ ℝ .
However we would still like to solve equation (5.1) somehow. To do this, we introduce the imaginary num-
bers or complex numbers. We define 𝑖 to be that number such that

𝑖2 = −1 .

Formally, we can also think of 𝑖 = √−1. We can use this speacial number to define the square root of a negative
number 𝑥 < 0:
√𝑥 ∶= 𝑖√−𝑥 .
Note that √−𝑥 is properly defined in ℝ, because −𝑥 > 0 if 𝑥 < 0.

5.1 The field ℂ

We would like to be able to do calculations with the newly introduced complex numbers, and investigate their
properties. We can introduce them rigorously as a field, as we did for ℝ.

Definition 5.2: Complex Numbers

The set of complex numbers ℂ is defined as

118
Numbers, Sequences and Series Page 119

ℂ ∶= ℝ ⊕ 𝑖ℝ ∶= {𝑥 ⊕ 𝑖𝑦 ∶ 𝑥, 𝑦 ∈ ℝ} .

In the above the symbol ⊕ is used to denote the pair

𝑥 ⊕ 𝑖𝑦 = (𝑥, 𝑦)

with 𝑥, 𝑦 ∈ ℝ. This means that 𝑥 and 𝑦 play different roles.

Definition 5.3

For a complex number

𝑧 = 𝑥 ⊕ 𝑖𝑦 ∈ ℂ
we say that

• 𝑥 is the real part of 𝑧, and denote it by

𝑥 = Re(𝑧)

• 𝑦 is the imaginary part of 𝑧, and denote it by

𝑦 = Im(𝑧)

We say that

• If Re 𝑧 = 0 then 𝑧 is a purely imaginary number.

• If Im 𝑧 = 0 then 𝑧 is a real number.

In order to make the set ℂ into a field, we first have to define the two binary operations of addition + and
multiplication ⋅,
+, ⋅ ∶ ℂ × ℂ → ℂ .
Then we need to prove that these operations satisfy all the field axioms.

Definition 5.4: Addition in ℂ

Let 𝑧1 , 𝑧2 ∈ ℂ, so that
𝑧1 = 𝑥1 ⊕ 𝑖𝑦1 , 𝑧2 = 𝑥2 ⊕ 𝑖𝑦2 ,
for some 𝑥1 , 𝑥2 , 𝑦1 , 𝑦2 ∈ ℝ. We define the sum of 𝑧1 and 𝑧2 as

𝑧1 + 𝑧2 = (𝑥1 ⊕ 𝑖𝑦1 ) + (𝑥2 ⊕ 𝑖𝑦2 ) ∶= (𝑥1 + 𝑥2 ) ⊕ 𝑖 (𝑦1 + 𝑦2 )

where the + symbol on the right hand side is the addition operator in ℝ.

Clearly, 𝑧1 + 𝑧2 as defined above is an element of ℂ. Therefore + defines a binary operation over ℂ.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 120

Notation 5.5

From the above definition, we have that, for all 𝑥, 𝑦 ∈ ℝ,

(𝑥 ⊕ 𝑖0) + (0 ⊕ 𝑖𝑦) = 𝑥 ⊕ 𝑖𝑦.

To simplify notation, we will write

𝑥 ⊕ 𝑖0 = 𝑥 , 0 ⊕ 𝑖𝑦 = 𝑖𝑦

and
𝑥 ⊕ 𝑖𝑦 = 𝑥 + 𝑖𝑦 .
We will also often swap 𝑖 and 𝑦, writing equivalently

𝑥 + 𝑖𝑦 = 𝑥 + 𝑦𝑖 .

We now want to define multiplication between complex numbers.

Remark 5.6: Formal calculation for multiplication in ℂ

How to define multiplication in ℂ? Whatever the definition may be, at least it has to give that that

𝑖2 = 𝑖 ⋅ 𝑖 = −1 .

Keeping the above in mind, let us do some formal calculations: For 𝑧1 = 𝑥1 + 𝑖𝑦1 , 𝑧2 = 𝑥2 + 𝑖𝑦2 we have

𝑧1 ⋅ 𝑧2 = (𝑥1 + 𝑖𝑦1 ) ⋅ (𝑥2 + 𝑖𝑦2 )

= 𝑥1 ⋅ 𝑥2 + 𝑥1 ⋅ 𝑖𝑦2 + 𝑥2 ⋅ 𝑖𝑦1 + 𝑦1 ⋅ 𝑖2 𝑦2
= (𝑥1 ⋅ 𝑥2 − 𝑦1 ⋅ 𝑦2 ) + 𝑖 (𝑥1 ⋅ 𝑦2 + 𝑥2 ⋅ 𝑦1 )

Remark 0.156 motivates the following definition of multiplication.

Definition 5.7: Multiplication in ℂ

Let 𝑧1 , 𝑧2 ∈ ℂ, so that
𝑧1 = 𝑥1 ⊕ 𝑖𝑦1 , 𝑧2 = 𝑥2 ⊕ 𝑖𝑦2 ,
for some 𝑥1 , 𝑥2 , 𝑦1 , 𝑦2 ∈ ℝ. We define the multiplication of 𝑧1 and 𝑧2 as

𝑧1 ⋅ 𝑧2 = (𝑥1 + 𝑖𝑦1 ) ⋅ (𝑥2 + 𝑖𝑦2 ) ∶= (𝑥1 ⋅ 𝑥2 − 𝑦1 ⋅ 𝑦2 ) + 𝑖 (𝑥1 ⋅ 𝑦2 + 𝑥2 ⋅ 𝑦1 ) ,

where the operations + and ⋅ on the right hand side are the operations in ℝ.

Clearly, 𝑧1 ⋅ 𝑧2 as defined above is an element of ℂ. Therefore ⋅ defines a binary operation over ℂ.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 121

Remark 5.8

To check that we have given a good definition of product, we should have that

𝑖2 = −1 ,

as expected. Indeed:

𝑖2 = (0 + 1𝑖) ⋅ (0 + 1𝑖)
= (0 ⋅ 0 − 1 ⋅ 1) + (0 ⋅ 1 + 0 ⋅ 1)𝑖 = −1 .

Important

In view of Remark 5.8, we see that he formal calculations in Remark 0.156 are compatible with the defini-
tion of multiplication of complex numbers. Therefore, it is not necessary to memorize the multiplication
formula, but it suffices to carry out calculations as usual, and replace 𝑖2 by −1.

Example 5.9

Suppose we want to multiply the complex numbers

𝑧 = −2 + 3𝑖 , 𝑤 = 1 − 𝑖.

Using the definition we compute

𝑧 ⋅ 𝑤 = (−2 + 3𝑖) ⋅ (1 − 𝑖)
= (−2 − (−3)) + (2 + 3)𝑖
= 1 + 5𝑖 .

Alternatively, we can proceed formally as in Remark 0.156. We just need to recall that 𝑖2 has to be replaced
with −1:

𝑧 ⋅ 𝑤 = (−2 + 3𝑖) ⋅ (1 − 𝑖)
= −2 + 2𝑖 + 3𝑖 − 3𝑖2
= (−2 + 3) + (2 + 3)𝑖
= 1 + 5𝑖 .

We now want to check that

(ℂ, +, ⋅)
is a field. All the field axioms are trivial to check, except for the existence of additive and multiplicative
inverses.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 122

Proposition 5.10: Additive inverse in ℂ

The neutral element of addition in ℂ is the number

0 ∶= 0 + 0𝑖 .

For any 𝑧 = 𝑥 + 𝑖𝑦 ∈ ℂ, the unique additive inverse is given by

−𝑧 ∶= −𝑥 − 𝑖𝑦 .

The proof is immediate and is left as an exercise. The multiplication requires more care.

Remark 5.11: Formal calculation for multiplicative inverse

Let us first carry our some formal calculations. Let

𝑧 = 𝑥 + 𝑖𝑦 ∈ ℂ , 𝑧 ≠ 0.

First, note that

𝑧 ⋅ 1 = (𝑥 + 𝑖𝑦) ⋅ (1 + 0𝑖) = 𝑥 + 𝑖𝑦 = 𝑧 ,
and therefore 1 is the neutral element of multiplication. Thus, the inverse of 𝑧 should be a complex
number 𝑧 −1 ∈ ℂ such that
𝑧 ⋅ 𝑧 −1 = 1 .
We would like to define
1
𝑧 −1 = .
𝑥 + 𝑖𝑦
Such number does not belong to ℂ, as it is not of the form 𝑎 + 𝑖𝑏 for some 𝑎, 𝑏 ∈ ℝ. However it is what
the inverse should look like. Proceeding formally:
1 1
= ⋅1
𝑥 + 𝑖𝑦 𝑥 + 𝑖𝑦
1 𝑥 − 𝑖𝑦
= ⋅
𝑥 + 𝑖𝑦 𝑥 − 𝑖𝑦
𝑥 − 𝑖𝑦
= 2
𝑥 − (𝑖𝑦)2
𝑥 − 𝑖𝑦
= 2
𝑥 + 𝑦2
𝑥 −𝑦
= 2 2
+𝑖 2 .
𝑥 +𝑦 𝑥 + 𝑦2

The right hand side is an element of ℂ, and looks like a good candidate for 𝑧 −1 .

Motivated by the above remark, we define inverses in ℂ in the following way.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 123

Proposition 5.12: Multiplicative inverse in ℂ

The neutral element of multiplication in ℂ is the number

1 ∶= 1 + 0𝑖 .

For any 𝑧 = 𝑥 + 𝑖𝑦 ∈ ℂ, the unique multiplicative inverse is given by

𝑥 −𝑦
𝑧 −1 ∶= +𝑖 2 .
𝑥2 +𝑦 2 𝑥 + 𝑦2

Proof
It is immediate to check that 1 is the neutral element of multiplication in ℂ. For the remaining part of
the statement, set
𝑥 −𝑦
𝑤 ∶= 2 2
+𝑖 2 .
𝑥 +𝑦 𝑥 + 𝑦2
We need to check that 𝑧 ⋅ 𝑤 = 1
𝑥 −𝑦
𝑧 ⋅ 𝑤 = (𝑥 + 𝑖𝑦) ⋅ ( + 𝑖 )
𝑥2 + 𝑦2 𝑥2 + 𝑦2
𝑥2 𝑦 ⋅ (−𝑦) 𝑥 ⋅ (−𝑦) 𝑥𝑦
=( 2 − ) + 𝑖 ( + )
𝑥 + 𝑦2 𝑥2 + 𝑦2 𝑥2 + 𝑦2 𝑥2 + 𝑦2
= 1,

so indeed 𝑧 −1 = 𝑤.

Important

It is not necessary to memorize the formula for 𝑧 −1 . Indeed one can just remember the trick of multiplying
by
𝑥 − 𝑖𝑦
1= ,
𝑥 − 𝑖𝑦
and proceed formally, as done in Remark 0.161.

Example 5.13

Let 𝑧 = 3 + 2𝑖. We want to compute 𝑧 −1 . By the formula in Propostion 0.162 we immediately get
3 −2 3 2
𝑧 −1 = + 2 𝑖= − 𝑖.
32 +2 2 3 +2 2 13 13

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 124

Alternatively, we can proceed formally as in Remark 0.161

1
(3 + 2𝑖)−1 =
3 + 2𝑖
1 3 − 2𝑖
=
3 + 2𝑖 3 − 2𝑖
3 − 2𝑖
= 2
3 + 22
3 2
= − 𝑖,
13 13
and obtain the same result.

We can now prove that ℂ is a field.

Theorem 5.14

(ℂ, +, ⋅) is a field.

Proof
We need to check that all field axioms hold. For the addition we have

• (A1) To show that + is commutative, note that

(𝑥 + 𝑖𝑦) + (𝑎 + 𝑖𝑏) = (𝑥 + 𝑎) + 𝑖(𝑦 + 𝑏)

= (𝑎 + 𝑥) + 𝑖(𝑏 + 𝑦)
= (𝑎 + 𝑖𝑏) + (𝑥 + 𝑖𝑦) ,

where we used Definition 0.154 in the first and last equality, and the commutative property of the
real numbers (which holds since by definition ℝ is a field) in the second equality. Associativity can
be checked in the same way.
• (A2) The neutral element of addition is 0, as stated in Proposition 0.160.
• (A3) Existence of additive inverses is given by Proposition 0.160.

For multiplication we have:

• (M1) Commutativity and associativity of product in ℂ can be checked using Definition 0.157 and
commutativity and associativity of sum and multiplication in ℝ.
• (M2) The neutral element of multiplication is 1, as stated in Propostion 0.162.
• (M3) Existence of multiplicative inverses is guaranteed by Proposition 0.162.

Finally one should check the associative property (AM). This is left as an exercise.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 125

5.1.1 Division in ℂ

Suppose we want to divide two complex numbers 𝑤, 𝑧 ∈ ℂ, 𝑧 ≠ 0, with

𝑧 = 𝑥 + 𝑖𝑦 , 𝑤 = 𝑎 + 𝑖𝑏 .

We have two options:

1. Use the formula for the inverse from Proposition 0.162 and compute
𝑥 −𝑦
𝑧 −1 ∶= + 𝑖 .
𝑥2 + 𝑦2 𝑥2 + 𝑦2

Then we use the multiplication formula of Definition 0.157 to compute

𝑤
= 𝑤 ⋅ 𝑧 −1
𝑧
𝑥 −𝑦
= (𝑎 + 𝑖𝑏) ⋅ ( +𝑖 2 )
𝑥2
+𝑦 2 𝑥 + 𝑦2
(𝑎𝑥 + 𝑏𝑦) + 𝑖(𝑏𝑥 − 𝑎𝑦)
=
𝑥2 + 𝑦2

2. Proceed formally as in Remark 0.161, using the multiplication by 1 trick. We would have

𝑤 𝑎 + 𝑖𝑏
=
𝑧 𝑥 + 𝑖𝑦
𝑎 + 𝑖𝑏 𝑥 − 𝑖𝑦
=
𝑥 + 𝑖𝑦 𝑥 − 𝑖𝑦
(𝑎𝑥 + 𝑏𝑦) + 𝑖(𝑏𝑥 − 𝑎𝑦)
=
𝑥2 + 𝑦2

Example 5.15
𝑤
Let 𝑤 = 1 + 𝑖 and 𝑧 = 3 − 𝑖. We compute 𝑧
using the two options we have:

1. Using the formula for the inverse from Proposition 0.162 we compute
𝑥 −𝑦
𝑧 −1 = + 𝑖
𝑥2 + 𝑦2 𝑥2 + 𝑦2
3 −1
= 2 2
−𝑖 2
3 +1 3 + 12
3 1
= + 𝑖
10 10

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 126

and therefore
𝑤
= 𝑤 ⋅ 𝑧 −1
𝑧
3 1
= (1 + 𝑖) ( + 𝑖)
10 10
3 1 1 3
= ( − ) + ( + )𝑖
10 10 10 10
2 4
= + 𝑖
10 10
1 2
= + 𝑖
5 5

2. We proceed formally, using the multiplication by 1 trick. We have

𝑤 1+𝑖
=
𝑧 3−𝑖
1+𝑖3+𝑖
=
3−𝑖3+𝑖
3 − 1 + (3 + 1)𝑖
=
32 + 1 2
2 4
= + 𝑖
10 10
1 2
= + 𝑖
5 5

5.1.2 ℂ is not ordered

We have seen that (ℂ, +, ⋅) is a field. One might wonder whether ℂ is also an ordered field. It turns out that
this is not the case.

Theorem 5.16

The field (ℂ, +, ⋅) is not ordered.

Proof
Suppose that ℂ is an ordered field, that is, there exists an order relation ≤ on ℂ compatible with the
operations + and ⋅. By axiom (MO) it follows that for all elements 𝑧 ∈ ℂ, 𝑧 ≠ 0, we have that 𝑧 2 > 0. But
since 𝑖2 = −1 < 0, we get a contradiction.

Hence, it is not possible to compare two complex numbers.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 127

5.1.3 Completeness of ℂ

One might also wonder whether ℂ is complete. Our definition of completeness uses the notion of supremum,
which only makes sense if the field is ordered. This is not the case for ℂ as we have seen in Theorem 5.16.
Still, it is possible to give a different definition of completeness using the notion of Cauchy sequence. In
ordered fields, this new definition of completeness is equivalent to the definition which uses the supremum.
The new definition of completeness with Cauchy sequences also makes sense in non-ordered fields. We will
see that ℂ is a complete field, according to this new definition.

5.2 Complex conjugates

When computing inverses, we used the trick to multiply by 1:

1 1 𝑥 − 𝑖𝑦
𝑧 −1 = ⋅1= ⋅ .
𝑧 𝑥 + 𝑖𝑦 𝑥 − 𝑖𝑦

The complex number 𝑥 − 𝑖𝑦 is obtained by changing the sign to the imaginary part of 𝑧 = 𝑥 + 𝑖𝑦. We give a
name to this operation.

Definition 5.17: Complex conjugate

Let 𝑧 = 𝑥 + 𝑖𝑦. We call the complex conjugate of 𝑧, denoted by 𝑧,̄ the complex number

𝑧 ̄ = 𝑥 − 𝑖𝑦 .

Example 5.18

We have the following conjugates:

3 + 4𝑖 = 3 − 4𝑖 , 3 − 4𝑖 = 3 + 4𝑖 ,
−3 + 4𝑖 = −3 − 4𝑖 , −3 − 4𝑖 = −3 + 4𝑖 ,
3 = 3, 4𝑖 = −4𝑖 .

Complex conjugates have the following properties:

Theorem 5.19

For all 𝑧1 , 𝑧2 ∈ ℂ it holds:

• 𝑧1 + 𝑧 2 = 𝑧1 + 𝑧2

• 𝑧1 ⋅ 𝑧 2 = 𝑧1 ⋅ 𝑧2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 128

Proof
Let 𝑧1 , 𝑧2 ∈ ℂ. Then
𝑧1 = 𝑥1 + 𝑖𝑦1 , 𝑧2 = 𝑥2 + 𝑖𝑦2 ,
for some 𝑥1 , 𝑦1 , 𝑥2 , 𝑦2 ∈ ℝ.

• Using the definition of addition in ℂ and of conjugate,

𝑧1 + 𝑧2 = (𝑥1 + 𝑖𝑦1 ) + (𝑥2 + 𝑖𝑦2 )

= (𝑥1 + 𝑥2 ) + 𝑖 (𝑦1 + 𝑦2 )
= (𝑥1 + 𝑥2 ) − 𝑖 (𝑦1 + 𝑦2 )
= (𝑥1 − 𝑖𝑦1 ) + (𝑥2 − 𝑖𝑦2 )
= 𝑥1 + 𝑖𝑦1 + 𝑥2 + 𝑖𝑦2
= 𝑧1 + 𝑧2 .

• Using the definition of multiplication in ℂ and of conjugate,

𝑧1 ⋅ 𝑧2 = (𝑥1 + 𝑖𝑦1 ) ⋅ (𝑥2 + 𝑖𝑦2 )

= (𝑥1 𝑥2 − 𝑦1 𝑦2 ) + 𝑖 (𝑥1 𝑦2 + 𝑥2 𝑦1 )
= (𝑥1 𝑥2 − 𝑦1 𝑦2 ) − 𝑖 (𝑥1 𝑦2 + 𝑥2 𝑦1 )
= (𝑥1 − 𝑖𝑦1 ) ⋅ (𝑥2 − 𝑖𝑦2 )
= 𝑧1 ⋅ 𝑧2

Example 5.20

Let 𝑧1 = 3 − 4𝑖 and 𝑧2 = −2 + 5𝑖. Then

• Let us check that

𝑧1 + 𝑧2 = 𝑧1 + 𝑧2
Indeed, we have
𝑧1 + 𝑧 2 = 1 + 𝑖 ⟹ 𝑧1 + 𝑧 2 = 1 − 𝑖 .
On the other hand

𝑧1 = 3 + 4𝑖 , 𝑧2 = −2 − 5𝑖 ⟹ 𝑧1 + 𝑧2 = 1 − 𝑖 .

• Let us check that

𝑧1 ⋅ 𝑧 2 = 𝑧1 ⋅ 𝑧2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 129

Indeed,

𝑧1 ⋅ 𝑧2 = (3 + 4𝑖) ⋅ (−2 + 5𝑖)

= (−6 + 20) + (8 + 15)𝑖
= 14 + 23𝑖

so that
𝑧1 ⋅ 𝑧2 = 14 − 23𝑖
On the other hand:

𝑧1 ⋅ 𝑧2 = (3 + 4𝑖) ⋅ (−2 − 5𝑖)

= (−6 + 20) + (−15 − 8)𝑖
= 14 − 23𝑖

5.3 The complex plane

We can depict a real number 𝑥 as a point on the one-dimensional real line ℝ. The distance between two real
numbers 𝑥, 𝑦 ∈ ℝ on the real line is given by |𝑥 − 𝑦|, see Figure 5.1.

Figure 5.1: Two points 𝑥 and 𝑦 on the real line ℝ. Their distance is |𝑥 − 𝑦|.

We would like to do something similar for the complex numbers, but the point

𝑧 = 𝑥 + 𝑖𝑦 , 𝑥, 𝑦 ∈ ℝ .

We therefore depict 𝑧 = 𝑧 + 𝑖𝑦 in the two-dimensional plane at the point with (Cartesian) coordinates (𝑥, 𝑦).
This two-dimensional plane in which we can depict all complex numbers is called the complex plane. The
origin of such plane, with coordinates (0, 0), corresponds to the complex number

0 + 0𝑖 = 0 ,

see Figure 5.2.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 130

Figure 5.2: A point 𝑧 = 𝑥 + 𝑖𝑦 ∈ ℂ can be represented on the complex plane by the point of coordinates (𝑥, 𝑦).
The distance between 𝑧 and 0 is given by |𝑧| = √𝑧 2 + 𝑦 2 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 131

5.3.1 Distance on ℂ

The Cartesian representation allows us to introduce a distance between two complex numbers. Let us start
with the distance between a complex number 𝑧 = 𝑥 + 𝑖𝑦 and 0. By Pythagoras Theorem this distance is given
by
2 2
√𝑥 + 𝑦 ,
see Figure 5.2. We give a name to this quantity.

Definition 5.21: Modulus

The modulus of a complex number 𝑧 = 𝑥 + 𝑖𝑦 is defined by

|𝑧| ∶= √𝑥 2 + 𝑦 2 .

Note that the distance between 𝑧 and 0 is always a non-negative number.

Remark 5.22: Modulus of Real numbers

A real number 𝑥 ∈ ℝ can be written as

𝑥 = 𝑥 + 0𝑖 ∈ ℂ .
Hence the modulus of 𝑥 is given by
|𝑥| = √𝑥 2 + 02 = √𝑥 2 .
The above coincides with the absolute value of 𝑥. This explains why the notation for modulus in ℂ is the
same as the one for absolute value in ℝ.

We can now define the distance between two complex numbers.

Definition 5.23: Distance in ℂ

Given 𝑧1 , 𝑧2 ∈ ℂ, we define the distance between 𝑧1 and 𝑧2 as the quantity

|𝑧1 − 𝑧2 | .

The geometric intuition of why the quantity |𝑧1 − 𝑧2 | is defined as the distance between 𝑧1 and 𝑧2 is given in
Figure 5.3.

Theorem 5.24

Given 𝑧1 , 𝑧2 ∈ ℂ, we have
|𝑧1 − 𝑧2 | = √(𝑥1 − 𝑥2 )2 + (𝑦1 − 𝑦2 )2 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 132

Figure 5.3: The difference 𝑧1 − 𝑧2 of the two points 𝑧1 , 𝑧2 ∈ ℂ is given by the magenta vector. We define |𝑧1 − 𝑧2 |
as the distance between 𝑧1 and 𝑧2 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 133

Proof
We have
𝑧1 − 𝑧2 = (𝑥1 − 𝑥2 ) + 𝑖(𝑦1 − 𝑦2 ) .
Therefore, by definition of modulus,

|𝑧1 − 𝑧2 | = √(𝑥1 − 𝑥2 )2 + (𝑦1 − 𝑦2 )2 .

Example 5.25

The distance between

𝑧 = 2 − 4𝑖 , 𝑤 = −5 + 𝑖
is given by

|𝑧 − 𝑤| = |(2 − 4𝑖) − (−5 + 𝑖)|

= |7 − 5𝑖|
= √72 + (−5)2
= √74

5.3.2 Properties of modulus

The modulus has the following properties.

Theorem 5.26

Let 𝑧, 𝑧1 , 𝑧2 ∈ ℂ. Then

1. |𝑧1 ⋅ 𝑧2 | = |𝑧1 | |𝑧2 |

2. |𝑧 𝑛 | = |𝑧|𝑛 for all 𝑛 ∈ ℕ

3. 𝑧 ⋅ 𝑧 ̄ = |𝑧|2

Proof
Part 1. We have

𝑧1 ⋅ 𝑧2 = (𝑥1 + 𝑖𝑦1 ) ⋅ (𝑥2 + 𝑖𝑦2 )

= (𝑥1 𝑥2 − 𝑦1 𝑦2 ) + 𝑖(𝑥2 𝑦1 + 𝑥1 𝑦2 )

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 134

and therefore

|𝑧1 ⋅ 𝑧2 | = √(𝑥1 𝑥2 − 𝑦1 𝑦2 )2 + (𝑥2 𝑦1 + 𝑥1 𝑦2 )2

= √𝑥12 𝑥22 + 𝑦12 𝑦22 + 𝑥22 𝑦12 + 𝑥12 𝑦22 .

On the other hand,

|𝑧1 | = √𝑥12 + 𝑦12 , |𝑧2 | = √𝑥22 + 𝑦22
so that

|𝑧1 ||𝑧2 | = √𝑥12 + 𝑦12 √𝑥22 + 𝑦22

= √𝑥12 𝑥22 + 𝑦12 𝑦22 + 𝑥22 𝑦12 + 𝑥12 𝑦22

proving that |𝑧1 ⋅ 𝑧2 | = |𝑧1 | |𝑧2 |.

Part 2. Exercise. It easily follows from Point 1 and induction.
Part 3. Let 𝑧 = 𝑥 + 𝑖𝑦 for some 𝑥, 𝑦 ∈ ℝ. Then,

𝑧 ⋅ 𝑧 ̄ = (𝑥 + 𝑖𝑦)(𝑥 − 𝑖𝑦)
= 𝑥 2 − (𝑖𝑦)2
= 𝑥2 + 𝑦2
= |𝑧|2

The modulus in ℂ satisfies the triangle inequality.

Theorem 5.27: Triangle inequality in ℂ

For all 𝑥, 𝑦, 𝑧 ∈ ℂ,

1. |𝑥 + 𝑦| ≤ |𝑥| + |𝑦|

2. |𝑥 − 𝑧| ≤ |𝑥 − 𝑦| + |𝑦 − 𝑧|

Proof
Part 1. Suppose that 𝑥 = 𝑎 + 𝑖𝑏 and 𝑦 = 𝑐 + 𝑖𝑑 for 𝑎, 𝑏, 𝑐, 𝑑 ∈ ℝ. Then,

|𝑥 + 𝑦| = |(𝑎 + 𝑐) + 𝑖(𝑏 + 𝑑)| = √(𝑎 + 𝑐)2 + (𝑏 + 𝑑)2 .

Therefore the inequality

|𝑥 + 𝑦| ≤ |𝑥| + |𝑦| (5.2)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 135

is equivalent to
√(𝑎 + 𝑐)2 + (𝑏 + 𝑑)2 ≤ √𝑎2 + 𝑏 2 + √𝑐 2 + 𝑑 2 . (5.3)
Now note that, for 𝐴, 𝐵 ∈ ℝ, we have that

𝐴2 ≤ 𝐵 2 ⟹ |𝐴| ≤ |𝐵| . (5.4)

In the two reverse implications ⟸ below we will use (5.4):

√(𝑎 + 𝑐)2 + (𝑏 + 𝑑)2 ≤ √𝑎2 + 𝑏 2 + √𝑐 2 + 𝑑 2

2
⟸ (𝑎 + 𝑐)2 + (𝑏 + 𝑑)2 ≤ (√𝑎2 + 𝑏 2 + √𝑐 2 + 𝑑 2 )

⟺ 𝑎2 + 2𝑎𝑐 + 𝑐 2 + 𝑏 2 + 2𝑏𝑑 + 𝑑 2 ≤ 𝑎2 + 𝑏 2 + 2√𝑎2 + 𝑏 2 √𝑐 2 + 𝑑 2 + 𝑐 2 + 𝑑 2

⟺ 𝑎𝑐 + 𝑏𝑑 ≤ √𝑎2 + 𝑏 2 √𝑐 2 + 𝑑 2
⟸ (𝑎𝑐 + 𝑏𝑑)2 ≤ (𝑎2 + 𝑏 2 ) (𝑐 2 + 𝑑 2 )
⟺ 𝑎2 𝑐 2 + 2𝑎𝑏𝑐𝑑 + 𝑏 2 𝑑 2 ≤ 𝑎2 𝑐 2 + 𝑎2 𝑑 2 + 𝑏 2 𝑐 2 + 𝑏 2 𝑑 2
⟺ 𝑎2 𝑑 2 + 𝑏 2 𝑐 2 − 2𝑎𝑏𝑐𝑑 ≥ 0
⟺ (𝑎𝑑 − 𝑏𝑐)2 ≥ 0.

This last statement is clearly true, since 𝑎𝑑 − 𝑏𝑐 ∈ ℝ. Therefore (5.3) holds, and so (5.2) follows.
Part 2. Using (5.2) we estimate

|𝑥 − 𝑧| = |𝑥 − 𝑦 + 𝑦 − 𝑧| ≤ |𝑥 − 𝑦| + |𝑦 − 𝑧|.

Remark 5.28: Geometric interpretation of triangle inequality

We finally have a justification of why the inequality

|𝑥 − 𝑧| ≤ |𝑥 − 𝑦| + |𝑦 − 𝑧|

is called triangle inequality: By drawing three points 𝑥, 𝑦, 𝑧 ∈ ℂ in the complex plane, the distance
between 𝑥 and 𝑧 is shorter than the distance to go from 𝑥 to 𝑧 via the point 𝑦, see Figure 5.4.

5.4 Polar coordinates

We have seen that we can identify a complex number 𝑧 = 𝑥 +𝑖𝑦 by a point in the complex plane with Cartesian
coordinates (𝑥, 𝑦). We can also specify the point (𝑥, 𝑦) by using the so-called polar coordinates (𝜌, 𝜃), where
• 𝜌 is the distance between 𝑧 and the origin

𝜌 = |𝑧| = √𝑥 2 + 𝑦 2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 136

Figure 5.4: Let 𝑥, 𝑦, 𝑧 ∈ ℂ. The distance between 𝑥 and 𝑧 is shorter than the distance to go from 𝑥 to 𝑧 via the
point 𝑦.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 137

• 𝜃 is the angle between the line connecting the origin and 𝑧 and the positive real axis, see Figure 5.5.

Figure 5.5: Polar coordinates (𝜌, 𝜃) for the complex number 𝑧 ∈ ℂ.

We give such angle a name.

Definition 5.29: Argument

Let 𝑧 ∈ ℂ. The angle 𝜃 between the line connecting the origin and 𝑧 and the positive real axis is called
the argument of 𝑧, and is denoted by
𝜃 ∶= arg(𝑧) .

Warning

We always use angles in radians, not degrees. Make sure your calculator is set to radians if you want to
use it to compute angles.

Remark 5.30: Principal Value

The argument of a complex number is not uniquely defined. We can always add an integer number
of times 2𝜋 to the argument to specify the same point. We usually use the convention to choose the
argument in the interval (−𝜋, 𝜋]. This is called the principal value of the argument function. Therefore
the complex numbers in the upper half plane have a positive argument, and in the lower half plane have

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 138

a negative argument.

Example 5.31

We have the following arguments:

𝜋
arg(1) = 0 arg(𝑖) =
2
𝜋
arg(−1) = 𝜋 arg(−𝑖) = −
2
1 3
arg(1 + 𝑖) = 𝜋 arg(−1 − 𝑖) = − 𝜋
4 4

We can represent non-zero complex numbers in polar coordinates.

Theorem 5.32: Polar coordinates

Let 𝑧 ∈ ℂ with 𝑧 = 𝑥 + 𝑖𝑦 and 𝑧 ≠ 0. Then

𝑥 = 𝜌 cos(𝜃) , 𝑦 = 𝜌 sin(𝜃) ,

where
𝜌 = √𝑥 2 + 𝑦 2 , 𝜃 = arg(𝑧) .

The proof of Theorem 0.182 is trivial, and is based on basic trigonometry and definition of arg(𝑧). Complex
numbers in polar form can be useful. We give a name to such polar form.

Definition 5.33: Trigonometric form

Let 𝑧 ∈ ℂ. The trigonometric form of 𝑧 is

𝑧 = |𝑧| [cos(𝜃) + 𝑖 sin(𝜃)] ,

where 𝜃 = arg(𝑧).

Let us make an example.

Example 5.34

Suppose that we have polar coordinates

3
𝜌 = √8 , 𝜃= 𝜋
4

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 139

We compute

3 √8√2
𝑥 = 𝜌 cos(𝜃) = √8 cos ( 𝜋) = − = −2
4 2
3 √8√2
𝑦 = 𝜌 sin(𝜃) = √8 sin ( 𝜋) = = 2.
4 2
The complex number 𝑧 corresponding to the polar coordinates (𝜌, 𝜃) is

𝑧 = 𝑥 + 𝑖𝑦 = −2 + 2𝑖 .

The trigonometric form of 𝑧 is

3 3
𝑧 = √8 [cos ( 𝜋) + 𝑖 sin ( 𝜋)] .
4 4

As a consequence of Theorem 0.182 we obtain a formula for computing the argument.

Corollary 5.35: Computing arg(𝑧)

Let 𝑧 ∈ ℂ with 𝑧 = 𝑥 + 𝑖𝑦 and 𝑧 ≠ 0. Then

𝑦
⎧arctan ( 𝑥 ) if 𝑥 > 0
⎪ 𝑦
⎪arctan ( ) + 𝜋 if 𝑥 < 0 and 𝑦 ≥ 0
⎪ 𝑥𝑦
arg(𝑧) = arctan ( ) − 𝜋 if 𝑥 < 0 and 𝑦 < 0
⎨ 𝑥
⎪𝜋 if 𝑥 = 0 and 𝑦 > 0
⎪2
⎪ 𝜋
⎩− 2 if 𝑥 = 0 and 𝑦 < 0

where arctan is the inverse of tan.

Proof
Using the polar coordinates formulas from Theorem 0.182 we have

𝑦 𝜌 sin(𝜃)
= = tan(𝜃) .
𝑥 𝜌 cos(𝜃)

The thesis can be obtained by carefully inverting the tangent.

Example 5.36

We want to compute the arguments of the complex numbers

3 + 4𝑖 , 3 − 4𝑖 , −3 + 4𝑖 , −3 − 4𝑖 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 140

Using the formula for arg in Corollary 0.185 we have

4
arg(3 + 4𝑖) = arctan ( )
3
4 4
arg(3 − 4𝑖) = arctan (− ) = − arctan ( )
3 3
4 4
arg(−3 + 4𝑖) = arctan (− ) + 𝜋 = − arctan ( ) + 𝜋
3 3
4
arg(−3 − 4𝑖) = arctan ( ) − 𝜋
3

5.5 Exponential form

We have seen that we can represent complex numbers in

• Cartesian form
• Trigonometric form

We now introduce a third way of representing complex numbers: the exponential form. For this, we need
Euler’s identity:

Theorem 5.37: Euler’s identity

For all 𝜃 ∈ ℝ it holds

𝑒 𝑖𝜃 = cos(𝜃) + 𝑖 sin(𝜃) .

Proof
The proof of this theorem uses Taylor power series. Note that we have not introduced what series are,
yet, so we just assume that everything below makes sense and actually exists. We have the following
Taylor series at 𝑥0 = 0 that you might know from calculus:

𝑥2 𝑥3 𝑥4 𝑥5 𝑥6 𝑥7
𝑒𝑥 = 1 + 𝑥 + + + + + + +…
2! 3! 4! 5! 6! 7!
𝑥 𝑥3 𝑥5 𝑥7
sin(𝑥) = − + − +…
1! 3! 5! 7!
𝑥2 𝑥4 𝑥6
cos(𝑥) = 1 − + − +…
2! 4! 6!

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 141

The above identities also holds for 𝑥 ∈ ℂ. Hence we can substitute 𝑥 = 𝑖𝜃 in the series for 𝑒 𝑥 to obtain
(𝑖𝜃)2 (𝑖𝜃)3 (𝑖𝜃)4 (𝑖𝜃)5 (𝑖𝜃)6 (𝑖𝜃)7
𝑒 𝑖𝜃 = 1 + 𝑖𝜃 + + + + + + …
2! 3! 4! 5! 6! 7!
𝜃2 𝜃3 𝜃4 𝜃5 𝜃6 𝜃7
= 1 + 𝑖𝜃 − −𝑖 + +𝑖 − −𝑖 +…
2! 3! 4! 5! 6! 7!
= cos(𝜃) + 𝑖 sin(𝜃),

where we used that 𝑖2 = −1 in the second equality, and the third equality follows by observing that all
terms with an even power of 𝜃 are exactly the terms in the expansion of cos(𝜃) and all terms with an odd
power of 𝜃 are exactly the terms in the expansion of sin(𝜃) multiplied by 𝑖.

Theorem 5.38

For all 𝜃 ∈ ℝ it holds

|𝑒 𝑖𝜃 | = 1 .

Proof
From Euler’s identity in Theorem 0.187 we get

|𝑒 𝑖𝜃 | = | cos(𝜃) + 𝑖 sin(𝜃)| = √cos2 (𝜃) + sin2 (𝜃) = 1 .

Theorem 5.39

Let 𝑧 ∈ ℂ with 𝑧 = 𝑥 + 𝑖𝑦 and 𝑧 ≠ 0. Then

𝑧 = 𝜌𝑒 𝑖𝜃 ,
where
𝜌 = √𝑥 2 + 𝑦 2 = |𝑧| , 𝜃 = arg(𝑧) .

Proof
By Theorem 0.182 we have
𝑥 = 𝜌 cos(𝜃) , 𝑦 = 𝜌 sin(𝜃) .
Hence

𝑧 = 𝑥 + 𝑖𝑦
= 𝜌 cos(𝜃) + 𝑖𝜌 sin(𝜃)
= 𝜌𝑒 𝑖𝜃 ,

where in the last line we used Euler’s identity in Theorem 0.187.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 142

Definition 5.40: Exponential form

A complex number 𝑧 ∈ ℂ is in exponential form if

𝑧 = 𝜌𝑒 𝑖𝜃 = |𝑧| 𝑒 𝑖 arg(𝑧) .

Example 5.41

From Example 5.34 we know that

𝑧 = −2 + 2𝑖
can be written in trigonometric form as
3 3
𝑧 = √8 [cos ( 𝜋) + 𝑖 sin ( 𝜋)] .
4 4
By Euler’s identity we hence obtain the exponential form

𝑖 34 𝜋
𝑧 = √8𝑒 .

Remark 5.42: Periodicity of exponential

For all 𝑘 ∈ ℤ we have

𝑒 𝑖𝜃 = 𝑒 𝑖(𝜃+2𝜋𝑘) . (5.5)
As we did for the principal value of the argument, also for the exponential form we select 𝜃 ∈ (−𝜋, 𝜋]. In
particular equation (5.5) is saying that the complex exponential is 2𝜋-periodic.

Equation (5.5) follows immediately by Euler’s identity and periodicity of cos and sin, since

𝑒 𝑖(𝜃+2𝜋𝑘) = cos(𝜃 + 2𝜋𝑘) + 𝑖 sin(𝜃 + 2𝜋𝑘)

= cos(𝜃) + 𝑖 sin(𝜃) = 𝑒 𝑖𝜃 .

The exponential for is very useful for computing products and powers of complex numbers.

Proposition 5.43

Let 𝑧, 𝑧1 , 𝑧2 ∈ ℂ and suppose that

𝑧 = 𝜌𝑒 𝑖𝜃 , 𝑧1 = 𝜌1 𝑒 𝑖𝜃1 , 𝑧2 = 𝜌2 𝑒 𝑖𝜃2 .

We have
𝑧1 ⋅ 𝑧2 = 𝜌1 𝜌2 𝑒 𝑖(𝜃1 +𝜃2 ) , 𝑧 𝑛 = 𝜌 𝑛 𝑒 𝑖𝑛𝜃 ,
for all 𝑛 ∈ ℕ.

The proof follows immediately by the properties of the exponential. Let us see some applications of Propostion
5.43.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 143

Example 5.44

Suppose we want to compute (−2 + 2𝑖)4 . We could do this by means of the binomial theorem:

4 4 4
(−2 + 2𝑖)4 = (−2)4 + ( ) (−2)3 ⋅ 2𝑖 + ( ) (−2)2 ⋅ (2𝑖)2 + ( ) (−2) ⋅ (2𝑖)3 + (2𝑖)4
1 2 3
= 16 − 4 ⋅ 8 ⋅ 2𝑖 − 6 ⋅ 4 ⋅ 4 + 4 ⋅ 2 ⋅ 8𝑖 + 16
= 16 − 64𝑖 − 96 + 64𝑖 + 16 = −64 .

Using the exponential form simplifies this calculation. Indeed, we know that

𝑖 34 𝜋
−2 + 2𝑖 = √8𝑒

by Example 5.41. Hence

4
𝑖 34 𝜋
(−2 + 2𝑖)4 = (√8𝑒 ) = 82 𝑒 𝑖3𝜋 = −64 ,
where we used that
𝑒 𝑖3𝜋 = 𝑒 𝑖𝜋 = cos(𝜋) + 𝑖 sin(𝜋) = −1
by 2𝜋 periodicity of 𝑒 𝑧 and Euler’s identity.

Example 5.45

Suppose we want to compute

𝑖𝑖 .
It is not even clear how to do this calculation in Cartesian form. However, we know that
𝜋
|𝑖| = 1 , arg(𝑖) = .
2
Hence we can write 𝑖 in exponential form
𝑖𝜋
𝑖 = |𝑖|𝑒 𝑖 arg(𝑖) = 𝑒 2 .

Therefore
𝑖
𝑖𝜋 𝑖2 𝜋2 − 𝜋2
𝑖𝑖 = (𝑒 2 ) = 𝑒 =𝑒 .

5.6 Fundamental Theorem of Algebra

We started the introduction to complex numbers with the following question:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 144

Question 5.46

Is there a number 𝑥 ∈ ℝ such that

𝑥 2 = −1 ? (5.6)

The answer is no. For this reason we introduced the complex number 𝑖, which satisfies

𝑖2 = −1 .

Therefore (5.6) has solution in ℂ, with 𝑥 = 𝑖. We also have that

(−𝑖)2 = (−1)2 𝑖2 = −1 .

Hence (5.6) has two solutions in ℂ, given by

𝑥1 = 𝑖 , 𝑥2 = −𝑖 .

It turns out that the set ℂ is so large that we are not only able to solve (5.6), but in fact any polynomial
equation.

Theorem 5.47: Fundamental theorem of algebra

Let 𝑝𝑛 (𝑥) be a polynomial of degree 𝑛 with complex coefficients, i.e.,

𝑝𝑛 (𝑥) = 𝑎𝑛 𝑥 𝑛 + 𝑎𝑛−1 𝑥 𝑛−1 + … + 𝑎1 𝑥 + 𝑎0 ,

for some coefficients 𝑎𝑛 , … , 𝑎0 ∈ ℂ with 𝑎𝑛 ≠ 0. Then there exist 𝑧1 , … , 𝑧𝑛 ∈ ℂ such that

𝑝𝑛 (𝑥) = 𝑎𝑛 (𝑥 − 𝑧1 ) (𝑥 − 𝑧2 ) ⋯ (𝑥 − 𝑧𝑛 ) . (5.7)

Hence, the polynomial equation

𝑎𝑛 𝑥 𝑛 + 𝑎𝑛−1 𝑥 𝑛−1 + … + 𝑎1 𝑥 + 𝑎0 = 0 (5.8)

has solutions 𝑧1 , … , 𝑧𝑛 .

Theorem 0.197 says that every polynomial of degree 𝑛 has 𝑛 zeros, sometimes also called roots, i.e., 𝑛 solutions
to (5.8). We call the expression (5.7) a factorization of the polynomial 𝑝𝑛 .
Several proofs of Theorem 0.197 exist in the literature, but they all use mathematical tools which are out of
reach for now. Therefore we will not show a proof. For example one can prove Theorem 0.197 by

• Liouville’s theorem (complex analysis)

• Homotopy arguments (general topology)
• Fundamental Theorem of Galois Theory (algebra)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 145

Example 5.48

The equation
𝑥 2 = −1 (5.9)
is equivalent to
𝑝(𝑥) = 0 , 𝑝(𝑥) ∶= 𝑥 2 + 1 .
Since 𝑝 has degree 𝑛 = 2, the Fundamental Theorem of Algebra tells us that there are two solutions to
(5.9). We have already seen that these two solutions are 𝑥 = 𝑖 and 𝑥 = −𝑖. Then

𝑝(𝑥) = 𝑥 2 + 1 = (𝑥 − 𝑖)(𝑥 + 𝑖) .

Example 5.49

Suppose we want to solve

𝑥4 − 1 = 0 . (5.10)
The associated polynomial equation is

𝑝(𝑥) = 0 , 𝑝(𝑥) ∶= 𝑥 4 − 1 .

Since 𝑝 has degree 𝑛 = 4, the Fundamental Theorem of Algebra tells us that there are 4 solutions to (5.10).
Let us find such solutions. We use the well known formula

𝑎2 − 𝑏 2 = (𝑎 + 𝑏)(𝑎 − 𝑏) , ∀ 𝑎, 𝑏 ∈ ℝ ,

to factorize 𝑝. We get:
𝑝(𝑥) = (𝑥 4 − 1) = (𝑥 2 + 1)(𝑥 2 − 1) .
We know that
𝑥2 + 1 = 0
has solutions 𝑥 = ±𝑖. Instead

𝑥2 − 1 = 0
has solutions 𝑥 = ±1. Hence, the four solutions of (5.10) are given by 𝑥 = 1, −1, 𝑖, −𝑖 and

𝑝(𝑥) = 𝑥 4 − 1 = (𝑥 − 1)(𝑥 + 1)(𝑥 − 𝑖)(𝑥 + 𝑖) .

Definition 5.50

Suppose that the polynomial 𝑝𝑛 factorizes as

𝑝𝑛 (𝑥) = 𝑎𝑛 (𝑥 − 𝑧1 )𝑘1 (𝑥 − 𝑧2 )𝑘2 ⋯ (𝑥 − 𝑧𝑚 )𝑘𝑚

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 146

with 𝑎𝑛 ≠ 0, 𝑧1 , … , 𝑧𝑚 ∈ ℂ and 𝑘1 , … , 𝑘𝑚 ∈ ℕ, 𝑘𝑖 ≥ 1. In this case 𝑝𝑛 has degree

𝑚
𝑛 = 𝑘 1 + … + 𝑘 𝑚 = ∑ 𝑘𝑖 .
𝑖=1

We have that 𝑧𝑖 is solves the equation

𝑝𝑛 (𝑥) = 0
exactly 𝑘𝑖 times. We call 𝑘𝑖 the multiplicity of the solution 𝑧𝑖 .

Example 5.51

The equation
(𝑥 − 1)(𝑥 − 2)2 (𝑥 + 𝑖)3 = 0
has 6 solutions:

• 𝑥 = 1 with multiplicity 1
• 𝑥 = 2 with multiplicity 2
• 𝑥 = −𝑖 with multiplicity 3

5.7 Solving polynomial equations

The non-factorized version of the polynomial
𝑝(𝑥) = (𝑥 − 1)(𝑥 − 2)2 (𝑥 + 𝑖)3
from Example 5.51 is
𝑝(𝑥) =𝑥 6 − (5 − 3𝑖)𝑥 5 + (5 − 15𝑖)𝑥 4
+ (11 + 23𝑖)𝑥 3 − (24 + 7𝑖)𝑥 2 + (12 − 8𝑖)𝑥 + 4𝑖
We therefore have the following natural question.

Question 5.52

The Fundamental Theorem of Algebra states that

𝑝𝑛 (𝑥) = 0 (5.11)

has 𝑛 complex solutions. How do we find such solutions?

The answer is that there is no general way to solve (5.11) when 𝑛 ≥ 5. This is the content of the Abel-Ruffini
Theorem.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 147

Theorem 5.53: Abel-Ruffini

There is no elementary solution formula to the polynomial equation

𝑝𝑛 (𝑥) = 0

with 𝑝𝑛 polynomial of degree 𝑛, with 𝑛 ≥ 5.

Similarly to the Fundamental Theorem of Algebra, the proof of the Abel-Ruffini Theorem is out of reach for
our current mathematical knowledge. A proof can be carried out, for example, using Galois Theory.
There are however explicit formulas for solving (5.11) when 𝑝𝑛 has degree 𝑛 = 2, 3, 4. For 𝑛 = 2 we can use
the well-known quadratic formula.

Proposition 5.54: Quadratic formula

Let 𝑎, 𝑏, 𝑐 ∈ ℝ, 𝑎 ≠ 0 and consider the equation

𝑎𝑥 2 + 𝑏𝑥 + 𝑐 = 0 . (5.12)

Define
Δ ∶= 𝑏 2 − 4𝑎𝑐 .
The following hold:

• If Δ > 0 then (5.12) has two distinct real solutions given by

−𝑏 − √Δ −𝑏 + √Δ
𝑥1 = , 𝑥2 = .
2𝑎 2𝑎
• If Δ = 0 then (5.12) has one solution with multiplicity 2. Such solution is given by

−𝑏
𝑥1 = .
2𝑎
• If Δ < 0 then (5.12) has two distinct complex solutions given by

−𝑏 − 𝑖√−Δ −𝑏 + 𝑖√−Δ
𝑥1 = , 𝑥2 = ,
2𝑎 2𝑎
where √−Δ is a real number, since −Δ > 0.

Moreover, if Δ ≠ 0 we have
𝑎𝑥 2 + 𝑏𝑥 + 𝑐 = 𝑎(𝑥 − 𝑥1 )(𝑥 − 𝑥2 ) ,
while if Δ = 0 then
𝑎𝑥 2 + 𝑏𝑥 + 𝑐 = 𝑎(𝑥 − 𝑥1 )2 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 148

Example 5.55

Suppose we want to solve

3𝑥 2 − 6𝑥 + 2 = 0 .
We have that
Δ = (−6)2 − 4 ⋅ 3 ⋅ 2 = 12 > 0
Therefore the equation has two distinct real solutions, given by

−(−6) ± √12 6 ± √12 √3

𝑥= = =1±
2⋅3 6 3
In particular we have the factorization

√3 √3
3𝑥 2 − 6𝑥 + 2 = 3 (𝑥 − 1 − ) (𝑥 − 1 + ) .
3 3

Example 5.56

Suppose we want to solve

4𝑥 2 − 8𝑥 + 4 = 0 .
We have that
Δ = (−8)2 − 4 ⋅ 4 ⋅ 4 = 0 .
Therefore there exists one solution with multiplicity 2. This is given by

−(−8)
𝑥= = 1.
2⋅4
In particular we have the factorization

4𝑥 2 − 8𝑥 + 4 = 4(𝑥 − 1)2 .

Example 5.57

Consider
𝑥 2 + 2𝑥 + 3 = 0 .
We have
Δ = 22 − 4 ⋅ 1 ⋅ 3 = −8 < 0 .
Therefore there are two complex solutions given by
−2 ± 𝑖√8
𝑥= = −1 ± 𝑖√2 .
2⋅1
In particular we have the factorization
𝑥 2 + 2𝑥 + 3 = (𝑥 + 1 − 𝑖√2)(𝑥 + 1 + 𝑖√2) .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 149

So far we have considered the polynomial equation

𝑎𝑥 2 + 𝑏𝑥 + 𝑐 = 0 ,

for 𝑎, 𝑏, 𝑐 ∈ ℝ and 𝑎 ≠ 0.

Question 5.58

What if 𝑎, 𝑏, 𝑐 ∈ ℂ?

If 𝑎, 𝑏, 𝑐 ∈ ℂ then we might have

Δ ∶= 𝑏 2 − 4𝑎𝑐 ∈ ℂ .
Therefore it is not clear how to compute
√Δ .
However, we can still use the quadratic equation.

Proposition 5.59: Generalization of quadratci formula

Let 𝑎, 𝑏, 𝑐 ∈ ℂ, 𝑎 ≠ 0. The two solutions to

𝑎𝑥 2 + 𝑏𝑥 + 𝑐 = 0

are given by
−𝑏 + 𝑆1 −𝑏 + 𝑆2
𝑥1 = , 𝑥2 = ,
2𝑎 2𝑎
where 𝑆1 and 𝑆2 are the two solutions to

𝑧2 = Δ , Δ ∶= 𝑏 2 − 4𝑎𝑐 .

Remark 5.60

Suppose that Δ ∈ ℝ. The equation

𝑧2 = Δ
has the following solutions:

• If Δ > 0 there are two real solutions

𝑆1 = −√Δ , 𝑆2 = √Δ

• If Δ = 0 then 0 is the only solution with multiplicity 2. Hence

𝑆1 = 𝑆 2 = 0

• If Δ < 0 there are two complex solutions

𝑆1 = −𝑖√−Δ , 𝑆2 = 𝑖√−Δ

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 150

Therefore the solutions

−𝑏 + 𝑆1 −𝑏 + 𝑆2
𝑥1 = , 𝑥2 = ,
2𝑎 2𝑎
given in Proposition 0.204 coincide with the ones given in Proposition 0.209.

Example 5.61

Let us see an application of Proposition 0.209. Consider the equation

1 2
𝑥 − (3 + 𝑖)𝑥 + (4 − 𝑖) = 0 . (5.13)
2
We have
1
Δ = (−(3 + 𝑖))2 − 4 ⋅ ⋅ (4 − 𝑖)
2
= 8 + 6𝑖 − 8 + 2𝑖
= 8𝑖 .
Therefore Δ ∈ ℂ. We have to find solutions 𝑆1 and 𝑆2 to the equation
𝑧 2 = Δ = 8𝑖 . (5.14)
We look for solutions of the form 𝑧 = 𝑥 + 𝑖𝑦. Then we must have that
𝑧 2 = (𝑎 + 𝑖𝑏)2 = 𝑎2 − 𝑏 2 + 2𝑎𝑏𝑖 = 8𝑖 .
Thus
𝑎2 − 𝑏 2 = 0 , 2𝑎𝑏 = 8 .
From the first equation we conclude that |𝑎| = |𝑏|. From the second equation we have that 𝑎𝑏 = 4, and
therefore 𝑎 and 𝑏 must have the same sign. Hence 𝑎 = 𝑏, and
2𝑎𝑏 = 8 ⟹ 𝑎 = 𝑏 = ±2 .
From this we conclude that the solutions to (5.14) are
𝑆1 = 2 + 2𝑖 , 𝑆2 = −2 − 2𝑖 .
Hence the solutions to (5.13) are
3 + 𝑖 + 𝑆1
𝑥1 = 1
= 3 + 𝑖 + 𝑆1
2⋅ 2
= 3 + 𝑖 + 2 + 2𝑖 = 5 + 3𝑖 ,
and
3 + 𝑖 + 𝑆2
𝑥2 = 1
= 3 + 𝑖 + 𝑆2
2⋅ 2
= 3 + 𝑖 − 2 − 2𝑖 = 1 − 𝑖 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 151

In the above example it was a bit laborious to compute 𝑆1 and 𝑆2 . In the next section we will see an easier
way to solve problems of the form 𝑧 2 = Δ.

Remark 5.62: Polynomial equations of order 𝑛 = 3, 4

For polynomial equations

𝑝𝑛 (𝑥) = 0
with 𝑝𝑛 of degree 𝑛 = 3, 4 similar methods exist. However the solution formulas for such equations are
really complicated, and we do not cover them here.
Still, it is sometimes possible to solve equations of degree higher than 2, in case it is obvious from inspec-
tion that a certain number is a solution, e.g., when 𝑥 = −1, 0, 1 is a solution.

Example 5.63

Consider the equation

𝑥 3 − 7𝑥 2 + 6𝑥 = 0 .
It is clear that 𝑥 = 0 is a solution and that we can write

𝑥 3 − 7𝑥 2 + 6𝑥 = 𝑥 (𝑥 2 − 7𝑥 + 6) .

We could now use the quadratic formula to find the remaining two roots, but we can also directly observe
that also 𝑥 = 1 is a solution, so that 𝑥 − 1 divides 𝑥 2 − 7𝑥 + 6. Using polynomial long division, we find
that
𝑥 2 − 7𝑥 + 6
= 𝑥 − 6,
𝑥 −1
see Figure 5.6. Therefore the last solution is 𝑥 = 6, and

𝑥 3 − 7𝑥 2 + 6𝑥 = 𝑥(𝑥 − 1)(𝑥 − 6) .

Figure 5.7: Example of polynomial long division between 6𝑥 3 + 5𝑥 2 − 7 and 3𝑥 2 − 2𝑥 − 1.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 152

Figure 5.6: Polynomial long division between 𝑥 2 − 7𝑥 + 6 and 𝑥 − 1.

Example 5.64

Suppose we want to solve

𝑥 3 − 7𝑥 + 6 = 0 .
It is easy to see 𝑥 = 1 is a solution. This means that 𝑥 − 1 divides 𝑥 3 − 7𝑥 + 6, so we can compute by using
polynomial long division,

𝑥 3 − 7𝑥 + 6
= 𝑥2 + 𝑥 − 6 ,
𝑥 −1
see Figure 5.8. For the remaining two solutions, we can use the quadratic formula to obtain that also
𝑥 = 2 and 𝑥 = −3 are solutions. Thus

𝑥 3 − 7𝑥 + 6 = (𝑥 − 1)(𝑥 − 2)(𝑥 + 3) .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 153

Figure 5.8: Polynomial long division between 𝑥 3 − 7𝑥 + 6 and 𝑥 − 1.

5.8 Roots of unity

Problem
Let 𝑛 ∈ ℕ. We want to find all complex solutions to

𝑧𝑛 = 1 . (5.15)

Note that 𝑧 = 1 is always a solution to (5.15) if 𝑛 is even. In such case also 𝑧 = −1 is a solution. If we were
only looking for solutions in ℝ, these two would be the only solutions.
However, the Fundamental Theorem of Algebra, see Theorem 0.197, tells us that there are 𝑛 complex solutions
to (5.15).

Question 5.65

Is there a way to find all 𝑛 solutions?

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 154

Example 5.66

We have seen in Example 5.49 that the solutions to

𝑥4 = 1

are 𝑥 = −1, 1, 𝑖, −𝑖. However we deduced this with a procedure which does not seem to generalize well
to other exponents.

The trick to find all 𝑛 solutions to (5.15) is to use the exponential form.

Theorem 5.67

Let 𝑛 ∈ ℕ and consider the equation

𝑧𝑛 = 1 . (5.16)
All the 𝑛 solutions to (5.16) are given by

2𝜋𝑘
𝑧𝑘 = exp (𝑖 ), 𝑘 = 0, … , 𝑛 − 1 ,
𝑛
where exp(𝑥) denotes 𝑒 𝑥 .

Proof
Rewrite 1 in exponential form:
1 = |1|𝑒 𝑖 arg(1) = 𝑒 𝑖2𝜋𝑘 , 𝑘 ∈ ℤ.
Therefore (5.16) is equivalent to
𝑧 𝑛 = 𝑒 𝑖2𝜋𝑘 .
By the properties of the exponential, we see that the above is solved by

2𝜋𝑘
𝑧𝑘 = exp (𝑖 ), 𝑘 ∈ ℤ.
𝑛
By choosing 𝑘 = 0, … , 𝑛 − 1 we obtain 𝑛 different solutions.

Definition 5.68

The solutions to
𝑧𝑛 = 1
are called the roots of unity.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 155

Example 5.69

The solutions to
𝑧4 = 1
are given by
2𝜋𝑘 𝜋𝑘
𝑧𝑘 = exp (𝑖 ) = exp (𝑖 ) .
4 2
By taking 𝑘 = 0, 1, 2, 3, we obtain the four solutions
𝑖 𝜋2
𝑧0 = 𝑒 𝑖0 = 1 , 𝑧1 = 𝑒 = 𝑖,
𝑖 3𝜋2
𝑧2 = 𝑒 𝑖𝜋 = −1 , 𝑧3 = 𝑒 = −𝑖 .

Note that for 𝑘 = 4 we would again get the solution 𝑧 = 𝑒 𝑖2𝜋 = 1.

Example 5.70

The solutions to
𝑧3 = 1
are given by
2𝜋𝑘
𝑧𝑘 = exp (𝑖 ).
3
By taking 𝑘 = 0, 1, 2, we obtain the three solutions

𝑖 2𝜋3 𝑖 4𝜋3
𝑧0 = 𝑒 𝑖0 = 1, 𝑧1 = 𝑒 , 𝑧2 = 𝑒 .

We can also convert 𝑧1 and 𝑧2 to trigonometric form:

𝑖 2𝜋3 2𝜋 2𝜋 1 √3
𝑧1 = 𝑒 = cos ( ) + 𝑖 sin ( ) = − + 𝑖
3 3 2 2
and
𝑖 4𝜋3 4𝜋 4𝜋 1 √3
𝑧2 = 𝑒 = cos ( ) + 𝑖 sin ( ) = − − 𝑖
3 3 2 2

5.9 Roots in ℂ

Problem
Let 𝑛 ∈ ℕ and 𝑐 ∈ ℂ. We want to find the 𝑛-th roots of 𝑐. This means we want to find all complex solutions
to
𝑧𝑛 = 𝑐 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 156

The Fundamental Theorem of Algebra ensures that the above has 𝑛 complex solutions. To find these solutions,
we pass to the exponential form.

Theorem 5.71

Let 𝑛 ∈ ℕ, 𝑐 ∈ ℂ and consider the equation

𝑧𝑛 = 𝑐 . (5.17)
All the 𝑛 solutions to (5.17) are given by

𝑛 𝜃 + 2𝜋𝑘
𝑧𝑘 = √ |𝑐| exp (𝑖 ), 𝑘 = 0, … , 𝑛 − 1 ,
𝑛

where √
𝑛
|𝑐| is the 𝑛-th root of the real number |𝑐|, and 𝜃 = arg(𝑐).

Proof
Write 𝑐 in exponential form:
𝑐 = |𝑐|𝑒 𝑖𝜃 = |𝑐|𝑒 𝑖(𝜃+2𝜋𝑘) , 𝑘 ∈ ℤ,
where 𝜃 = arg(𝑐). Therefore (5.17) is equivalent to

𝑧 𝑛 = |𝑐|𝑒 𝑖(𝜃+2𝜋𝑘) .

By the properties of the exponential, we see that the above is solved by

𝑛 𝜃 + 2𝜋𝑘
𝑧𝑘 = √ |𝑐| exp (𝑖 ), 𝑘 ∈ ℤ.
𝑛
By choosing 𝑘 = 0, … , 𝑛 − 1 we obtain 𝑛 different solutions.

Example 5.72

We want to find all the 𝑧 ∈ ℂ such that

𝑧 5 = −32 .
Let 𝑐 = −32. We have
|𝑐| = | − 32| = 32 = 25 , 𝜃 = arg(−32) = 𝜋 .
Hence, the solutions are given by
1
1 + 2𝑘
𝑧𝑘 = (25 ) 5 exp (𝑖𝜋 ), 𝑘 ∈ ℤ.
5
By taking 𝑘 = 0, 1, 2, 3, 4 we get the solutions
𝑖 𝜋5 𝑖 3𝜋5
𝑧0 = 2𝑒 𝑧1 = 2𝑒
𝑖 7𝜋5
𝑧2 = 2𝑒 𝑖𝜋 = −2 𝑧3 = 2𝑒
𝑖 9𝜋5
𝑧4 = 2𝑒

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 157

Example 5.73

We want to find all the 𝑧 ∈ ℂ such that

𝜋 𝜋
𝑧 4 = 9 (cos ( ) + 𝑖 sin ( )) .
3 3
Set
𝜋 𝜋
𝑐 ∶= 9 (cos ( ) + 𝑖 sin ( )) .
3 3
The complex number 𝑐 is already in the trigonometric form, so that we can immediately obtain
𝜋
|𝑐| = 9 , 𝜃 = arg(𝑐) = .
3
Hence the solutions are given by

4 𝜋/3 + 2𝜋𝑘
𝑧𝑘 = √ 9 exp (𝑖 )
4
1 + 6𝑘
= √3 exp (𝑖𝜋 )
12
for 𝑘 ∈ ℤ. Choosing 𝑘 = 0, 1, 2, 3 gives the 4 solutions
1 7
𝑖𝜋 12 𝑖𝜋 12
𝑧0 = √3𝑒 𝑧1 = √3𝑒
13
𝑖𝜋 12 𝑖𝜋 19
𝑧2 = √3𝑒 𝑧3 = √3𝑒 12

Dr. Silvio Fanzon [email protected]

6 Sequences in ℝ
A sequence is an infinite list of real numbers. For example, the following are sequences:

• (1, 2, 3, 4, …)
• (−1, 1, −1, 1, …)
• (1, 12 , 31 , 41 , 15 , …)

Remark 6.1
• The order of elements in a sequence matters.

For example
(1, 2, 3, 4, 5, 6, …) ≠ (2, 1, 4, 3, 6, 5, …)

• A sequence is not a set.

For example
{−1, 1, −1, 1, −1, 1, …} = {−1, 1}
but we cannot make a similar statement for the sequence

(−1, 1, −1, 1, −1, 1, …) .

• The above notation is ambiguous.

For example the sequence

(1, 2, 3, 4, …)
can continue as

(1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 4, …) .

• In the sequence
1 1 1 1
(1, , , , , …)
2 3 4 5
the elements get smaller and smaller, and closer and closer to 0. We say that this sequence con-
verges to 0, or has 0 as a limit.

We would like to make the notions of sequence and convergence more precise.

158
Numbers, Sequences and Series Page 159

6.1 Definition of sequence

We start with the definition of sequence of Real numbers.

Definition 6.2: Sequence of Real numbers

A sequence 𝑎 in ℝ is a function
𝑎∶ ℕ → ℝ.
For 𝑛 ∈ ℕ, we denote the 𝑛-th element of the sequence 𝑎 by

𝑎𝑛 = 𝑎(𝑛)

and write the sequence as

(𝑎𝑛 )𝑛∈ℕ .

Notation 6.3

We will sometimes omit the subscript 𝑛 ∈ ℕ and simply write

(𝑎𝑛 ) .

In certain situations, we will also write

∞
(𝑎𝑛 )𝑛=1 .

Example 6.4

• In general (𝑎𝑛 )𝑛∈ℕ is the sequence

(𝑎1 , 𝑎2 , 𝑎3 , …) .
• Consider the function
𝑎∶ ℕ → ℕ, 𝑛 ↦ 2𝑛 .
This is also a sequence of real numbers. It can be written as
(2𝑛)𝑛∈ℕ
and it represents the sequence of even numbers
(2, 4, 6, 8, 10, …) .
• Let
𝑎𝑛 = (−1)𝑛
Then (𝑎𝑛 ) is the sequence
(−1, 1, −1, 1, −1, 1, …) .
• ( 𝑛1 ) is the sequence
𝑛∈ℕ
1 1 1
(1, , , , …) .
2 4 5

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 160

6.2 Convergent sequences

We have notice that the sequence ( 𝑛1 ) gets close to 0 as 𝑛 gets large. We would like to say that 𝑎𝑛 converges
𝑛∈ℕ
to 0 as 𝑛 tends to infinity.
To make this precise, we first have to say what it means for two numbers to be close. For this we use the
notion of absolute value, and say that:

• 𝑥 and 𝑦 are close if |𝑥 − 𝑦| is small.

• |𝑥 − 𝑦| is called the distance between 𝑥 and 𝑦
• For 𝑥 to be close to 0, we need that |𝑥 − 0| = |𝑥| is small.

Saying that |𝑥| is small is not very precise. Let us now give the formal definition of convergent sequence.

Definition 6.5: Convergent sequence

We say that a sequence (𝑎𝑛 )𝑛∈ℕ in ℝ converges to 𝑎 ∈ ℝ, or equivalently has limit 𝑎, denoted by

lim 𝑎𝑛 = 𝑎
𝑛→∞

if for all 𝜀 ∈ ℝ, 𝜀 > 0, there exists 𝑁 ∈ ℕ such that for all 𝑛 ∈ ℕ, 𝑛 ≥ 𝑁 it holds that

|𝑎𝑛 − 𝑎| < 𝜀 .

Using quantifiers, we can write this as

∀ 𝜀 > 0, ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 − 𝑎| < 𝜀 .

If there exists 𝑎 ∈ ℝ such that lim𝑛→∞ 𝑎𝑛 = 𝑎, then we say that the sequence (𝑎𝑛 )𝑛∈ℕ is convergent.

Notation 6.6
We will often write
𝑎𝑛 → 𝑎
in place of
lim 𝑎𝑛 = 𝑎 .
𝑛→∞

Remark 6.7

• Informally, Definition 0.228 says that, no matter how small we choose 𝜀 (as long as it is strictly
positive), we always have that 𝑎𝑛 has a distance to 𝑎 of less than or equal to 𝜀 from a certain point
onwards (i.e., from 𝑁 onward). The sequence (𝑎𝑛 ) may fluctuate wildly in the beginning, but from
𝑁 onward it should stay within a distance of 𝜀 of 𝑎.

• In general 𝑁 depends on 𝜀. If 𝜀 is chosen smaller, we might have to take 𝑁 larger: this means we

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 161

need to wait longer before the sequence stays within a distance 𝜀 from 𝑎.

We now prove that the sequence

1
𝑎𝑛 =
𝑛
converges to 0, according to Definition 0.228.

Theorem 6.8

The sequence ( 𝑛1 ) converges to 0 , i.e.,

𝑛∈ℕ
1
lim = 0.
𝑛→∞ 𝑛

We give two proofs of the above theorem:

• Long proof, with all the details.
• Short proof, with less details, but still acceptable.

Proof: Proof of Theorem 6.8 (Long version)

We have to show that

1
lim
= 0,
𝑛→∞ 𝑛
which by definition is equivalent to showing that
1
∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , | − 0| < 𝜀 . (6.1)
𝑛
Let 𝜀 ∈ ℝ with 𝜀 > 0. Choose 𝑁 ∈ ℕ such that
1
𝑁 > .
𝜀
Such natural number 𝑁 exists thanks to the Archimedean property. The above implies
1
<𝜀, (6.2)
𝑁
Let 𝑛 ∈ ℕ with 𝑛 ≥ 𝑁 . By (6.2) we have
1 1
≤ <𝜀.
𝑛 𝑁
From this we deduce
1 1 1
| − 0| = ≤ <𝜀
𝑛 𝑛 𝑁
Since 𝑛 ∈ ℕ, 𝑛 ≥ 𝑁 was arbitrary, we have proven that
1
| − 0| < 𝜀 , ∀𝑛 ≥ 𝑁 . (6.3)
𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 162

Condition (6.3) holds for all 𝜀 > 0, for the choice of 𝑁 ∈ ℕ such that
1
𝑁 > .
𝜀
We have hence shown (6.1), and the proof is concluded.

As the above proof is quite long and includes lots of details, it is acceptable to shorten it. For example:

• We skip some intermediate steps.

• We do not mention the Archimedean property.
• We leave out the conclusion when it is obvious that the statement has been proven.

Proof: Proof of Theorem 6.8 (Short version)

We have to show that

1
∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , | − 0| < 𝜀 .
𝑛
Let 𝜀 > 0. Choose 𝑁 ∈ ℕ such that
1
𝑁 > .
𝜀
Let 𝑛 ≥ 𝑁 . Then
1 1 1
| − 0| = ≤ <𝜀.
𝑛 𝑛 𝑁

In Theorem 6.8 we showed that

1
lim = 0.
𝑛→∞ 𝑛
We can generalise this statement to prove that
1
lim =0
𝑛→∞ 𝑛𝑝
for any 𝑝 > 0 fixed.

Theorem 6.9

For all 𝑝 > 0, the sequence ( 𝑛1𝑝 ) converges to 0 , i.e.,

𝑛∈ℕ
1
lim =0
𝑛→∞ 𝑛𝑝

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 163

Proof
Let 𝑝 > 0. We have to show that
1
∀𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , | − 0| < 𝜀 .
𝑛𝑝
Let 𝜀 > 0. Choose 𝑁 ∈ ℕ such that
1
𝑁 > . (6.4)
𝜀 1/𝑝
Let 𝑛 ≥ 𝑁 . Since 𝑝 > 0, we have 𝑛𝑝 ≥ 𝑁 𝑝 , which implies
1 1
𝑝 ≤ 𝑝.
𝑛 𝑁
By (6.4) we deduce
1
<𝜀.
𝑁𝑝
Then
1 1 1
| 𝑝 − 0| = 𝑝 ≤ 𝑝 < 𝜀 .
𝑛 𝑛 𝑁

Question 6.10

Why did we choose 𝑁 ∈ ℕ such that

1
𝑁 >
𝜀𝑝
in the above proof?

The answer is: because it works. Finding a number 𝑁 that makes the proof work requires some rough work:
Specifically, such rough work consists in finding 𝑁 ∈ ℕ such that the inequality

|𝑎𝑁 − 𝑎| < 𝜀

is satisfied.

Important

Any rough work required to prove convergence must be shown before the formal proof (in assignments).

Example 6.11

Prove that, as 𝑛 → ∞,
𝑛 1
→ .
2𝑛 + 3 2
Part 1. Rough Work.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 164

Let 𝜀 > 0. We want to find 𝑁 ∈ ℕ such that

𝑛 1
| − | < 𝜀 , ∀𝑛 ≥ 𝑁 .
2𝑛 + 3 2
To this end, we compute:
𝑛 1 2𝑛 − (2𝑛 + 3)
| − |=| |
2𝑛 + 3 2 2(2𝑛 + 3)
−3
=| |
4𝑛 + 6
3
= .
4𝑛 + 6
Therefore
𝑛 1 3
| − |<𝜀 ⟺ <𝜀
2𝑛 + 3 2 4𝑛 + 6
4𝑛 + 6 1
⟺ >
3 𝜀
3
⟺ 4𝑛 + 6 >
𝜀
3
⟺ 4𝑛 > − 6
𝜀
3 6
⟺ 𝑛> − .
4𝜀 4
Looking at the above equivalences, it is clear that 𝑁 ∈ ℕ has to be chosen so that
3 6
𝑁 > − .
4𝜀 4
Part 2. Formal Proof. We have to show that
𝑛 1
∀𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , | − |<𝜀.
2𝑛 + 3 2
Let 𝜀 > 0. Choose 𝑁 ∈ ℕ such that
3 6
𝑁 > − . (6.5)
4𝜀 4
By the rough work shown above, inequality (6.5) is equivalent to
3
<𝜀.
4𝑁 + 6
Let 𝑛 ≥ 𝑁 . Then
𝑛 1 −3
| − |=| |
2𝑛 + 3 2 2(2𝑛 + 3)
3
=
4𝑛 + 6
3
≤
4𝑁 + 6
<𝜀,
where in the third line we used that 𝑛 ≥ 𝑁 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 165

We conclude by showing that constant sequences always converge.

Theorem 6.12
Let 𝑐 ∈ ℝ and define the constant sequence

𝑎𝑛 ∶= 𝑐 , ∀𝑛 ∈ ℕ.

We have that
lim 𝑎𝑛 = 𝑐 .
𝑛→∞

Proof
We have to prove that
∀𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 − 𝑐| < 𝜀 . (6.6)
Let 𝜀 > 0. We have
|𝑎𝑛 − 𝑐| = |𝑐 − 𝑐| = 0 < 𝜀 , ∀𝑛 ∈ ℕ.
Therefore we can choose 𝑁 = 1 and (6.6) is satisfied.

6.3 Divergent sequences

The opposite of convergent sequences are divergent sequences.

Definition 6.13: Divergent sequence

We say that a sequence (𝑎𝑛 )𝑛∈ℕ in ℝ is divergent if it is not convergent.

Remark 6.14

Proving that a sequence (𝑎𝑛 ) is divergent is more complicated than showing it is convergent: To show
that (𝑎𝑛 ) is divergent, we need to show that (𝑎𝑛 ) cannot converge to 𝑎 for any 𝑎 ∈ ℝ.
In other words, we have to show that there does not exist an 𝑎 ∈ ℝ such that

lim 𝑎𝑛 = 𝑎 .
𝑛→∞

Using quantifiers, this means

∄ 𝑎 ∈ ℝ s.t. ∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 − 𝑎| < 𝜀 .

The above is equivalent to showing that

∀ 𝑎 ∈ ℝ , ∃ 𝜀 > 0 s.t. ∀ 𝑁 ∈ ℕ , ∃ 𝑛 ≥ 𝑁 s.t. |𝑎𝑛 − 𝑎| ≥ 𝜀 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 166

Theorem 6.15

Let (𝑎𝑛 ) be the sequence defined by

𝑎𝑛 = (−1)𝑛 .
Then (𝑎𝑛 ) does not converge.

Proof
To prove that (𝑎𝑛 ) does not converge, we have to show that

∀ 𝑎 ∈ ℝ , ∃ 𝜀 > 0 s.t. ∀ 𝑁 ∈ ℕ , ∃ 𝑛 ≥ 𝑁 s.t. |𝑎𝑛 − 𝑎| ≥ 𝜀 .

Let 𝑎 ∈ ℝ. Choose
1
𝜀= .
2
Let 𝑁 ∈ ℕ. We distinguish two cases:

• 𝑎 ≥ 0: Choose 𝑛 = 2𝑁 + 1. Note that 𝑛 ≥ 𝑁 . Then

|𝑎𝑛 − 𝑎| = |𝑎2𝑁 +1 − 𝑎|
= |(−1)2𝑁 +1 − 𝑎|
= | − 1 − 𝑎|
=1+𝑎
≥1
1
> =𝜀,
2
where we used that 𝑎 ≥ 0, and therefore

| − 1 − 𝑎| = 1 + 𝑎 ≥ 1 .

• 𝑎 < 0: Choose 𝑛 = 2𝑁 . Note that 𝑛 ≥ 𝑁 . Then

|𝑎𝑛 − 𝑎| = |𝑎2𝑁 − 𝑎|
= |(−1)2𝑁 − 𝑎|
= |1 − 𝑎|
=1−𝑎
>1
1
> =𝜀,
2
where we used that 𝑎 < 0, and therefore

|1 − 𝑎| = 1 − 𝑎 > 1 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 167

6.4 Uniqueness of limit

In Definition 0.228 of convergence, we used the notation

lim 𝑎𝑛 = 𝑎 .
𝑛→∞

The above notation makes sense only if the limit is unique, that is, if we do not have that

lim 𝑎𝑛 = 𝑏 ,
𝑛→∞
for some
𝑎 ≠ 𝑏.
In the next theorem we will show that the limit is unique, if it exists.

Theorem 6.16: Uniqueness of limit

Let (𝑎𝑛 )𝑛∈ℕ be a sequence. Suppose that

lim 𝑎𝑛 = 𝑎 , lim 𝑎𝑛 = 𝑏 .
𝑛→∞ 𝑛→∞

Then 𝑎 = 𝑏.

Proof
Assume that,
lim 𝑎𝑛 = 𝑎 , lim 𝑎𝑛 = 𝑏 .
𝑛→∞ 𝑛→∞
Suppose by contradiction that
𝑎 ≠ 𝑏.
Choose
1
𝜀 ∶= |𝑎 − 𝑏| .
2
Therefore 𝜀 > 0, since |𝑎 − 𝑏| > 0. By the convergence 𝑎𝑛 → 𝑎,

∃ 𝑁1 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁1 , |𝑎𝑛 − 𝑎| < 𝜀 .

By the convergence 𝑎𝑛 → 𝑏,
∃ 𝑁2 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁2 , |𝑎𝑛 − 𝑏| < 𝜀 .
Define
𝑁 ∶= max{𝑁1 , 𝑁2 } .
Choose an 𝑛 ∈ ℕ such that 𝑛 ≥ 𝑁 . In particular

𝑛 ≥ 𝑁1 , 𝑛 ≥ 𝑁2 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 168

Then

2𝜀 = |𝑎 − 𝑏|
= |𝑎 − 𝑎𝑛 + 𝑎𝑛 − 𝑏|
≤ |𝑎 − 𝑎𝑛 | + |𝑎𝑛 − 𝑏|
<𝜀+𝜀
= 2𝜀 ,

where we used the triangle inequality in the first inequality. Hence 2𝜀 < 2𝜀, which gives a contradiction.

Example 6.17

Prove that

𝑛2 − 1 1
2
lim =
𝑛→∞ 2𝑛 − 3 2
According to Theorem 0.239, it suffices to show that the sequence

𝑛2 − 1
( )
2𝑛2 − 3 𝑛∈ℕ

converges to 21 , since then 12 can be the only limit.

Part 1. Rough Work.
Let 𝜀 > 0. We want to find 𝑁 ∈ ℕ such that
𝑛2 − 1 1
| 2
− |<𝜀, ∀𝑛 ≥ 𝑁 .
2𝑛 − 3 2
To this end, we compute:

𝑛2 − 1 1 2 (𝑛2 − 1) − (2𝑛2 − 3)
| − | = | |
2𝑛2 − 3 2 2 (2𝑛2 − 3)
1
=| 2 |
4𝑛 − 6
1
= 2
4𝑛 − 6
1
= 2
3𝑛 + 𝑛2 − 6
1
≤ 2
3𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 169

which holds if 𝑛 ≥ 3, since in this case 𝑛2 − 6 ≥ 0. Therefore

𝑛2 − 1 1 1
| 2
− |<𝜀 ⟸ <𝜀
2𝑛 − 3 2 3𝑛2
1
⟺ 3𝑛2 >
𝜀
2 1
⟺ 𝑛 >
3𝜀
1
⟺ 𝑛> .
√3𝜀
Looking at the above implications, it is clear that 𝑁 ∈ ℕ has to be chosen so that
1
𝑁 > .
√3𝜀
Moreover we need to recall that 𝑁 has to satisfy
𝑁 ≥3
for the estimates to hold.
Part 2. Formal Proof. We have to show that
𝑛2 − 1 1
∀𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , | 2
− |<𝜀.
2𝑛 − 3 2
Let 𝜀 > 0. Choose 𝑁 ∈ ℕ such that
1
𝑁 > max { , 3} .
√3𝜀
Let 𝑛 ≥ 𝑁 . Then
𝑛2 − 1 1 1
| 2
− |= 2
2𝑛 − 3 2 4𝑛 − 6
1
= 2
3𝑛 + 𝑛2 − 6
1
≤ 2
3𝑛
1
≤
3𝑁 2
<𝜀,
where we used that
𝑛≥𝑁 ≥3
which implies
𝑛2 − 6 ≥ 0 ,
in the third line. The last inequality holds, since it is equivalent to
1
𝑁 > .
√3𝜀

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 170

6.5 Bounded sequences

An important property of sequences is boundedness.

Definition 6.18: Bounded sequence

A sequence (𝑎𝑛 )𝑛∈ℕ is called bounded if there exists a constant 𝑀 ∈ ℝ, with 𝑀 > 0, such that

|𝑎𝑛 | ≤ 𝑀 , ∀𝑛 ∈ ℕ.

Definition 0.241 says that a sequence is bounded, if we can find some constant 𝑀 > 0 (possibly very large),
such that for all elements of the sequence it holds that
|𝑎𝑛 | ≤ 𝑀 ,
or equivalently, that
−𝑀 ≤ 𝑎𝑛 ≤ 𝑀 .

We now show that any sequence that converges is also bounded

Theorem 6.19

If a sequence (𝑎𝑛 )𝑛∈ℕ converges, then the sequence is bounded.

Proof
Suppose the sequence (𝑎𝑛 )𝑛∈ℕ converges and let

𝑎 ∶= lim 𝑎𝑛
𝑛→∞

By definition of convergence we have that

∀ 𝜀 > 0 , ∃𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 − 𝑎| < 𝜀 .

In particular, we can choose

𝜀=1
and let 𝑁 ∈ ℕ be that value such that

|𝑎𝑛 − 𝑎| < 1 , ∀𝑛 ≥ 𝑁 .

If 𝑛 ≥ 𝑁 we have, by the triangle inequality,

|𝑎𝑛 | = |𝑎𝑛 − 𝑎 + 𝑎|
≤ |𝑎𝑛 − 𝑎| + |𝑎|
< 1 + |𝑎| .

Set
𝑀 ∶= max {|𝑎1 | , |𝑎2 | , … , |𝑎𝑁 −1 | , 1 + |𝑎|} .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 171

Note that such maximum exists, being the set finite. Then

|𝑎𝑛 | ≤ 𝑀 , ∀𝑛 ∈ ℕ,

showing that (𝑎𝑛 ) is bounded.

The choice of 𝑀 in the above proof says that the sequence can behave wildly for a finite number of terms.
After that, it will stay close to the value of the limit, if the latter exists.

Example 6.20

In Theorem 6.8 we have shown that

1
lim =0
𝑛→∞ 𝑛
Hence, it follows from Theorem 6.19 that the sequence (1/𝑛) is bounded.
Indeed, we have that
1 1
| | = ≤ 1, ∀𝑛 ∈ ℕ,
𝑛 𝑛
since 𝑛 ≥ 1 for all 𝑛 ∈ ℕ.

Warning

The converse of Theorem 6.19 does not hold: There exist sequences (𝑎𝑛 ) which are bounded, but not
convergent.

Example 6.21

Define the sequence

𝑎𝑛 = (−1)𝑛 .
We have proven in Theorem 6.15 that (𝑎𝑛 ) is not convergent. However (𝑎𝑛 ) is bounded, with 𝑀 = 1, since

|𝑎𝑛 | = |(−1)𝑛 | = 1 = 𝑀 , ∀𝑛 ∈ ℕ.

Taking the contrapositive of the statement in Theorem 6.19 we get the following corollary:

Corollary 6.22

If a sequence (𝑎𝑛 )𝑛∈ℕ is not bounded, then the sequence does not converge.

Remark 6.23

For a sequence (𝑎𝑛 ) to be unbounded, it means that

∀ 𝑀 > 0 , ∃ 𝑛 ∈ ℕ s.t. |𝑎𝑛 | > 𝑀 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 172

The above is saying that no real number 𝑀 > 0 can be a bound for |𝑎𝑛 |, since there is always an index
𝑛 ∈ ℕ such that
|𝑎𝑛 | > 𝑀 .

We can use Corollary 6.22 to show that certain sequences do not converge.

Theorem 6.24

For all 𝑝 > 0, the sequence (𝑛𝑝 )𝑛∈ℕ does not converge.

Proof
Let 𝑝 > 0. We prove that the sequence (𝑛𝑝 )𝑛∈ℕ is unbounded, that is,

∀ 𝑀 > 0 , ∃ 𝑛 ∈ ℕ s.t. |𝑎𝑛 | > 𝑀 .

To this end, let 𝑀 > 0. Choose 𝑛 ∈ ℕ such that

𝑛 > 𝑀 1/𝑝 .

Then 𝑝
𝑎𝑛 = 𝑛𝑝 > (𝑀 1/𝑝 ) = 𝑀 .
This proves that the sequence (𝑛𝑝 ) is unbounded. Hence (𝑛𝑝 ) cannot converge, by Corollary 6.22.

Theorem 6.25

The sequence (log 𝑛)𝑛∈ℕ does not converge.

Proof
Let us show that (log 𝑛)𝑛∈ℕ is unbounded, that is,

∀ 𝑀 > 0 , ∃ 𝑛 ∈ ℕ s.t. |𝑎𝑛 | > 𝑀 .

To this end let 𝑀 > 0. Choose 𝑛 ∈ ℕ such that

𝑛 ≥ 𝑒 𝑀+1 .

Then
|𝑎𝑛 | = | log 𝑛| ≥ |log 𝑒 𝑀+1 | = 𝑀 + 1 > 𝑀 .
This proves that the sequence (log 𝑛) is unbounded. Hence (log 𝑛) cannot converge, by Corollary 6.22.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 173

6.6 Algebra of limits

Proving convergence using Definition 0.228 can be a tedious task. In this section we discuss how to prove
convergence, starting from known convergence results.

Theorem 6.26: Algebra of limits

Let (𝑎𝑛 )𝑛∈ℕ and (𝑏𝑛 )𝑛∈ℕ be sequences in ℝ. Suppose that

lim 𝑎𝑛 = 𝑎 , lim 𝑏𝑛 = 𝑏 ,
𝑛→∞ 𝑛→∞

for some 𝑎, 𝑏 ∈ ℝ. Then,

1. Limit of sum is the sum of limits:

lim (𝑎𝑛 ± 𝑏𝑛 ) = 𝑎 ± 𝑏
𝑛→∞

2. Limit of product is the product of limits:

lim (𝑎𝑛 𝑏𝑛 ) = 𝑎𝑏
𝑛→∞

3. If 𝑏𝑛 ≠ 0 for all 𝑛 ∈ ℕ and 𝑏 ≠ 0, then

𝑎𝑛 𝑎
lim ( )=
𝑛→∞ 𝑏𝑛 𝑏

Proof
Let (𝑎𝑛 )𝑛∈ℕ and (𝑏𝑛 )𝑛∈ℕ be sequences in ℝ and let 𝑐 ∈ ℝ. Suppose that, for some 𝑎, 𝑏 ∈ ℝ
lim 𝑎𝑛 = 𝑎 , lim 𝑏𝑛 = 𝑏 .
𝑛→∞ 𝑛→∞
Proof of Point 1.
We need to show that
lim (𝑎𝑛 ± 𝑏𝑛 ) = 𝑎 ± 𝑏 .
𝑛→∞
We only give a proof of the formula with +, since the case with − follows with a very similar proof.
Hence, we need to show that
∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |(𝑎𝑛 + 𝑏𝑛 ) − (𝑎 + 𝑏)| < 𝜀 .
Let 𝜀 > 0 and set
𝜀
𝜀 ̃ ∶=
.
2
Since 𝑎𝑛 → 𝑎, 𝑏𝑛 → 𝑏, and 𝜀 ̃ > 0, there exist 𝑁1 , 𝑁2 ∈ ℕ such that
|𝑎𝑛 − 𝑎| < 𝜀 ̃ , ∀ 𝑛 ≥ 𝑁1 ,
|𝑏𝑛 − 𝑏| < 𝜀 ̃ , ∀ 𝑛 ≥ 𝑁2 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 174

Define
𝑁 ∶= max{𝑁1 , 𝑁2 } .
For all 𝑛 ≥ 𝑁 we have, by the triangle inequality,

|(𝑎𝑛 + 𝑏𝑛 ) − (𝑎 + 𝑏)| = |(𝑎𝑛 − 𝑎) + (𝑏𝑛 − 𝑏)|

≤ |𝑎𝑛 − 𝑎| + |𝑏𝑛 − 𝑏|
< 𝜀̃ + 𝜀̃
=𝜀.

Proof of Point 2.
We need to show that
lim (𝑎𝑛 𝑏𝑛 ) = 𝑎𝑏 ,
𝑛→∞
which is equivalent to
∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 𝑏𝑛 − 𝑎𝑏| < 𝜀 .
Let 𝜀 > 0. The sequence (𝑎𝑛 ) converges, and hence is bounded, by Theorem 6.19. This means there exists
some 𝑀 > 0 such that
|𝑎𝑛 | ≤ 𝑀 , ∀ 𝑛 ∈ ℕ .
Define
𝜀
𝜀̃ = .
𝑀 + |𝑏|
Since 𝑎𝑛 → 𝑎, 𝑏𝑛 → 𝑏, and 𝜀 ̃ > 0, there exist 𝑁1 , 𝑁2 ∈ ℕ such that

|𝑎𝑛 − 𝑎| < 𝜀 ̃ , ∀ 𝑛 ≥ 𝑁1 .
|𝑏𝑛 − 𝑏| < 𝜀 ̃ , ∀ 𝑛 ≥ 𝑁2 .

Let
𝑁 ∶= max{𝑁1 , 𝑁2 } .
For all 𝑛 ≥ 𝑁 we have

|𝑎𝑛 𝑏𝑛 − 𝑎𝑏| = |𝑎𝑛 𝑏𝑛 − 𝑎𝑛 𝑏 + 𝑎𝑛 𝑏 − 𝑎𝑏|

≤ |𝑎𝑛 𝑏𝑛 − 𝑎𝑛 𝑏| + |𝑎𝑛 𝑏 − 𝑎𝑏|
= |𝑎𝑛 | |𝑏𝑛 − 𝑏| + |𝑏| |𝑎𝑛 − 𝑎|
< 𝑀 𝜀 ̃ + |𝑏| 𝜀 ̃
= (𝑀 + |𝑏|) 𝜀 ̃
=𝜀.

Proof of Point 3.
Suppose in addition that 𝑏𝑛 ≠ 0 and 𝑏 ≠ 0. We need to show that
𝑎𝑛 𝑎
lim = ,
𝑛→∞ 𝑏𝑛 𝑏

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 175

which is equivalent to
𝑎𝑛 𝑎
∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , | − |<𝜀.
𝑏𝑛 𝑏
We suppose in addition that 𝑏 > 0. The proof is very similar for the case 𝑏 < 0, and is hence omitted. Let
𝜀 > 0. Set
𝑏
𝛿 ∶= .
2
Since 𝑏𝑛 → 𝑏 and 𝛿 > 0, there exists 𝑁1 ∈ ℕ such that

|𝑏𝑛 − 𝑏| < 𝛿 ∀ 𝑛 ≥ 𝑁1 .

In particular we have
𝑏 𝑏
𝑏𝑛 > 𝑏 − 𝛿 = 𝑏 − = ∀ 𝑛 ≥ 𝑁1 .
2 2
Define
𝑏2
𝜀 ̃ ∶= 𝜀.
2(𝑏 + |𝑎|)
Since 𝜀 ̃ > 0 and 𝑎𝑛 → 𝑎, 𝑏𝑛 → 𝑏, there exist 𝑁2 , 𝑁3 ∈ ℕ such that

|𝑎𝑛 − 𝑎| < 𝜀 ̃ , ∀ 𝑛 ≥ 𝑁2 ,
|𝑏𝑛 − 𝑏| < 𝜀 ̃ , ∀ 𝑛 ≥ 𝑁3 .

Define
𝑁 ∶= max{𝑁1 , 𝑁2 , 𝑁3 } .
For all 𝑛 ≥ 𝑁 we have
𝑎𝑛 𝑎 𝑎 𝑏 − 𝑎𝑏𝑛
| − |=| 𝑛 |
𝑏𝑛 𝑏 𝑏𝑛 𝑏
1
= |𝑎𝑛 𝑏 − 𝑎𝑏 + 𝑎𝑏 − 𝑎𝑏𝑛 |
|𝑏𝑛 𝑏|
1
= |(𝑎𝑛 − 𝑎)𝑏 + 𝑎(𝑏 − 𝑏𝑛 )|
|𝑏𝑛 𝑏|
1
≤ (|𝑎𝑛 − 𝑎||𝑏| + |𝑎||𝑏 − 𝑏𝑛 |)
|𝑏𝑛 𝑏|
1
< (𝜀 ̃ 𝑏 + 𝜀|𝑎|)
̃
𝑏
𝑏
2
2(𝑏 + |𝑎|)
= 𝜀̃
𝑏2
=𝜀.

In the future we will refer to Theorem 0.249 as the Algebra of Limits. We now show how to use Theorem
0.249 for computing certain limits.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 176

Example 6.27

Prove that
3𝑛 3
lim = .
𝑛→∞ 7𝑛 + 4 7
Proof . We can rewrite
3𝑛 3
=
7𝑛 + 4 7 + 4
𝑛
By Theorem 6.12 we know that

3 → 3, 4 → 4, 7 → 7.

From Theorem 6.8 we know that

1
→ 0.
𝑛
Hence, it follows from Theorem 0.249 Point 2 that
4 1
= 4 ⋅ → 4 ⋅ 0 = 0.
𝑛 𝑛
By Theorem 0.249 Point 1 we have
4
7+ → 7 + 0 = 7.
𝑛
Finally we can use Theorem 0.249 Point 3 to infer
3𝑛 3 3
= → .
7𝑛 + 4 7 + 4 7
𝑛

Important

The technique shown in Example 6.27 is useful to compute limits of fractions of polynomials. To identify
the possible limit, if it exists, it is often best to divide by the largest power of 𝑛 in the denominator.

Example 6.28

Prove that
𝑛2 − 1 1
lim 2
= .
𝑛→∞ 2𝑛 − 3 2
Proof . Factor 𝑛2 to obtain
1
1−
𝑛2 −1 𝑛2
= .
2𝑛 − 3 2 − 3
2
𝑛2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 177

By Theorem 6.9 we have

1
→ 0.
𝑛2
We can then use the Algebra of Limits Theorem 0.249 Point 2 to infer
3
→3⋅0=0
𝑛2
and Theorem 0.249 Point 1 to get
1 3
1− → 1 − 0 = 1, 2− → 2 − 0 = 2.
𝑛2 𝑛2
Finally we use Theorem 0.249 Point 3 and conclude
1
1−
𝑛2 1
→ .
3 2
2− 2
𝑛
Therefore
1
1− 2
𝑛2 − 1 𝑛 1
lim = lim = .
𝑛→∞ 2𝑛2 − 3 𝑛→∞ 3 2
2− 2
𝑛

We can also use the Algebra of Limits to prove that certain limits do not exist.

Example 6.29

Prove that the sequence

4𝑛3 + 8𝑛 + 1
𝑎𝑛 =
7𝑛2 + 2𝑛 + 1
does not converge.

Proof . To show that the sequence (𝑎𝑛 ) does not converge, we divide by the largest power in
the denominator, which in this case is 𝑛2
4𝑛3 + 8𝑛 + 1
𝑎𝑛 =
7𝑛2 + 2𝑛 + 1
8 1
4𝑛 + + 2
𝑛 𝑛
=
2 1
7+ + 2
𝑛 𝑛
𝑏𝑛
=
𝑐𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 178

where we set
8 1 2 1
𝑏𝑛 ∶= 4𝑛 ++ 2 , 𝑐𝑛 ∶= 7 + + 2 .
𝑛 𝑛 𝑛 𝑛
Using the Algebra of Limits Theorem 0.249 we see that
2 1
𝑐𝑛 = 7 + + 2 → 7.
𝑛 𝑛
Suppose by contradiction that
𝑎𝑛 → 𝑎
for some 𝑎 ∈ ℝ. Then, by the Algebra of Limits Theorem 0.249 we would infer

𝑏𝑛 = 𝑐𝑛 ⋅ 𝑎𝑛 → 7𝑎 ,

concluding that 𝑏𝑛 is convergent to 7𝑎. We have that

8 1
𝑏𝑛 = 4𝑛 + 𝑑𝑛 , 𝑑𝑛 ∶= + 2.
𝑛 𝑛
Again by the Algebra of Limits Theorem 0.249 we get that
8 1
𝑑𝑛 = + 2 → 0,
𝑛 𝑛
and hence
4𝑛 = 𝑏𝑛 − 𝑑𝑛 → 7𝑎 − 0 = 7𝑎 .
This is a contradiction, since the sequence (4𝑛) is unbounded, and hence cannot be conver-
gent. Hence (𝑎𝑛 ) is not convergent.

Warning

Consider the sequence

4𝑛3 + 8𝑛 + 1
𝑎𝑛 =
7𝑛2 + 2𝑛 + 1
from the previous example. We have proven that (𝑎𝑛 ) is not convergent, by making use of the Algebra
of Limits.
Let us review a faulty argument to conclude that (𝑎𝑛 ) is not convergent. Write
𝑏𝑛
𝑎𝑛 = , 𝑏𝑛 ∶= 4𝑛3 + 8𝑛 + 1 , 𝑐𝑛 ∶= 7𝑛2 + 2𝑛 + 1 .
𝑐𝑛
The numerator
𝑏𝑛 = 4𝑛3 + 8𝑛 + 1
and denominator
𝑐𝑛 = 7𝑛2 + 2𝑛 + 1

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 179

are both unbounded, and hence (𝑏𝑛 ) and (𝑐𝑛 ) do not converge. One might be tempted to conclude that
(𝑎𝑛 ) does not converge. However this is false in general: as seen in Example 6.28, we have

𝑛2 − 1 1
lim 2
= ,
𝑛→∞ 2𝑛 − 3 2
while numerator and denominator are unbounded.

Sometimes it is useful to rearrange the terms of a sequence, before applying the Algebra of Limits.

Example 6.30

Define
2𝑛3 + 7𝑛 + 1 8𝑛 + 9
𝑎𝑛 ∶= ⋅ 3 .
5𝑛 + 9 6𝑛 + 8𝑛2 + 3
Prove that
8
lim 𝑎𝑛 = .
𝑛→∞ 15

Proof. The first fraction in (𝑎𝑛 ) does not converge, as it is unbounded. Therefore we cannot
use Point 2 in Theorem 0.249 directly. However, we note that

2𝑛3 + 7𝑛 + 1 8𝑛 + 9
𝑎𝑛 = ⋅ 3
5𝑛 + 9 6𝑛 + 8𝑛2 + 3
3
8𝑛 + 9 2𝑛 + 7𝑛 + 1
= ⋅ .
5𝑛 + 9 6𝑛3 + 8𝑛2 + 3

Factoring out 𝑛 and 𝑛3 , respectively, and using the Algebra of Limits, we see that

8𝑛 + 9 8 + 9/𝑛 8+0 8
= → =
5𝑛 + 9 5 + 9/𝑛 5+0 5
and
2 + 7/𝑛2 + 1/𝑛3 2+0+0 1
3
→ =
6 + 8/𝑛 + 3/𝑛 6+0+0 3
Therefore Theorem 0.249 Point 2 ensures that
8 1 8
𝑎𝑛 → ⋅ = .
5 3 15

6.7 Fractional powers

The Algebra of Limits Theorem 0.249 can also be used when fractional powers of 𝑛 are involved.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 180

Example 6.31

Prove that
𝑛7/3 + 2√𝑛 + 7
𝑎𝑛 =
4𝑛3/2 + 5𝑛
does not converge.

Proof . The largest power of 𝑛 in the denominator is 𝑛3/2 . Hence we factor out 𝑛3/2

𝑛7/3 + 2√𝑛 + 7
𝑎𝑛 =
4𝑛3/2 + 5𝑛
𝑛7/3−3/2 + 2𝑛1/2−3/2 + 7𝑛−3/2
=
4 + 5𝑛−3/2
𝑛5/6 + 2𝑛−1 + 7𝑛−3/2
=
4 + 5𝑛−3/2
𝑏
= 𝑛
𝑐𝑛
where we set
𝑏𝑛 ∶= 𝑛5/6 + 2𝑛−1 + 7𝑛−3/2 , 𝑐𝑛 ∶= 4 + 5𝑛−3/2 .
We see that 𝑏𝑛 is unbounded while 𝑐𝑛 → 4. By the Algebra of Limits (and usual contradiction
argument) we conclude that (𝑎𝑛 ) is divergent.

We now present a general result about the square root of a sequence.

Theorem 6.32

Let (𝑎𝑛 )𝑛∈ℕ be a sequence in ℝ such that

lim 𝑎𝑛 = 𝑎 ,
𝑛→∞
for some 𝑎 ∈ ℝ. If 𝑎𝑛 ≥ 0 for all 𝑛 ∈ ℕ and 𝑎 ≥ 0, then

lim 𝑎𝑛 = √𝑎 .
𝑛→∞ √

Proof
Let 𝜀 > 0. We the two cases 𝑎 > 0 and 𝑎 = 0:

• 𝑎 > 0: Define
𝑎
𝛿 ∶=
.
2
Since 𝛿 > 0 and 𝑎𝑛 → 𝑎, there exists 𝑁1 ∈ ℕ such that

|𝑎𝑛 − 𝑎| < 𝛿 , ∀ 𝑛 ≥ 𝑁1 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 181

In particular
𝑎 𝑎
𝑎𝑛 > 𝑎 − 𝛿 = 𝑎 − = , ∀ 𝑛 ≥ 𝑁1 ,
2 2
from which we infer
√𝑎𝑛 > √𝑎/2 , ∀ 𝑛 ≥ 𝑁1 ,
Now set
𝜀 ̃ ∶= (√𝑎/2 + √𝑎) 𝜀 .
Since 𝜀 ̃ > 0 and 𝑎𝑛 → 𝑎, there exists 𝑁2 ∈ ℕ such that

|𝑎𝑛 − 𝑎| < 𝜀 ̃ , ∀ 𝑛 ≥ 𝑁2 .

Let
𝑁 ∶= max{𝑁1 , 𝑁2 } .
For 𝑛 ≥ 𝑁 we have

(√𝑎𝑛 − √𝑎) (√𝑎𝑛 + √𝑎)

|√𝑎𝑛 − √𝑎| = | |
√𝑎𝑛 + √𝑎
|𝑎 − 𝑎|
= 𝑛
√𝑎𝑛 + √𝑎
𝜀̃
<
√𝑎/2 + √𝑎
=𝜀.

• 𝑎 = 0: In this case
𝑎𝑛 → 𝑎 = 0 .
Since 𝜀 2 > 0, there exists 𝑁 ∈ ℕ such that

|𝑎𝑛 − 0| = |𝑎𝑛 | < 𝜀 2 , ∀𝑛 ≥ 𝑁 .

Therefore
|√𝑎𝑛 − √0| = |√𝑎𝑛 | < √𝜀 2 = 𝜀 , ∀𝑛 ≥ 𝑁 .

Let us show an application of Theorem 6.32.

Example 6.33

Define the sequence

𝑎𝑛 = √9𝑛2 + 3𝑛 + 1 − 3𝑛 .
Prove that
1
lim 𝑎𝑛 = .
𝑛→∞ 2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 182

Proof . We first rewrite

𝑎𝑛 = √9𝑛2 + 3𝑛 + 1 − 3𝑛
(√9𝑛2 + 3𝑛 + 1 − 3𝑛) (√9𝑛2 + 3𝑛 + 1 + 3𝑛)
=
√9𝑛2 + 3𝑛 + 1 + 3𝑛
9𝑛2 + 3𝑛 + 1 − (3𝑛)2
=
√9𝑛2 + 3𝑛 + 1 + 3𝑛
3𝑛 + 1
= .
√9𝑛2 + 3𝑛 + 1 + 3𝑛
The biggest power of 𝑛 in the denominator is 𝑛. Therefore we factor out 𝑛:

𝑎𝑛 = √9𝑛2 + 3𝑛 + 1 − 3𝑛
3𝑛 + 1
=
√9𝑛2 + 3𝑛 + 1 + 3𝑛
1
3+
𝑛
= .
3 1
9+ + 2 +3
√ 𝑛 𝑛
By the Algebra of Limits we have
3 1
9+ + 2 → 9 + 0 + 0 = 9.
𝑛 𝑛
Therefore we can use Theorem 6.32 to infer

3 1
9+ + 2 → √9 .
√ 𝑛 𝑛

By the Algebra of Limits we conclude:

1
3+
𝑛 3+0 1
𝑎𝑛 = → = .
3 1
9+ + 2 +3 √9 + 3 2
√ 𝑛 𝑛

Example 6.34

Prove that the sequence

𝑎𝑛 = √9𝑛2 + 3𝑛 + 1 − 2𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 183

does not converge.

Proof. We rewrite 𝑎𝑛 as

𝑎𝑛 = √9𝑛2 + 3𝑛 + 1 − 2𝑛
(√9𝑛2 + 3𝑛 + 1 − 2𝑛)(√9𝑛2 + 3𝑛 + 1 + 2𝑛)
=
√9𝑛2 + 3𝑛 + 1 + 2𝑛
9𝑛2 + 3𝑛 + 1 − (2𝑛)2
=
√9𝑛2 + 3𝑛 + 1 + 2𝑛
5𝑛2 + 3𝑛 + 1
=
√9𝑛2 + 3𝑛 + 1 + 2𝑛
1
5𝑛 + 3 +
𝑛
=
3 1
9+ + 2 +2
√ 𝑛 𝑛
𝑏𝑛
= ,
𝑐𝑛
where we factored 𝑛, being it the largest power of 𝑛 in the denominator, and defined

1 3 1
𝑏𝑛 ∶= 5𝑛 + 3 + , 𝑐𝑛 ∶= 9+ + + 2.
𝑛 √ 𝑛 𝑛2

Note that
3 1
9+ + 2 →9
𝑛 𝑛
by the Algebra of Limits. Therefore

3 1
9 + + 2 → √9 = 3
√ 𝑛 𝑛

by Theorem 6.32. Hence

3 1
𝑐𝑛 = 9+ + 2 + 2 → 3 + 2 = 5.
√ 𝑛 𝑛

The numerator
1
𝑏𝑛 = 5𝑛 + 3 +
𝑛
is instead unbounded. Therefore (𝑎𝑛 ) is not convergent, by the Algebra of Limits and the
usual contradiction argument.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 184

6.8 Limit Tests

In this section we discuss a number of Tests to determine whether a sequence converges or not. These are
known as Limit Tests.

6.8.1 Squeeze Theorem

When a sequence (𝑎𝑛 ) oscillates, it is difficult to compute the limit. Examples of terms which produce oscilla-
tions are
(−1)𝑛 , sin(𝑛) , cos(𝑛) .
In such instance it might be useful to compare (𝑎𝑛 ) with other sequences whose limit is known. If we can
prove that (𝑎𝑛 ) is squeezed between two other sequences with the same limiting value, then we can show that
also (𝑎𝑛 ) converges to this value.

Theorem 6.35: Squeeze theorem

Let (𝑎𝑛 ) , (𝑏𝑛 ) and (𝑐𝑛 ) be sequences in ℝ. Suppose that

𝑏𝑛 ≤ 𝑎𝑛 ≤ 𝑐𝑛 , ∀𝑛 ∈ ℕ,

and that
lim 𝑏𝑛 = lim 𝑐𝑛 = 𝐿 .
𝑛→∞ 𝑛→∞
Then
lim 𝑎𝑛 = 𝐿 .
𝑛→∞

Proof
Let 𝜀 > 0. Since 𝑏𝑛 → 𝐿 and 𝑐𝑛 → 𝐿 , there exist 𝑁1 , 𝑁2 ∈ ℕ such that

−𝜀 < 𝑏𝑛 − 𝐿 < 𝜀 , ∀ 𝑛 ≥ 𝑁1 ,
−𝜀 < 𝑐𝑛 − 𝐿 < 𝜀 , ∀ 𝑛 ≥ 𝑁2 .

Set
𝑁 ∶= max{𝑁1 , 𝑁2 } .
Let 𝑛 ≥ 𝑁 . Using the assumption that 𝑏𝑛 ≤ 𝑎𝑛 ≤ 𝑐𝑛 , we get

𝑏𝑛 − 𝐿 ≤ 𝑎𝑛 − 𝐿 ≤ 𝑐𝑛 − 𝐿 .

In particular
−𝜀 < 𝑏𝑛 − 𝐿 ≤ 𝑎𝑛 − 𝐿 ≤ 𝑏𝑛 − 𝐿 < 𝜀 .
The above implies
−𝜀 < 𝑎𝑛 − 𝐿 < 𝜀 ⟹ |𝑎𝑛 − 𝐿| < 𝜀 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 185

Example 6.36

Prove that
(−1)𝑛
lim = 0.
𝑛→∞ 𝑛
Proof. For all 𝑛 ∈ ℕ we can estimate

−1 ≤ (−1)𝑛 ≤ 1 .

Therefore
−1 (−1)𝑛 1
≤ ≤ , ∀𝑛 ∈ ℕ.
𝑛 𝑛 𝑛
Moreover
−1 1
lim
= −1 ⋅ 0 = 0 , lim = 0.
𝑛→∞ 𝑛 𝑛→∞ 𝑛
By the Squeeze Theorem 0.258 we conclude

(−1)𝑛
lim = 0.
𝑛→∞ 𝑛

Example 6.37

Prove that
cos(3𝑛) + 9𝑛2 9
lim 2
= .
𝑛→∞ 11𝑛 + 15 sin(17𝑛) 11
Proof. We know that
−1 ≤ cos(𝑥) ≤ 1 , −1 ≤ sin(𝑥) ≤ 1 , ∀𝑥 ∈ ℝ.
Therefore, for all 𝑛 ∈ ℕ
−1 ≤ cos(3𝑛) ≤ 1 , −1 ≤ sin(17𝑛) ≤ 1 .
We can use the above to estimate the numerator in the given sequence:
−1 + 9𝑛2 ≤ cos(3𝑛) + 9𝑛2 ≤ 1 + 9𝑛2 . (6.7)
Concerning the denominator, we have
11𝑛2 − 15 ≤ 11𝑛2 + 15 sin(17𝑛) ≤ 11𝑛2 + 15
and therefore
1 1 1
≤ ≤ . (6.8)
11𝑛2 2 2
+ 15 11𝑛 + 15 sin(17𝑛) 11𝑛 − 15
Putting together (6.7)-(6.8) we obtain
−1 + 9𝑛2 cos(3𝑛) + 9𝑛2 1 + 9𝑛2
≤ ≤ .
11𝑛2 + 15 11𝑛2 + 15 sin(17𝑛) 11𝑛2 − 15

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 186

By the Algebra of Limits we infer

1
−
+9
−1 + 9𝑛2 𝑛2 0+9 9
= → =
2
11𝑛 + 15 11 + 15 11 + 0 11
𝑛2
and
1
+9
1 + 9𝑛2 𝑛2 0+9 9
= → = .
11𝑛 − 15 11 − 15
2 11 + 0 11
𝑛2
Applying the Squeeze Theorem 0.258 we conclude

cos(3𝑛) + 9𝑛2 9
lim = .
𝑛→∞ 11𝑛2 + 15 sin(17𝑛) 11

Warning

Suppose that the sequences (𝑎𝑛 ), (𝑏𝑛 ), (𝑐𝑛 ) satisfy

𝑏𝑛 ≤ 𝑎 𝑛 ≤ 𝑐 𝑛 , ∀𝑛 ∈ ℕ ,

and
𝑏𝑛 → 𝐿1 , 𝑐𝑛 → 𝐿2 , 𝐿1 ≠ 𝐿2 .
In general, we cannot conclude that 𝑎𝑛 converges.

Example 6.38

Consider the sequence

1
𝑎𝑛 = (1 + ) (−1)𝑛 .
𝑛
For all 𝑛 ∈ ℕ we can bound
1 1 1
−1 − ≤ (1 + ) (−1)𝑛 ≤ 1 + .
𝑛 𝑛 𝑛
However
1
−1 − ⟶ −1 − 0 = −1
𝑛
and
1
1+ ⟶ 1 + 0 = 1.
𝑛
Since
−1 ≠ 1 ,

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 187

we cannot apply the Squeeze Theorem 0.258 to conclude convergence of (𝑎𝑛 ). Indeed, (𝑎𝑛 ) is a divergent
sequence.

Proof. Suppose by contradiction that 𝑎𝑛 → 𝑎. We have

(−1)𝑛
𝑎𝑛 = (−1)𝑛 + = 𝑏 𝑛 + 𝑐𝑛
𝑛
where
(−1)𝑛
𝑏𝑛 ∶= (−1)𝑛 , 𝑐𝑛 ∶= .
𝑛
We have seen in Example 6.37 that 𝑐𝑛 → 0. Therefore, by the Algebra of Limits, we have

𝑏𝑛 = 𝑎 𝑛 − 𝑐 𝑛 ⟶ 𝑎 − 0 = 𝑎 .

However, Theorem 6.15 says that the sequence 𝑏𝑛 = (−1)𝑛 diverges. Contradiction. Hence
(𝑎𝑛 ) diverges.

6.8.2 Geometric sequences

Definition 6.39

A sequence (𝑎𝑛 ) is called a geometric sequence if

𝑎𝑛 = 𝑥 𝑛 ,

for some 𝑥 ∈ ℝ.

The value of |𝑥| determines whether or not a geometric sequence converges, as shown in the following theo-
rem.

Theorem 6.40: Geometric Sequence Test

Let 𝑥 ∈ ℝ and let (𝑎𝑛 ) be the sequence defined by

𝑎𝑛 ∶= 𝑥 𝑛 .

We have:

1. If |𝑥| < 1, then

lim 𝑎𝑛 = 0 .
𝑛→∞

2. If |𝑥| > 1, then sequence (𝑎𝑛 ) is unbounded, and hence divergent.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 188

Warning

The Geometric Sequence Test in Theorem 0.263 does not address the case

|𝑥| = 1 .

This is because, in this case, the sequence

𝑎𝑛 = 𝑥 𝑛
might converge or diverge, depending on the value of 𝑥. Indeed,

|𝑥| = 1 ⟹ 𝑥 = ±1 .

We therefore have two cases:

• 𝑥 = 1: Then
𝑎𝑛 = 1 𝑛 = 1
so that 𝑎𝑛 → 1 and (𝑎𝑛 ) is convergent.

• 𝑥 = −1: Then
𝑎𝑛 = 𝑥 𝑛 = (−1)𝑛
which is divergent by Theorem 6.15.

To prove Theorem 0.263 we need the following inequality, known as Bernoulli’s inequality.

Lemma 6.41: Bernoulli’s inequality

Let 𝑥 ∈ ℝ with 𝑥 > −1. Then

(1 + 𝑥)𝑛 ≥ 1 + 𝑛𝑥 , ∀𝑛 ∈ ℕ. (6.9)

Proof
Let 𝑥 ∈ ℝ, 𝑥 > −1. We prove the statement by induction:
• Base case: (6.9) holds with equality when 𝑛 = 1.
• Induction hypothesis: Let 𝑘 ∈ ℕ and suppose that (6.9) holds for 𝑛 = 𝑘, i.e.,
(1 + 𝑥)𝑘 ≥ 1 + 𝑘𝑥 .
Then
(1 + 𝑥)𝑘+1 = (1 + 𝑥)𝑘 (1 + 𝑥)
≥ (1 + 𝑘𝑥)(1 + 𝑥)
= 1 + 𝑘𝑥 + 𝑥 + 𝑘𝑥 2
≥ 1 + (𝑘 + 1)𝑥 ,
where we used that 𝑘𝑥 2 ≥ 0. Then (6.9) holds for 𝑛 = 𝑘 + 1.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 189

By induction we conclude (6.9).

We are ready to prove Theorem 0.263.

Proof: Proof of Theorem 0.263

Part 1. The case |𝑥| < 1.

If 𝑥 = 0, then
𝑎𝑛 = 𝑥 𝑛 = 0
so that 𝑎𝑛 → 0. Hence assume 𝑥 ≠ 0. We need to prove that
∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑥 𝑛 − 0| < 𝜀 .
Let 𝜀 > 0. We have
1
|𝑥| < 1 ⟹ > 1.
|𝑥|
Therefore
1 1
|𝑥| = , 𝑢 ∶= − 1 > 0.
1+𝑢 |𝑥|
Let 𝑁 ∈ ℕ be such that
1
𝑁 > ,
𝜀𝑢
so that
1
<𝜀.
𝑁𝑢
Let 𝑛 ≥ 𝑁 . Then
|𝑥 𝑛 − 0| = |𝑥|𝑛
1 𝑛
=( )
1+𝑢
1
=
(1 + 𝑢)𝑛
1
≤
1 + 𝑛𝑢
1
≤
𝑛𝑢
1
≤
𝑁𝑢
<𝜀,
where we used Bernoulli’s inequality (6.9) in the first inequality.
Part 2. The case |𝑥| > 1.
To prove that (𝑎𝑛 ) does not converge, we prove that it is unbounded. This means showing that
∀ 𝑀 > 0 , ∃𝑛 ∈ ℕ s.t. |𝑎𝑛 | > 𝑀 .
Let 𝑀 > 0. We have two cases:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 190

• 0 < 𝑀 ≤ 1: Choose 𝑛 = 1. Then

|𝑎1 | = |𝑥| > 1 ≥ 𝑀 .

• 𝑀 > 1: Choose 𝑛 ∈ ℕ such that

log 𝑀
𝑛> .
log |𝑥|
Note that log |𝑥| > 0 since |𝑥| > 1. Therefore

log 𝑀
𝑛> ⟺ 𝑛 log |𝑥| > log 𝑀
log |𝑥|
⟺ log |𝑥|𝑛 > log 𝑀
⟺ |𝑥|𝑛 > 𝑀 .

Then
|𝑎𝑛 | = |𝑥 𝑛 | = |𝑥|𝑛 > 𝑀 .

Hence (𝑎𝑛 ) is unbounded. By Corollary 6.22 we conclude that (𝑎𝑛 ) is divergent.

Example 6.42

We can apply Theorem 0.263 to prove convergence or divergence for the following sequences.

1. We have
1 𝑛
( ) ⟶0
2
since
1 1
| | = < 1.
2 2
2. We have
−1 𝑛
( ) ⟶0
2
since
−1 1
| | = < 1.
2 2
3. The sequence
−3 𝑛
𝑎𝑛 = ( )
2
does not converge, since
−3 3
| | = > 1.
2 2
4. As 𝑛 → ∞,
3𝑛 3 𝑛
= (− ) ⟶0
(−5)𝑛 5

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 191

since
3 3
|− | = < 1 .
5 5
5. The sequence
(−7)𝑛
𝑎𝑛 =
22𝑛
does not converge, since
(−7)𝑛 (−7)𝑛 7 𝑛
= 𝑛 = (− )
22𝑛 (22 ) 4
and
7 7
|− | = > 1 .
4 4

6.8.3 Ratio Test

Theorem 6.43: Ratio Test

Let (𝑎𝑛 ) be a sequence in ℝ such that

𝑎𝑛 ≠ 0 , ∀𝑛 ∈ ℕ.

1. Suppose that the following limit exists:

𝑎𝑛+1
𝐿 ∶= lim | |.
𝑛→∞ 𝑎𝑛
Then,
• If 𝐿 < 1 we have
lim 𝑎𝑛 = 0 .
𝑛→∞

• If 𝐿 > 1, the sequence (𝑎𝑛 ) is unbounded, and hence does not converge.

2. Suppose that there exists 𝑁 ∈ ℕ and 𝐿 > 1 such that

𝑎𝑛+1
| | ≥ 𝐿, ∀𝑛 ≥ 𝑁 .
𝑎𝑛

Then the sequence (𝑎𝑛 ) is unbounded, and hence does not converge.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 192

Proof
Define the sequence 𝑏𝑛 = |𝑎𝑛 |. Then,

𝑎𝑛+1 |𝑎𝑛+1 | 𝑏𝑛+1

| |= =
𝑎𝑛 |𝑎𝑛 | 𝑏𝑛
Part 1. Suppose that there exists the limit
𝑎𝑛+1
𝐿 ∶= lim | |.
𝑛→∞ 𝑎𝑛
Therefore
𝑏𝑛+1
lim = 𝐿. (6.10)
𝑛→∞ 𝑏𝑛

• 𝐿 < 1: Choose 𝑟 > 0 such that

𝐿 < 𝑟 < 1.
Set
𝜀 ∶= 𝑟 − 𝐿
By the convergence at (6.10) there exists 𝑁 ∈ ℕ such that

𝑏𝑛+1
| − 𝐿| < 𝜀 = 𝑟 − 𝐿 , ∀𝑛 ≥ 𝑁 .
𝑏𝑛
In particular
𝑏𝑛+1
− 𝐿 < 𝑟 − 𝐿, ∀𝑛 ≥ 𝑁 ,
𝑏𝑛
which implies
𝑏𝑛+1 < 𝑟 𝑏𝑛 , ∀𝑛 ≥ 𝑁 . (6.11)
Let 𝑛 ≥ 𝑁 , we can use (6.11) recursively and obtain

𝑏𝑁
0 ≤ 𝑏𝑛 < 𝑟𝑏𝑛−1 < … < 𝑟 𝑛−𝑁 𝑏𝑁 = 𝑟 𝑛 .
𝑟𝑁
In particular, we have proven that

𝑏𝑁
0 ≤ 𝑏𝑛 < 𝑟 𝑛 , ∀𝑛 ∈ ℕ. (6.12)
𝑟𝑁
Since |𝑟| < 1, by the Geometric Sequence Test Theorem 0.263 we infer

𝑟𝑛 → 0 .

The Algebra of Limits the yields

𝑏𝑁
𝑟𝑛 → 0.
𝑟𝑁

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 193

By the Squeeze Theorem 0.258 applied to (6.12), it follows that

𝑏𝑛 = |𝑎𝑛 | → 0 .

Since
− |𝑎𝑛 | ≤ 𝑎𝑛 ≤ |𝑎𝑛 | ,
and
−|𝑎𝑛 | → 0 , |𝑎𝑛 | → 0 ,
we can again apply the Squeeze Theorem 0.258 to infer

𝑎𝑛 → 0 .

• 𝐿 > 1: Choose 𝑟 > 0 such that

1 < 𝑟 < 𝐿.
Define
𝜀 ∶= 𝐿 − 𝑟 > 0 .
By the convergence (6.10), there exists 𝑁 ∈ ℕ such that

𝑏𝑛+1
| − 𝐿| < 𝜀 = 𝐿 − 𝑟 , ∀𝑛 ≥ 𝑁 .
𝑏𝑛
In particular,
𝑏𝑛+1
−(𝐿 − 𝑟) < − 𝐿, ∀𝑛 ≥ 𝑁 ,
𝑏𝑛
which implies
𝑏𝑛+1 > 𝑟 𝑏𝑛 , ∀𝑛 ≥ 𝑁 . (6.13)
Let 𝑛 ≥ 𝑁 . Applying (6.13) recursively we get

𝑏𝑁
𝑏𝑛 > 𝑟 𝑛−𝑁 𝑏𝑁 = 𝑟 𝑛 , ∀𝑛 ≥ 𝑁 . (6.14)
𝑟𝑁
Since |𝑟| > 1, by the Geometric Sequence Test we have that the sequence

(𝑟 𝑛 )

is unbounded. Therefore also the right hand side of (6.14) is unbounded, proving that (𝑏𝑛 ) is un-
bounded. Since
𝑏𝑛 = |𝑎𝑛 | ,
we conclude that (𝑎𝑛 ) is unbounded. By Corollary 6.22 we conclude that (𝑎𝑛 ) does not converge.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 194

Part 2. Suppose that there exists 𝑁 ∈ ℕ and 𝐿 > 1 such that

𝑎𝑛+1
| | ≥ 𝐿, ∀𝑛 ≥ 𝑁 .
𝑎𝑛

Since 𝑏𝑛 = |𝑎𝑛 |, we infer

𝑏𝑛+1 ≥ 𝐿 𝑏𝑛 , ∀𝑛 ≥ 𝑁 .
Arguing as above, we obtain
𝑏𝑁
𝑏𝑛 ≥ 𝐿 𝑛 , ∀𝑛 ≥ 𝑁 .
𝐿𝑁
Since 𝐿 > 1, we have that the sequence
𝑏𝑁
𝐿𝑛
𝐿𝑁
is unbounded, by the Geometric Sequence Test. Hence also (𝑏𝑛 ) is unbounded, from which we conclude
that (𝑎𝑛 ) is unbounded. By Corollary 6.22 we conclude that (𝑎𝑛 ) does not converge.

Let us apply the Ratio Test to some concrete examples.

Example 6.44

Let
3𝑛
,
𝑎𝑛 =
𝑛!
where we recall that 𝑛! (pronounced 𝑛 factorial) is defined by

𝑛! ∶= 𝑛 ⋅ (𝑛 − 1) ⋅ (𝑛 − 2) ⋅ … ⋅ 3 ⋅ 2 ⋅ 1 .

Prove that
lim 𝑎𝑛 = 0 .
𝑛→∞

Proof . We have
3𝑛+1
( )
𝑎𝑛+1 (𝑛 + 1)!
| |=
𝑎𝑛 3𝑛
( )
𝑛!
3𝑛+1 𝑛!
= 𝑛
3 (𝑛 + 1)!
3 ⋅ 3𝑛 𝑛!
= 𝑛
3 (𝑛 + 1)𝑛!
3
= ⟶ 𝐿 = 0.
𝑛+1
Hence, 𝐿 = 0 < 1 so 𝑎𝑛 → 0 by the Ratio Test in Theorem 0.266.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 195

Example 6.45

Consider the sequence

𝑛! ⋅ 3𝑛
𝑎𝑛 = .
√(2𝑛)!
Prove that (𝑎𝑛 ) is divergent.

Proof. We have

𝑎𝑛+1 (𝑛 + 1)! ⋅ 3𝑛+1 √(2𝑛)!

| |=
𝑎𝑛 𝑛! ⋅ 3𝑛
√(2(𝑛 + 1))!
(𝑛 + 1)! 3𝑛+1 √(2𝑛)!
= ⋅ 𝑛 ⋅
𝑛! 3
√(2(𝑛 + 1))!
For the first two fractions we have
(𝑛 + 1)! 3𝑛+1
⋅ 𝑛 = 3(𝑛 + 1) ,
𝑛! 3
while for the third fraction

√(2𝑛)! (2𝑛)!
=
(2𝑛 + 2)!
√(2(𝑛 + 1))! √
(2𝑛)!
=
√ (2𝑛 + 2) ⋅ (2𝑛 + 1) ⋅ (2𝑛)!
1
= .
√(2𝑛 + 1)(2𝑛 + 2)
Therefore, using the Algebra of Limits,

𝑎𝑛+1 3(𝑛 + 1)
| |=
𝑎𝑛
√(2𝑛 + 1)(2𝑛 + 2)
1
3𝑛 (1 + )
𝑛
=
1 2
𝑛2 (2 + ) (2 + )
√ 𝑛 𝑛
1
3 (1 + )
𝑛 3 3
= ⟶ = > 1.
1
(2 + ) (2 + )
2 √4 2
√ 𝑛 𝑛
By the Ratio Test we conclude that (𝑎𝑛 ) is divergent.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 196

Example 6.46

Let
𝑛!
𝑎𝑛 = .
100𝑛
Prove that (𝑎𝑛 ) is divergent.

Proof.
𝑎𝑛+1 100𝑛 (𝑛 + 1)! 𝑛 + 1
| |= = .
𝑎𝑛 100𝑛+1 𝑛! 100
Choose 𝑁 = 101. Then for all 𝑛 ≥ 𝑁 ,
𝑎𝑛+1 101
| |≥ > 1.
𝑎𝑛 100
Hence 𝑎𝑛 is divergent by the Ratio Test.

Warning

The Ratio Test in Theorem 0.266 does not address the case

𝐿 = 1.

This is because, in this case, the sequence (𝑎𝑛 ) might converge or diverge.
For example:

• Define the sequence

1
𝑎𝑛 = .
𝑛
We have
𝑎𝑛+1 𝑛
| |= → 𝐿 = 1.
𝑎𝑛 𝑛+1
Hence we cannot apply the Ratio Test. However we know that
1
lim = 0.
𝑛→∞ 𝑛

• Consider the sequence

𝑎𝑛 = 𝑛 .
We have
|𝑎𝑛+1 | 𝑛 + 1
= → 𝐿 = 1.
|𝑎𝑛 | 𝑛
Hence we cannot apply the Ratio Test. However we know that (𝑎𝑛 ) is unbounded, and thus diver-
gent.

If the sequence (𝑎𝑛 ) is geometric, the Ratio Test of Theorem 0.266 will give the same answer as the Geometric

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 197

Sequence Test of Theorem 0.263. This is the content of the following remark.

Remark 6.47

Let 𝑥 ∈ ℝ and define the geometric sequence

𝑎𝑛 = 𝑥 𝑛 .

Then
𝑎𝑛+1 |𝑥 𝑛+1 | |𝑥|𝑛+1
| |= = = |𝑥| → |𝑥| .
𝑎𝑛 |𝑥 𝑛 | |𝑥|𝑛
Hence:

• If |𝑥| < 1, the sequence (𝑎𝑛 ) converges by the Ratio Test

• If |𝑥| > 1, the sequence (𝑎𝑛 ) diverges by the Ratio Test.
• If |𝑥| = 1, the sequence (𝑎𝑛 ) might be convergent or divergent.

These results are in agreement with the Geometric Sequence Test.

6.9 Monotone sequences

We have seen in Theorem 6.19 that convergent sequences are bounded. We noted that the converse statement
is not true. For example the sequence
𝑎𝑛 = (−1)𝑛
is bounded but not convergent, as shown in Theorem 6.15. On the other hand, if a bounded sequence is
monotone, then it is convergent.

Definition 6.48: Monotone sequence

Let (𝑎𝑛 ) be a real sequence. We say that:

1. (𝑎𝑛 ) is increasing if
𝑎𝑛 ≤ 𝑎𝑛+1 , ∀𝑛 ≥ 𝑁 .

2. (𝑎𝑛 ) is decreasing if
𝑎𝑛 ≥ 𝑎𝑛+1 , ∀𝑛 ≥ 𝑁 .

3. (𝑎𝑛 ) is monotone if it is either increasing or decreasing.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 198

Example 6.49

• The sequence below is increasing

𝑛−1
𝑏𝑛 = .
𝑛
We have
𝑛 𝑛−1
𝑏𝑛+1 = > = 𝑏𝑛 ,
𝑛+1 𝑛
where the inequality holds because
𝑛 𝑛−1
> ⟺ 𝑛2 > (𝑛 − 1)(𝑛 + 1)
𝑛+1 𝑛
⟺ 𝑛2 > 𝑛2 − 1
⟺ 0 > −1

• The sequence below is decreasing

1
𝑎𝑛 = .
𝑛
We have
1 1
𝑎𝑛 = > = 𝑎𝑛+1 .
𝑛 𝑛+1

The main result about monotone sequences is the Monotone Convergence Theorem.

Theorem 6.50: Monotone Convergence Theorem

Let (𝑎𝑛 ) be a sequence in ℝ. Suppose that (𝑎𝑛 ) is bounded and monotone. Then (𝑎𝑛 ) converges.

Proof
Assume (𝑎𝑛 ) is bounded and monotone. Since (𝑎𝑛 ) is bounded, the set

𝐴 ∶= {𝑎𝑛 ∶ 𝑛 ∈ ℕ} ⊆ ℝ

is bounded below and above. By the Axiom of Completeness of ℝ there exist 𝑖, 𝑠 ∈ ℝ such that

𝑖 = inf 𝐴 , 𝑠 = sup 𝐴 .

We have two cases:

1. (𝑎𝑛 ) is increasing: We are going to prove that

lim 𝑎𝑛 = 𝑠 .
𝑛→∞

Equivalently, we need to prove that

∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 − 𝑠| < 𝜀 . (6.15)

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 199

Let 𝜀 > 0. Since 𝑠 is the smallest upper bound for 𝐴, this means

𝑠−𝜀

is not an upper bound. Therefore there exists 𝑁 ∈ ℕ such that

𝑠 − 𝜀 < 𝑎𝑁 . (6.16)

Let 𝑛 ≥ 𝑁 . Since 𝑎𝑛 is increasing, we have

𝑎𝑁 ≤ 𝑎 𝑛 , ∀𝑛 ≥ 𝑁 . (6.17)

Moreover 𝑠 is the supremum of 𝐴, so that

𝑎𝑛 ≤ 𝑠 < 𝑠 + 𝜀 , ∀𝑛 ∈ ℕ. (6.18)

Putting together estimates (6.16)-(6.17)-(6.18) we get

𝑠 − 𝜀 < 𝑎𝑁 ≤ 𝑎 𝑛 ≤ 𝑠 < 𝑠 + 𝜀 , ∀𝑛 ≥ 𝑁 .

The above implies

𝑠 − 𝜀 < 𝑎𝑛 < 𝑠 + 𝜀 , ∀𝑛 ≥ 𝑁 ,
which is equivalent to (6.15).

2. (𝑎𝑛 ) is decreasing: With a similar proof, one can show that

lim 𝑎𝑛 = 𝑖 .
𝑛→∞

This is left as an exercise.

6.9.1 Example: Euler’s Number

As an application of the Monotone Convergence Theorem we can give a formal definition for the Euler’s
Number
𝑒 = 2.71828182845904523536 …

Theorem 6.51

Consider the sequence

1 𝑛
𝑎𝑛 = (1 + ) .
𝑛
We have that:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 200

1. (𝑎𝑛 ) is monotone increasing,

2. (𝑎𝑛 ) is bounded.

In particular (𝑎𝑛 ) is convergent.

Proof
Part 1. We prove that (𝑎𝑛 ) is increasing
𝑎𝑛 ≥ 𝑎𝑛−1 , ∀𝑛 ∈ ℕ,
which by definition is equivalent to
1 𝑛 1 𝑛−1
(1 + ) ≥ (1 + ) , ∀𝑛 ∈ ℕ.
𝑛 𝑛−1
Summing the fractions we get
𝑛+1 𝑛 𝑛 𝑛−1
( ) ≥( ) .
𝑛 𝑛−1
Multiplying by ((𝑛 − 1)/𝑛)𝑛 we obtain
𝑛−1 𝑛 𝑛+1 𝑛 𝑛−1
( ) ( ) ≥ ,
𝑛 𝑛 𝑛
which simplifies to
1 𝑛 1
(1 − ) ≥ 1 − , ∀𝑛 ∈ ℕ. (6.19)
𝑛2 𝑛
Therefore (𝑎𝑛 ) is increasing if and only if (6.19) holds. Recall Bernoulli’s inequality from Lemma 0.264:
For 𝑥 ∈ ℝ, 𝑥 > −1, it holds
(1 + 𝑥)𝑛 ≥ 1 + 𝑛𝑥 , ∀ 𝑛 ∈ ℕ .
Appliying Bernoulli’s inequality with
1
𝑥=−
𝑛2
yields
1 𝑛 1 1
(1 − 2
) ≥ 1 + 𝑛 (− 2 ) = 1 − ,
𝑛 𝑛 𝑛
which is exactly (6.19). Then (𝑎𝑛 ) is increasing.
Part 2. We have to prove that (𝑎𝑛 ) is bounded, that is, that there exists 𝑀 > 0 such that
|𝑎𝑛 | ≤ 𝑀 , ∀𝑛 ∈ ℕ.
To this end, introduce the sequence (𝑏𝑛 ) by setting
1 𝑛+1
𝑏𝑛 ∶= (1 + ) .
𝑛
The sequence (𝑏𝑛 ) is decreasing.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 201

To prove (𝑏𝑛 ) is decreasing, we need to show that

𝑏𝑛−1 ≥ 𝑏𝑛 , ∀𝑛 ∈ ℕ.
By definition of 𝑏𝑛 , the above reads
1 𝑛 1 𝑛+1
(1 + ) ≥ (1 + ) , ∀𝑛 ∈ ℕ.
𝑛−1 𝑛
Summing the terms inside the brackets, the above is equivalent to
𝑛 𝑛 𝑛+1 𝑛 𝑛+1
() ≥( ) ( ).
𝑛−1 𝑛 𝑛
Multiplying by (𝑛/(𝑛 + 1))𝑛 we get
𝑛
𝑛2 𝑛+1
( 2 ) ≥( ).
𝑛 −1 𝑛
The above is equivalent to
𝑛
1 1
(1 +
2
) ≥ (1 + ) . (6.20)
𝑛 −1 𝑛
Therefore (𝑏𝑛 ) is decreasing if and only if (6.20) holds for all 𝑛 ∈ ℕ. By choosing
1
𝑥=
𝑛2 −1
in Bernoulli’s inequality, we obtain
𝑛
1 1
(1 + 2
) ≥ 1+𝑛( 2 )
𝑛 −1 𝑛 −1
𝑛
=1+ 2
𝑛 −1
1
≥1+ ,
𝑛
where in the last inequality we used that
𝑛 1
> ,
𝑛2 −1 𝑛
which holds, being equivalent to 𝑛2 > 𝑛2 − 1. We have therefore proven (6.20), and hence
(𝑏𝑛 ) is decreasing.
We now observe that For all 𝑛 ∈ ℕ
1 𝑛+1
𝑏𝑛 = (1 + )
𝑛
1 𝑛 1
= (1 + ) (1 + )
𝑛 𝑛
1
= 𝑎𝑛 (1 + )
𝑛
> 𝑎𝑛 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 202

Since (𝑎𝑛 ) is increasing and (𝑏𝑛 ) is decreasing, in particular

𝑎𝑛 ≥ 𝑎1 , 𝑏𝑛 ≤ 𝑏1 .

Therefore
𝑎1 ≤ 𝑎 𝑛 < 𝑏 𝑛 ≤ 𝑏 1 , ∀𝑛 ∈ ℕ.
We compute
𝑎1 = 2 , 𝑏1 = 4 ,
from which we get
2 ≤ 𝑎𝑛 ≤ 4 , ∀𝑛 ∈ ℕ.
Therefore
|𝑎𝑛 | ≤ 4 , ∀𝑛 ∈ ℕ,
showing that (𝑎𝑛 ) is bounded.
Part 3. The sequence (𝑎𝑛 ) is increasing and bounded above. Therefore (𝑎𝑛 ) is convergent by the Monotone
Convergence Theorem 0.273.

Thanks to Theorem 6.51 we can define the Euler’s Number 𝑒.

Definition 6.52: Euler’s Number

The Euler’s number is defined as

1 𝑛
𝑒 ∶= lim (1 + ) .
𝑛→∞ 𝑛

Setting 𝑛 = 1000 in the formula for (𝑎𝑛 ), we get an approximation of 𝑒:

𝑒 ≈ 𝑎1000 = 2.7169 .

6.10 Some important limits

In this section we investigate limits of some sequences to which the Limit Tests do not apply.

Theorem 6.53

Let 𝑥 ∈ ℝ, with 𝑥 > 0. Then

lim √
𝑛
𝑥 = 1.
𝑛→∞

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 203

Proof
Step 1. Assume 𝑥 ≥ 1. In this case
𝑛
√ 𝑥 ≥ 1.
Define
𝑏𝑛 ∶= √
𝑛
𝑥 − 1,
so that 𝑏𝑛 ≥ 0. By Bernoulli’s Inequality we have

𝑥 = (1 + 𝑏𝑛 )𝑛 ≥ 1 + 𝑛𝑏𝑛 .

Therefore
𝑥 −1
0 ≤ 𝑏𝑛 ≤ .
𝑛
Since
𝑥 −1
⟶ 0,
𝑛
by the Squeeze Theorem we infer 𝑏𝑛 → 0, and hence
𝑛
√ 𝑥 = 1 + 𝑏𝑛 ⟶ 1 + 0 = 1 ,

by the Algebra of Limits.

Step 2. Assume 0 < 𝑥 < 1. In this case
1
> 1.
𝑥
Therefore
𝑛
lim √ 1/𝑥 = 1 .
𝑛→∞
by Step 1. Therefore
1 1
𝑛
√ 𝑥= ⟶ = 1,
1
√1/𝑥
𝑛

by the Algebra of Limits.

Theorem 6.54

Let (𝑎𝑛 ) be a sequence such that 𝑎𝑛 → 0. Then

sin(𝑎𝑛 ) → 0 , cos(𝑎𝑛 ) → 1 .

Proof
Assume that 𝑎𝑛 → 0 and set
𝜋
𝜀 ∶= .
2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 204

By the convergence 𝑎𝑚 → 0 there exists 𝑁 ∈ ℕ such that

𝜋
|𝑎𝑛 | < 𝜀 = ∀𝑛 ≥ 𝑁 . (6.21)
2
Step 1. We prove that
sin(𝑎𝑛 ) → 0 .
By elementary trigonometry we have
𝜋 𝜋
0 ≤ | sin(𝑥)| = sin |𝑥| ≤ |𝑥| , ∀ 𝑥 ∈ [− , ] .
2 2
Therefore, since (6.21) holds, we can substitute 𝑥 = 𝑎𝑛 in the above inequality to get

0 ≤ | sin(𝑎𝑛 )| ≤ |𝑎𝑛 | , ∀𝑛 ≥ ℕ.

Since 𝑎𝑛 → 0, we also have |𝑎𝑛 | → 0. Therefore | sin(𝑎𝑛 )| → 0 by the Squeeze Theorem. This immediately
implies sin(𝑎𝑛 ) → 0.
Step 2. We prove that
cos(𝑎𝑛 ) → 1 .
Inverting the relation
cos2 (𝑥) + sin2 (𝑥) = 1 ,
we obtain
cos(𝑥) = ±√1 − sin2 (𝑥) .
We have that cos(𝑥) ≥ 0 for −𝜋/2 ≤ 𝑥 ≤ 𝜋/2. Thus

𝜋 𝜋
cos(𝑥) = √1 − sin2 (𝑥) , ∀ 𝑥 ∈ [− , ] .
2 2
Since (6.21) holds, we can set 𝑥 = 𝑎𝑛 in the above inequality and obtain

cos(𝑎𝑛 ) = √1 − sin2 (𝑎𝑛 ) , ∀𝑛 ≥ 𝑁 .

By Step 1 we know that sin(𝑎𝑛 ) → 0. Therefore, by the Algebra of Limits,

1 − sin2 (𝑎𝑛 ) ⟶ 1 − 0 ⋅ 0 = 1 .

Using Theorem 6.32 we have

cos(𝑎𝑛 ) = √1 − sin2 (𝑎𝑛 ) ⟶ √1 = 1 ,
concluding the proof.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 205

Theorem 6.55

Suppose (𝑎𝑛 ) is such that 𝑎𝑛 → 0 and

𝑎𝑛 ≠ 0 , ∀𝑛 ∈ ℕ.
Then
sin(𝑎𝑛 )
lim = 1.
𝑛→∞ 𝑎𝑛

Proof
The following elementary trigonometric inequality holds:
𝜋
sin(𝑥) < 𝑥 < tan(𝑥) , ∀ 𝑥 ∈ [0, ] .
2
Note that sin 𝑥 > 0 for 0 < 𝑥 < 𝜋/2. Therefore we can divide the above inequality by sin(𝑥) and take the
reciprocals to get
sin(𝑥) 𝜋
cos(𝑥) < < 1 , ∀ 𝑥 ∈ (0, ] .
𝑥 2
If −𝜋/2 < 𝑥 < 0, we can apply the above inequality to −𝑥 to obtain

sin(−𝑥)
cos(−𝑥) < < 1.
−𝑥
Recalling that cos(−𝑥) = cos(𝑥) and sin(−𝑥) = − sin(𝑥), we get

sin(𝑥) 𝜋
cos(𝑥) < < 1, ∀𝑥 ∈ ( − , 0] .
𝑥 2
Thus
sin(𝑥) 𝜋 𝜋
cos(𝑥) < < 1, ∀ 𝑥 ∈ [− , ] ∖ {0} . (6.22)
𝑥 2 2
Let
𝜋
𝜀 ∶= .
2
Since 𝑎𝑛 → 0, there exists 𝑁 ∈ ℕ such that
𝜋
|𝑎𝑛 | < 𝜀 = , ∀𝑛 ≥ 𝑁 .
2
Since 𝑎𝑛 ≠ 0 by assumption, the above shows that
𝜋 𝜋
𝑎𝑛 ∈ [ − , ] ∖ {0} , ∀𝑛 ≥ ℕ.
2 2
Therefore we can substitute 𝑥 = 𝑎𝑛 into (6.22) to get

sin(𝑎𝑛 )
cos(𝑎𝑛 ) < < 1, ∀𝑛 ≥ 𝑁 .
𝑎𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 206

We have
cos(𝑎𝑛 ) → 1
by Theorem 6.54. By the Squeeze Theorem we conclude that

sin(𝑎𝑛 )
lim = 1.
𝑛→∞ 𝑎𝑛

Warning

You might be tempted to apply L’Hôpital’s rule (which we did not cover in these Lecture Notes) to com-
pute
sin(𝑥)
lim .
𝑥→0 𝑥
This would yield the correct limit

sin(𝑥) (sin(𝑥))′
lim = lim = lim cos(𝑥) = 1 .
𝑥→0 𝑥 𝑥→0 (𝑥)′ 𝑥→0

However this is a circular argument, since the derivative of sin(𝑥) at 𝑥 = 0 is defined as the limit

sin(𝑥)
lim .
𝑥→0 𝑥

Theorem 6.56

Suppose (𝑎𝑛 ) is such that 𝑎𝑛 → 0 and

𝑎𝑛 ≠ 0 , ∀𝑛 ∈ ℕ.
Then
1 − cos(𝑎𝑛 ) 1 1 − cos(𝑎𝑛 )
lim = , lim = 0.
𝑛→∞ (𝑎𝑛 )2 2 𝑛→∞ 𝑎𝑛

Proof
Step 1. By Theorem 6.54 and Theorem 6.55, we have

sin(𝑎𝑛 )
cos(𝑎𝑛 ) → 1 , → 1.
𝑎𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 207

Therefore
1 − cos(𝑎𝑛 ) 1 − cos(𝑎𝑛 ) 1 + cos(𝑎𝑛 )
=
(𝑎𝑛 )2 (𝑎𝑛 )2 1 + cos(𝑎𝑛 )
2
1 − cos (𝑎𝑛 ) 1
= 2
(𝑎𝑛 ) 1 + cos(𝑎𝑛 )
2
sin(𝑎𝑛 ) 1 1 1
=( ) ⟶1⋅ = ,
𝑎𝑛 1 + cos(𝑎𝑛 ) 1+1 2
where in the last line we use the Algebra of Limits.
Step 2. We have
1 − cos(𝑎𝑛 ) 1 − cos(𝑎𝑛 ) 1
= 𝑎𝑛 ⋅ 2
⟶ 0 ⋅ = 0,
𝑎𝑛 (𝑎𝑛 ) 2
using Step 1 and the Algebra of Limits.

Example 6.57

• We have
1
lim 𝑛 sin ( ) = 1 . (6.23)
𝑛→∞ 𝑛
This is because
1
sin ( )
1 𝑛
𝑛 sin ( ) = ⟶ 1,
𝑛 1
𝑛
by Theorem 6.55 with 𝑎𝑛 = 1/𝑛.

• We have
1 1
lim 𝑛2 (1 − cos ( )) = . (6.24)
𝑛→∞ 𝑛 2
Indeed,
1
1 − cos ( )
1 𝑛 1
𝑛2 (1 − cos ( )) = ⟶ ,
𝑛 1 2
𝑛 2

by applying Theorem 6.56 with 𝑎𝑛 = 1/𝑛.

• We have
1
𝑛 (1 − cos ( ))
𝑛 1
lim = .
𝑛→∞ 1 2
sin ( )
𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 208

Indeed, using (6.24)-(6.23) and the Algebra of Limits

1 1
𝑛 (1 − cos ( )) 𝑛2 (1 − cos ( ))
𝑛 𝑛 1/2 1
= ⟶ = .
1 1 1 2
sin ( ) 𝑛 sin ( )
𝑛 𝑛

• We have
2 2
lim 𝑛 cos ( ) sin ( ) = 2 .
𝑛→∞ 𝑛 𝑛
This is because
2
cos ( ) ⟶ 1 ,
𝑛
by Theorem 6.54 applied with 𝑎𝑛 = 2/𝑛. Moreover
2
sin ( )
𝑛
⟶ 1,
2
𝑛
by Theorem 6.55 applied with 𝑎𝑛 = 2/𝑛. Therefore
2
sin ( )
2 2 2 𝑛
𝑛 cos ( ) sin ( ) = 2 ⋅ cos ( ) ⋅ ⟶2⋅1⋅1=2
𝑛 𝑛 𝑛 2
𝑛
where we used the Algebra of Limits.

• We have
𝑛2 + 1 1
lim sin ( ) = 1 .
𝑛→∞ 𝑛+1 𝑛
To prove it, note that
1
⎛ 1+ 2⎞
𝑛2
+1 1 𝑛 1 1+0
sin ( ) = ⎜ ⎟ ⋅ (𝑛 sin ( )) ⟶ ⋅ 1 = 1,
𝑛+1 𝑛 ⎜ 1 ⎟ 𝑛 1 + 0
1+
⎝ 𝑛 ⎠
where we used (6.23) and the Algebra of Limits.

Dr. Silvio Fanzon [email protected]

7 Sequences in ℂ
The theory for sequences in ℂ is very similar to that of sequences in ℝ. In ℝ, we said that a sequence (𝑎𝑛 )
converges to some number 𝑎 ∈ ℝ if for all 𝜀 > 0, it holds
|𝑎𝑛 − 𝑎| < 𝜀
for all 𝑛 suffieciently large. The definition of convergence in ℂ is essentially the same, with the absolute value
replaced by the complex modulus.

7.1 Definition and convergence

First of all, let us give the formal definition of sequence in ℂ.

Definition 7.1: Sequence of Complex numbers

A sequence 𝑎 in ℂ is a function
𝑎∶ ℕ → ℂ.
For 𝑛 ∈ ℕ, we denote the 𝑛-th element of the sequence 𝑎 by

𝑎𝑛 = 𝑎(𝑛)

and write the sequence as

(𝑎𝑛 )𝑛∈ℕ or (𝑎𝑛 ) .

In the following we define convergent sequences in ℂ.

Definition 7.2: Convergent sequence in ℂ

We say that a sequence (𝑎𝑛 )𝑛∈ℕ in ℂ converges to 𝑎 ∈ ℂ, or equivalently has limit 𝑎, denoted by
lim 𝑎𝑛 = 𝑎 or 𝑎𝑛 → 𝑎 ,
𝑛→∞
if for all 𝜀 ∈ ℝ with 𝜀 > 0, there exists 𝑁 ∈ ℕ such that for all 𝑛 ∈ ℕ, 𝑛 ≥ 𝑁 it holds that
|𝑎𝑛 − 𝑎| < 𝜀 .
Using quantifiers, we can write this as
∀ 𝜀 > 0, ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 − 𝑎| < 𝜀 .
If there exists 𝑎 ∈ ℂ such that lim𝑛→∞ 𝑎𝑛 = 𝑎, then we say that the sequence (𝑎𝑛 )𝑛∈ℕ is convergent.

209
Numbers, Sequences and Series Page 210

Important

In Definition 0.282 we still take 𝜀 to be real. This makes sense, since

|𝑧| = √𝑥 2 + 𝑦 2 ∈ ℝ

for all 𝑧 = 𝑥 + 𝑖𝑦 ∈ ℂ.

Example 7.3

Using Definition 0.282, prove that

(3 + 𝑖)𝑛 − 7𝑖
lim = 3 + 𝑖.
𝑛→∞ 𝑛
Part 1. Rough Work. Let 𝜀 > 0. We would like to understand for which values of 𝑛 the following holds:

(3 + 𝑖)𝑛 − 7𝑖
| − (3 + 𝑖)| < 𝜀 .
𝑛
We have
(3 + 𝑖)𝑛 − 7𝑖 (3 + 𝑖)𝑛 − 7𝑖 − (3 + 𝑖)𝑛
| − (3 + 𝑖)| = | |
𝑛 𝑛
| − 7𝑖|
=
𝑛
7
= .
𝑛

Therefore
7 7
<𝜀 ⟺ 𝑛> .
𝑛 𝜀
Part 2. Formal Proof. We want to prove that

(3 + 𝑖)𝑛 − 7𝑖
∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , | − (3 + 𝑖)| < 𝜀 .
𝑛
Let 𝜀 > 0. Choose 𝑁 ∈ ℕ such that
7
𝑁 > .
𝜀
The above is equivalent to
7
<𝜀.
𝑁
For 𝑛 ≥ 𝑁 we have
(3 + 𝑖)𝑛 − 7𝑖 7 7
| − (3 + 𝑖)| = ≤ <𝜀.
𝑛 𝑛 𝑁

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 211

7.2 Boundedness
Boundedness plays an important role for complex sequences.

Definition 7.4: Bounded sequence in ℂ

A sequence (𝑎𝑛 ) in ℂ is called bounded if there exists a constant 𝑀 ∈ ℝ, with 𝑀 > 0, such that

|𝑎𝑛 | ≤ 𝑀 , ∀𝑛 ∈ ℕ.

As it happens in ℝ, we have that complex sequences which converge are also bounded.

Theorem 7.5

If a sequence (𝑎𝑛 ) in ℂ converges, then the sequence is bounded.

The proof is identical to the one in ℝ, and is hence omitted. Similarly to real sequences, we can define
divergent complex sequences.

Definition 7.6: Divergent sequences in ℂ

We say that a sequence (𝑎𝑛 ) in ℂ is divergent if it is not convergent.

As a corollary of Theorem 7.5 we have the following.

Corollary 7.7

Let (𝑎𝑛 ) be a complex sequence. If (𝑎𝑛 ) is not bounded, then it is divergent.

7.3 Algebra of limits in ℂ

Most of the results about limits that we have shown in ℝ also hold in ℂ. The first result is the Algebra of
Limits.

Theorem 7.8: Algebra of limits in ℂ

Let (𝑎𝑛 ) and (𝑏𝑛 ) be sequences in ℂ. Suppose that

lim 𝑎𝑛 = 𝑎 , lim 𝑏𝑛 = 𝑏 ,
𝑛→∞ 𝑛→∞

for some 𝑎, 𝑏 ∈ ℂ. Then,

1. Limit of sum is the sum of limits:

lim (𝑎𝑛 ± 𝑏𝑛 ) = 𝑎 ± 𝑏
𝑛→∞

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 212

2. Limit of product is the product of limits:

lim (𝑎𝑛 𝑏𝑛 ) = 𝑎𝑏
𝑛→∞

3. If 𝑏𝑛 ≠ 0 for all 𝑛 ∈ ℕ and 𝑏 ≠ 0, then

𝑎𝑛 𝑎
lim ( )=
𝑛→∞ 𝑏𝑛 𝑏

The proof of Theorem 0.288 follows word by word the proof of the Algebra of Limits for sequences in ℝ: one
just needs to replace the absolute value by the complex modulus.
We can use the Algebra of Limits to compute limits of complex sequences.

Example 7.9

Let (𝑎𝑛 ) be the sequence defined by

(2 − 𝑖)𝑛2 + 6𝑖𝑛 − 5 − 3𝑖
𝑎𝑛 ∶= .
(6 + 3𝑖)𝑛2 + 11𝑖

The largest power of 𝑛 in the denominator is 𝑛2 . We hence divide by 𝑛2 to obtain

6𝑖 5 3𝑖
(2 − 𝑖) + − 2− 2
(2 − 𝑖)𝑛2 + 6𝑖𝑛 − 5 − 3𝑖 𝑛 𝑛 𝑛
𝑎𝑛 = =
(6 + 3𝑖)𝑛2 + 11𝑖 11𝑖
(6 + 3𝑖) + 2
𝑛
By the Algebra of Limits the right hand side converges to

(2 − 𝑖) + 0 − 0 − 0 2−𝑖
=
(6 + 3𝑖) + 0 6 + 3𝑖
By performing the complex division, we can write the limit in the form 𝑥 + 𝑖𝑦:
2−𝑖
lim 𝑎𝑛 =
𝑛→∞ 6 + 3𝑖
(2 − 𝑖)(6 − 3𝑖)
=
(6 + 3𝑖)(6 − 3𝑖)
12 − 6𝑖 − 6𝑖 + 3𝑖2
=
36 − 9𝑖2
9 − 12𝑖
=
45
1 4
= − 𝑖
5 15

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 213

7.4 Convergence to zero

One of the results that cannot hold for complex sequences is the Squeeze Theorem. Indeed the chain of
inequalities
𝑏𝑛 ≤ 𝑎𝑛 ≤ 𝑐𝑛
would not make sense in ℂ, since there is no order relation.
We can however prove the following (weaker) result.

Theorem 7.10

Let (𝑎𝑛 ) be a sequence in ℂ and suppose that

lim |𝑎𝑛 | = 0 .
𝑛→∞

Then
lim 𝑎𝑛 = 0 .
𝑛→∞

Proof
Assume that |𝑎𝑛 | → 0. We need to show that

∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 − 0| < 𝜀 .

Let 𝜀 > 0. Since |𝑎𝑛 | → 0, there exists 𝑁 ∈ ℕ such that

||𝑎𝑛 | − 0| < 𝜀 , ∀𝑛 ≥ 𝑁 .

Let 𝑛 ≥ 𝑁 . Then,

|𝑎𝑛 − 0| = |𝑎𝑛 |
= |𝑎𝑛 | − 0
= ||𝑎𝑛 | − 0|
<𝜀.

Note that the sequence |𝑎𝑛 | is real. Therefore the convergence of |𝑎𝑛 | can be studied using convergence results
in ℝ.

Example 7.11

Consider the complex sequence

1 1 𝑛
𝑎𝑛 = ( + 𝑖) .
2 3
Prove that 𝑎𝑛 → 0.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 214

Proof. We have

1 1 𝑛
|𝑎𝑛 | = |( + 𝑖) |
2 3
1 1 𝑛
= | + 𝑖|
2 3
𝑛
1 2 1 2
=( ( ) +( ) )
√ 2 3
𝑛
13
=( ) .
√ 36
Since
13
| | < 1,
√ 36
by the Geometric Sequence Test for real sequences, we conclude that

|𝑎𝑛 | → 0 .

Hence 𝑎𝑛 → 0 by Theorem 7.10.

Although the Squeeze Theorem cannot be used for complex sequences, sometimes it can be used to deal with
real terms in a complex sequence.

Example 7.12

Consider the sequence

2𝑖 cos(3𝑛)𝑛 + (7 − 𝑖)𝑛2
𝑎𝑛 ∶= .
3𝑛2 + 2𝑖𝑛 + sin(2𝑛)
Prove that
7 1
lim 𝑎𝑛 = − 𝑖.
𝑛→∞ 3 3
Proof. We divide by the largest power in the denominator, to get
2𝑖 cos(3𝑛)
+ (7 − 𝑖)
𝑛
𝑎𝑛 = .
2𝑖 sin(2𝑛)
3+ +
𝑛 𝑛2
Notice that
−1 ≤ cos(3𝑛) ≤ 1 , ∀𝑛 ∈ ℕ,
and thus
2 2 cos(3𝑛) 2
− ≤ ≤ , ∀𝑛 ∈ ℕ.
𝑛 𝑛 𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 215

Since
2 2
− ⟶ 0, ⟶ 0,
𝑛 𝑛
by the Squeeze Theorem we conclude that also

2 cos(3𝑛)
→ 0.
𝑛
In particular we have shown that

2𝑖 cos(3𝑛) 2 cos(3𝑛)
| |=| | → 0.
𝑛 𝑛
Using Theorem 7.10 we infer
2𝑖 cos(3𝑛)
→ 0.
𝑛
Similarly,
1 sin(2𝑛) 1
− 2
≤ 2
≤− 2, ∀𝑛 ∈ ℕ.
𝑛 𝑛 𝑛
Since
1 1
−
⟶ 0, ⟶ 0,
𝑛2 𝑛2
by the Squeeze Theorem we conclude

sin(2𝑛)
⟶ 0.
𝑛2
Finally, we have
2𝑖 2
| | = ⟶ 0,
𝑛 𝑛
and therefore
2𝑖
⟶0
𝑛
by Theorem 7.10. Using the Algebra of Limits in ℂ we conclude

2𝑖 cos(3𝑛)
+ (7 − 𝑖)
𝑛 0 + (7 − 𝑖) 7 1
𝑎𝑛 = ⟶ = − 𝑖.
2𝑖 sin(2𝑛) 3+0+0 3 3
3+ + 2
𝑛 𝑛

7.5 Geometric sequence Test and Ratio Test in ℂ

The Geometric Sequence Test and Ratio Test can be generalized to complex sequences.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 216

Theorem 7.13: Geometric sequence Test in ℂ

Let 𝑥 ∈ ℂ and let (𝑎𝑛 )𝑛∈ℕ be the geometric sequence in ℂ defined by

𝑎𝑛 ∶= 𝑥 𝑛 .

We have:

1. If |𝑥| < 1, then

lim 𝑎𝑛 = 0 .
𝑛→∞

2. If |𝑥| > 1, then sequence (𝑎𝑛 ) is unbounded, and hence divergent.

The proof can be obtained as in the real case, replacing the absolute value by the modulus.

Example 7.14

• Let
(−1 + 4𝑖)𝑛
𝑎𝑛 = .
(7 + 3𝑖)𝑛
We first rewrite
(−1 + 4𝑖)𝑛 −1 + 4𝑖 𝑛
𝑎𝑛 = = ( )
(7 + 3𝑖)𝑛 7 + 3𝑖
Then, we compute

−1 + 4𝑖 | − 1 + 4𝑖|
| |=
7 + 3𝑖 |7 + 3𝑖|
2 2
√(−1) + 4
=
√72 + 32
√17
=
√58
17
=
√ 58
<1

By the Geometric Sequence Test 𝑎𝑛 → 0.

• Let
(−5 + 12𝑖)𝑛
𝑏𝑛 = .
(3 − 4𝑖)𝑛
We first rewrite
(−5 + 12𝑖)𝑛 −5 + 12𝑖 𝑛
𝑏𝑛 = = ( ) .
(3 − 4𝑖)𝑛 3 − 4𝑖

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 217

We compute

−5 + 12𝑖 | − 5 + 12𝑖|
| |=
3 − 4𝑖 |3 − 4𝑖|
2 2
√5 + (−12)
=
2 2
√3 + (−4)
13
=
5
> 1.

By the Geometric Sequence Test, the sequence (𝑏𝑛 ) does not converge.

• Let 𝑖𝜋
𝑛
𝑐𝑛 = 𝑒 2 .
We have 𝑖𝜋
𝑛
|𝑐𝑛 | = |𝑒 2 | = 1 ,
and hence the Geometric Sequence Test cannot be applied. However, we can see that

𝑐𝑛 = (𝑖, −1, −𝑖, 1, 𝑖, −1, −𝑖, 1, …) ,

that is, 𝑐𝑛 assumes only the values {𝑖, −1, −𝑖, 1}, and each of them is assumed infinitely many times.
Thus 𝑐𝑛 is oscillating and it is divergent.

We now provide the statement of the Ratio Test in ℂ.

Theorem 7.15: Ratio Test in ℂ

Let (𝑎𝑛 ) be a sequence in ℂ such that

𝑎𝑛 ≠ 0 , ∀𝑛 ∈ ℕ.

1. Suppose that the following limit exists:

𝑎𝑛+1
𝐿 ∶= lim | |.
𝑛→∞ 𝑎𝑛
Then,
• If 𝐿 < 1 we have
lim 𝑎𝑛 = 0 .
𝑛→∞

• If 𝐿 > 1, the sequence (𝑎𝑛 ) is unbounded, and hence does not converge.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 218

2. Suppose that there exists 𝑁 ∈ 𝑁 and 𝐿 > 1 such that

𝑎𝑛+1
| | ≥ 𝐿, ∀𝑛 ≥ 𝑁 .
𝑎𝑛

Then the sequence (𝑎𝑛 ) is unbounded, and hence does not converge.

The proof of Theorem 0.295 follows word by word the proof of the Ratio Test Theorem in ℝ, and only two
minor modifications are needed:

• Replace the absolute value with the complex modulus,

• Instead of the Squeeze Theorem, use Theorem 7.10.

Example 7.16

Let
(4 − 3𝑖)𝑛
𝑎𝑛 = .
(2𝑛)!
Prove that 𝑎𝑛 → 0.

Proof. We compute

𝑎𝑛+1 (4 − 3𝑖)𝑛+1 (2𝑛)!

| |=| |
𝑎𝑛 (2(𝑛 + 1))! (4 − 3𝑖)𝑛
|4 − 3𝑖|𝑛+1 (2𝑛)!
= 𝑛 ⋅
|4 − 3𝑖| (2𝑛 + 2)!
|4 − 3𝑖|
=
(2𝑛 + 2)(2𝑛 + 1)
2 2
√4 + (−3)
=
(2𝑛 + 2)(2𝑛 + 1)
5
=
(2𝑛 + 2)(2𝑛 + 1)
5
𝑛2
= ⟶ 𝐿 = 0.
2 1
(2 + ) (2 + )
𝑛 𝑛
Since 𝐿 = 0 < 1, by the Ratio Test in ℂ we infer 𝑎𝑛 → 0.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 219

7.6 Convergence of real and imaginary part

A complex number 𝑧 ∈ ℂ can be written as
𝑧 = 𝑎 + 𝑏𝑖
for some 𝑎, 𝑏 ∈ ℝ, where
• 𝑎 is the real part of 𝑧,
• 𝑏 the imaginary part of 𝑧.
We can prove that a complex sequence converges if and only if both the real parts and the imaginary parts
converge.

Theorem 7.17

Let (𝑧𝑛 )𝑛∈ℕ be a sequence in ℂ. For 𝑛 ∈ ℕ, let 𝑎𝑛 , 𝑏𝑛 ∈ ℝ such that

𝑧𝑛 = 𝑎 𝑛 + 𝑏 𝑛 𝑖 .

Let
𝑧 = 𝑎 + 𝑏𝑖
with 𝑎, 𝑏 ∈ ℝ. Then
lim 𝑧𝑛 = 𝑧 ⟺ lim 𝑎𝑛 = 𝑎 , lim 𝑏𝑛 = 𝑏.
𝑛→∞ 𝑛→∞ 𝑛→∞

Proof
Part 1. Suppose that
lim 𝑧𝑛 = 𝑧 .
𝑛→∞
To prove that 𝑎𝑛 → 𝑎 we need to show that

∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑎𝑛 − 𝑎| < 𝜀 .

Let 𝜀 > 0. Since 𝑧𝑛 → 𝑧, there exists 𝑁 ∈ ℕ such that

|𝑧𝑛 − 𝑧| < 𝜀 , ∀𝑛 ≥ 𝑁 .

Let 𝑛 ≥ 𝑁 . Then
2
|𝑎𝑛 − 𝑎| = √(𝑎𝑛 − 𝑎)
2 2
≤ √(𝑎𝑛 − 𝑎) + (𝑏𝑛 − 𝑏)
= |(𝑎𝑛 − 𝑎) + (𝑏𝑛 − 𝑏) 𝑖|
= |(𝑎𝑛 + 𝑏𝑛 𝑖) − (𝑎 + 𝑏𝑖)|
= |𝑧𝑛 − 𝑧|
<𝜀.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 220

The proof for 𝑏𝑛 → 𝑏 is similar.

Part 2. Suppose that
lim 𝑎𝑛 = 𝑎 , lim 𝑏𝑛 = 𝑏 .
𝑛→∞ 𝑛→∞
To prove that 𝑧𝑛 → 𝑧 we need to show that

∀ 𝜀 > 0 , ∃ 𝑁 ∈ ℕ s.t. ∀ 𝑛 ≥ 𝑁 , |𝑧𝑛 − 𝑧| < 𝜀 .

Let 𝜀 > 0. Since 𝑎𝑛 → 𝑎, there exists 𝑁1 ∈ ℕ such that

𝜀
|𝑎𝑛 − 𝑎| < , ∀ 𝑛 ≥ 𝑁1 .
2
Since 𝑏𝑛 → 𝑏, there exists 𝑁2 ∈ ℕ such that
𝜀
|𝑏𝑛 − 𝑏| < , ∀ 𝑛 ≥ 𝑁2 .
2
Let
𝑁 ∶= max {𝑁1 , 𝑁2 } .
Let 𝑛 ≥ 𝑁 . By the triangle inequality,

|𝑧𝑛 − 𝑧| = |(𝑎𝑛 + 𝑏𝑛 𝑖) − (𝑎 + 𝑏𝑖)|

= |(𝑎𝑛 − 𝑎) + (𝑏𝑛 − 𝑏) 𝑖|
≤ |𝑎𝑛 − 𝑎| + |𝑏𝑛 − 𝑏| ⋅ |𝑖|
= |𝑎𝑛 − 𝑎| + |𝑏𝑛 − 𝑏|
𝜀 𝜀
< +
2 2
=𝜀.

Example 7.18

Consider the complex sequence

(4𝑛 + 3𝑛2 𝑖) (2𝑛2 + 𝑖)
𝑧𝑛 ∶=
5𝑛4
Show that
6
lim 𝑧𝑛 = 𝑖 .
𝑛→∞ 5

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 221

Proof. Let us find the real and imaginary parts of 𝑧𝑛

(4𝑛 + 3𝑛2 𝑖) (2𝑛2 + 𝑖)

𝑧𝑛 =
5𝑛4
8𝑛3 + 4𝑛𝑖 + 6𝑛4 𝑖 + 3𝑛2 𝑖2
=
5𝑛4
3
8𝑛 − 3𝑛 2 6𝑛4 + 4𝑛
= + 𝑖
5𝑛4 5𝑛4
= 𝑎 𝑛 + 𝑏𝑛 𝑖 .

Using the Algebra of Limits for real sequences we have that

8 3
−
8𝑛3 − 3𝑛2 𝑛 𝑛2 0−0
𝑎𝑛 = 4
= ⟶ = 0,
5𝑛 5 5
and
4
6+
6𝑛4+ 4𝑛 𝑛3 6+0 6
𝑏𝑛 = = ⟶ = .
5𝑛4 5 5 5
By Theorem 7.17 we conclude
6 6
lim 𝑧𝑛 = lim 𝑎𝑛 + 𝑖 lim 𝑏𝑛 = 0 + 𝑖 = 𝑖 .
𝑛→∞ 𝑛→∞ 𝑛→∞ 5 5

Dr. Silvio Fanzon [email protected]

8 Series
A series is the sum of all terms in a sequence:

𝑎1 + 𝑎 2 + 𝑎 3 + … + 𝑎 𝑛 + …

Since we are dealing with an infinite amount of terms, we need to be careful. For example, consider the
series
∞
∑(−1)𝑛 = −1 + 1 − 1 + 1 − 1 + 1 − … (8.1)
𝑛=1
If we sum the terms in pairs, we obtain
∞
∑(−1)𝑛 = (−1 + 1) + (−1 + 1) + (−1 + 1) + … = 0 .
𝑛=1

If we reorder the terms, we obtain a different result

∞
∑(−1)𝑛 = (1 + 1) − 1 + (1 + 1) − 1 + …
𝑘=1
= (2 − 1) + (2 − 1) + …
=1+1+…
= ∞.

Therefore commutativity of the sum does not hold when summing infinitely many terms. We need a good
definition of convergence.

8.1 Convergent series

We will develop the theory of series for sequences in ℂ. All the results for complex series will also hold for
series in ℝ. Some results will only hold for series in ℝ: this will be the case for comparison tests in which the
order relation of ℝ is needed.
We start by defining partial sums.

222
Numbers, Sequences and Series Page 223

Definition 8.1: Partial sums

Let (𝑎𝑛 ) be a sequence in ℂ. The 𝑘-th partial sum of (𝑎𝑛 ) is

𝑘
𝑠𝑘 ∶= 𝑎1 + 𝑎2 + … + 𝑎𝑘 = ∑ 𝑎𝑛
𝑛=1

This sequence (𝑠𝑘 )𝑘∈ℕ is called the sequence of partial sums.

We can use the sequence of partial sums to define convergence of a series.

Definition 8.2: Convergent series

Let (𝑎𝑛 ) be a sequence in ℂ. We denote the series of (𝑎𝑛 )𝑛∈ℕ by

∞
∑ 𝑎𝑛
𝑛=1

We say that this series converges to 𝑠 ∈ ℂ if

𝑘
lim ∑ 𝑎𝑛 = lim 𝑠𝑘 = 𝑠 .
𝑘→∞ 𝑛=1 𝑘→∞

In this case we write

∞
∑ 𝑎𝑛 = 𝑠 .
𝑛=1

Definition 8.3: Divergent series

Let (𝑎𝑛 ) be a sequence in ℂ. The series

∞
∑ 𝑎𝑛
𝑛=1
is divergent if the sequence of partial sums (𝑠𝑘 ) is divergent.

Example 8.4

Consider the series

∞
1
∑ .
𝑛=1 𝑛(𝑛 + 1)
Note that
1 1 1
= − .
𝑛(𝑛 + 1) 𝑛 𝑛 + 1

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 224

Hence, we can compute the partial sums 𝑠𝑛 as follows:

𝑘
1
𝑠𝑘 = ∑
𝑛=1 𝑛(𝑛 + 1)
𝑘
1 1
= ∑( − )
𝑛=1 𝑛 𝑛 + 1
1 1 1 1 1 1 1 1
= − + − + − +…+ −
1 2 2 3 3 4 𝑘 𝑘+1
1
=1− .
𝑘+1
Since,
1
lim 𝑠𝑘 = lim (1 − ) = 1,
𝑘→∞ 𝑘→∞ 𝑘+1
the series converges to 1, that is,
∞
1
∑ = 1.
𝑛=1 𝑛(𝑛 + 1)
A series of this kind is called a telescopic sum, since we can fold the entire partial sum together, in such
a way that only two terms remain.

Let us go back to the series considered in (8.1).

Example 8.5

Consider the series

∞
∑(−1)𝑛 .
𝑛=1
The partial sums 𝑠𝑘 are given by

𝑘
−1 if 𝑛 is odd
𝑠𝑘 = ∑(−1)𝑛 = {
𝑛=1 0 if 𝑛 is even.

Therefore 𝑠𝑘 diverges, so also the series diverges.

In general, it is a difficult taks to compute the exact value of a series. Therefore, we will mainly focus our
effort on determining whether a series converges or not.
The following theorem shows that if the terms in the sequence do not converge to 0, then the series cannot
converge.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 225

Theorem 8.6: Necessary Condition for Convergence

Let (𝑎𝑛 ) be a sequence in ℂ. If the series

∞
∑ 𝑎𝑛
𝑛=1
converges, then
lim 𝑎𝑛 = 0 .
𝑛→∞

Proof
Suppose that
∞
∑ 𝑎𝑛
𝑛=1
converges. By definition of convergent series there exists some 𝑠 ∈ ℂ such that

𝑘
lim 𝑠𝑘 = lim ∑ 𝑎𝑛 = 𝑠 .
𝑘→∞ 𝑘→∞ 𝑛=1

Then also
lim 𝑠𝑘−1 = 𝑠 .
𝑘→∞
Hence, by the Algebra of Limits in ℂ, we have that

lim (𝑠𝑘 − 𝑠𝑘−1 ) = lim 𝑠𝑘 − lim 𝑠𝑘−1 = 𝑠 − 𝑠 = 0 .

𝑘→∞ 𝑘→∞ 𝑘→∞

Noting that
𝑠𝑘 − 𝑠𝑘−1 = 𝑎𝑘 , ∀𝑘 ∈ ℕ,
we obtain
lim 𝑎𝑘 = lim (𝑠𝑘 − 𝑠𝑘−1 ) = 0 .
𝑘→∞ 𝑘→∞

Important

Theorem 0.304 is saying that, if

lim 𝑎𝑛 ≠ 0 ,
𝑛→∞
then the series
∞
∑ 𝑎𝑛
𝑛=1
does not converge.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 226

Example 8.7

Consider the series

∞
∑(−1)𝑛 . (8.2)
𝑛=1
We have that
lim 𝑎𝑛 = lim (−1)𝑛 ≠ 0 ,
𝑛→∞ 𝑛→∞
being (𝑎𝑛 ) divergent. Therefore the series at (8.2) diverges by Theorem 0.304.

Example 8.8

Consider the series

∞
𝑛
∑ .
𝑛=1 5𝑛 + 11
Then
𝑛 1 1
𝑎𝑛 ∶= = ⟶ ≠ 0.
5𝑛 + 11 5 + 11 5
𝑛
Hence, the series does not converge.

Important
∞
Theorem 0.304 says that if ∑𝑛=1 𝑎𝑛 converges, then

𝑎𝑛 → 0 .

The converse is false: In general the condition 𝑎𝑛 → 0 does not guarantee convergence of the associated
series, as shown in the example below.

Example 8.9

Consider the series

∞
1
∑ 𝑎𝑛 , 𝑎𝑛 ∶= .
𝑛=1 √𝑛 + 1 + √𝑛
By the Algebra of Limits we have
lim 𝑎𝑛 = 0 .
𝑛→∞
However the partial sums are given by
𝑘
𝑠𝑘 = ∑ 𝑎𝑛 = √𝑘 + 1 − 1 .
𝑛=1
Therefore (𝑠𝑘 ) is divergent and so the series is divergent.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 227

Exercise: Prove by induction that

𝑘
∑ 𝑎𝑛 = √𝑘 + 1 − 1 .
𝑛=1

Remark 8.10
It is customary to sum a series starting at 𝑛 = 1. However one could start the sum at any 𝑛 = 𝑁 with
𝑁 ∈ ℕ. This does not affect the convergence of the series, in the sense that
∞ ∞
∑ 𝑎𝑛 converges ⟺ ∑ 𝑎𝑛 converges.
𝑛=1 𝑛=𝑁

In case of convergence, we would of course have

∞ ∞
∑ 𝑎𝑛 = ∑ 𝑎𝑛 − (𝑎1 + … + 𝑎𝑁 −1 ) .
𝑛=𝑁 𝑛=1

Example 8.11

We have seen that

∞
1
∑ = 1.
𝑛=1 𝑛(𝑛 + 1)
Hence also the series
∞
1
∑
𝑛=7 𝑛(𝑛 + 1)
converges. However, in this case, the partial sums are given by
𝑘
1
𝑠𝑘 = ∑
𝑛=7 𝑛(𝑛 + 1)
𝑘
1 1
= ∑( − )
𝑛=7 𝑛 𝑛+1
1 1 1 1 1 1
= − + − +…+ −
7 8 8 9 𝑘 𝑘+1
1 1
= − .
7 𝑘+1
Therefore
∞
1 1
∑ = lim 𝑠𝑘 = .
𝑛=7 𝑛(𝑛 + 1) 𝑘→∞ 7

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 228

8.2 Geometric series

Definition 8.12: Geometric Series in ℂ

Let 𝑥 ∈ ℂ. The geometric series of ratio 𝑥 is the series
∞
∑ 𝑥𝑛 .
𝑛=0

Geometric series are one of the few types of series that can be explicitly computed, as stated in the following
theorem.

Theorem 8.13: Geometric Series Test

Let 𝑥 ∈ ℂ. We have:

1. If |𝑥| < 1, then the geometric series of ratio 𝑥 converges, with

∞
1
∑ 𝑥𝑛 = . (8.3)
𝑛=0 1−𝑥

2. If |𝑥| ≥ 1, then the geometric series of ratio 𝑥 diverges.

Important

Recall that the Geometric Sequence Test does not cover the case |𝑥| = 1, since in general the sequence

𝑎𝑛 = 𝑥 𝑛

could be convergent or divergent. However the Geometric Series Test covers the case |𝑥| = 1, in which
case
∞
∑ 𝑥𝑛
𝑛=0
diverges.

Proof
Part 1. Suppose that |𝑥| < 1. By using induction we prove that

𝑘
1 − 𝑥 𝑘+1
𝑠𝑘 ∶= ∑ 𝑥 𝑛 = , ∀𝑘 ∈ ℕ. (8.4)
𝑛=0 1−𝑥

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 229

• Base case: For 𝑘 = 0, we get that

1 − 𝑥1
𝑠0 = 𝑥 0 = 1 = .
1−𝑥
• Induction step: Let 𝑘 ∈ ℕ ∪ {0} and suppose that

1 − 𝑥 𝑘+1
𝑠𝑘 = .
1−𝑥
Then,
𝑠𝑘+1 = 𝑠𝑘 + 𝑥 𝑘+1
1 − 𝑥 𝑘+1
= + 𝑥 𝑘+1
1−𝑥
1 − 𝑥 𝑘+1 + (1 − 𝑥)𝑥 𝑘+1
=
1−𝑥
1 − 𝑥 𝑘+1 + 𝑥 𝑘+1 − 𝑥 𝑘+2
=
1−𝑥
1−𝑥 𝑘+2
= ,
1−𝑥
concluding the proof of the inductive step.

By the Principle of Induction formula (8.4) holds for all 𝑘 ∈ ℕ. Since |𝑥| < 1, by the Geometric Sequence
Test we infer
lim 𝑥 𝑘 = 0 .
𝑘→∞
Hence
∞
∑ 𝑥 𝑛 = lim 𝑠𝑘
𝑛=0 𝑘→∞

1 − 𝑥 𝑘+1
= lim
𝑘→∞ 1 − 𝑥
1 − 𝑥 ⋅ 𝑥𝑘
= lim
𝑘→∞ 1 − 𝑥
1
= ,
1−𝑥
where the last equality follows from the Algebra of Limits.
Part 2. Suppose |𝑥| ≥ 1. Then
lim 𝑥 𝑛 ≠ 0 . (8.5)
𝑛→∞
Indeed, suppose by contradiction that 𝑥𝑛 → 0. Hence, for 𝜀 = 1/2, there exists 𝑁 ∈ ℕ such that
1
|𝑥 𝑛 − 0| < 𝜀 = , ∀𝑛 ≥ ℕ.
2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 230

However
|𝑥 𝑛 − 0| = |𝑥 𝑛 | = |𝑥|𝑛 ≥ 1 ,
which yields
1
1<𝜀= ,
2
contradiction. Then (8.5) holds and the series
∞
∑ 𝑥𝑛
𝑛=0

diverges by the Necessary Condition in Theorem 0.304.

Let us show some applications of the Geometric Series Test of Theorem 0.311.

Example 8.14

• Since | 12 | < 1 we have

∞
1 𝑛 1
∑( ) = =2
𝑛=0 2
1
1−
2

• Since | −3
2
|= 3
2
> 1 the series
∞
−3 𝑛
∑( )
𝑛=0 2
does not converge.

• Since | −3
4
|= 3
4
< 1, we have
∞
−3 𝑛 1 1 4
∑( ) = = =
4 −3 7 7
𝑛=0 1−
4 4
• Since | − 1| = 1, the series
∞
∑(−1)𝑛
𝑛=0
does not converge.

Remark 8.15

If the sum of a Geometric Sries does not start at 𝑛 = 0, we need to tweak the summation formula at (8.3).

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 231

For example, if |𝑥| < 1, and we start the series at 𝑛 = 1, we get

∞
1 𝑥
∑ 𝑥𝑘 = −1= .
𝑛=1 1−𝑥 1−𝑥

Example 8.16

We have that
1
∞
1 𝑛
∑( ) = 2 = 1.
𝑛=1 2
1
1−
2

The Geometric Series Test of Theorem 0.311 can be applied to complex geometric series as well.

Example 8.17

• Consider the series

∞
1
∑ 𝑛 .
𝑛=0 (1 + 𝑖)
Since
1 1 𝑛
= ( )
(1 + 𝑖)𝑛 1+𝑖
and
1 1 1
| |= = < 1,
1+𝑖 √2
√12 + 12
Therefore the series converges by the Geometric Series Test, and
∞
1 1
∑ 𝑛 =
𝑛=0 (1 + 𝑖)
1
1−
1+𝑖
1
=
1+𝑖−1
1+𝑖
1+𝑖
=
𝑖
(1 + 𝑖)𝑖
=
𝑖2
𝑖−1
=
−1
= 1 − 𝑖.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 232

• Consider the series

∞
−5 − 5𝑖 𝑛
∑( ) .
𝑛=0 3 + 3𝑖
Since
−5 − 5𝑖 | − 5 − 5𝑖|
| |=
3 + 3𝑖 |3 + 3𝑖|
2 2
√(−5) + (−5)
=
√32 + 32
50
=
√ 18
5
= > 1,
3
the above series does not converge.

• Consider the series

∞
2+𝑖 𝑛
∑( ) .
𝑛=0 3 − 2𝑖
We have
2+𝑖 |2 + 𝑖|
| |=
3 − 2𝑖 |3 − 2𝑖|
√22 + 12
=
2 2
√3 + (−2)
5
= < 1.
√ 13

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 233

Therefore the series converges by the Geometric Series Test, and

∞
2+𝑖 𝑛 1
∑( ) =
3 − 2𝑖 2+𝑖
𝑛=0 1−
3 − 2𝑖
1
=
3 − 2𝑖 − (2 + 𝑖)
3 − 2𝑖
3 − 2𝑖
=
1 − 3𝑖
3 − 2𝑖 1 + 3𝑖
=
1 − 3𝑖 1 + 3𝑖
3 − 2𝑖 + 9𝑖 − 6𝑖2
=
1 − 9𝑖2
9 + 7𝑖
=
10
9 7
= + 𝑖
10 10

8.3 Algebra of Limits for Series

We have proven the Algebra of Limit Theorem for sequences in ℂ. A similar results holds for series as well.

Theorem 8.18: Algebra of Limits for Series

Let (𝑎𝑛 )𝑛∈ℕ and (𝑏𝑛 )𝑛∈ℕ be sequences in ℂ and let 𝑐 ∈ ℂ. Suppose that
∞ ∞
∑ 𝑎𝑛 = 𝑎 , ∑ 𝑏𝑛 = 𝑏 .
𝑛=1 𝑛=1

Then:

1. The sum of series is the series of the sums:

∞
∑ (𝑎𝑛 ± 𝑏𝑛 ) = 𝑎 ± 𝑏 .
𝑛=1

2. The product of a series with a number obeys

∞
∑ 𝑐 ⋅ 𝑎𝑛 = 𝑐 ⋅ 𝑎 .
𝑛=1

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 234

Proof
Part 1. We prove the formula with the + sign, since in the other case the proof is the same. To this end,
define the partial sums
𝑘 𝑘 𝑘
𝑠𝑘 ∶= ∑ 𝑎𝑛 , 𝑡𝑘 ∶= ∑ 𝑏𝑛 , 𝑣𝑘 ∶= ∑ (𝑎𝑛 + 𝑏𝑛 ) .
𝑛=1 𝑛=1 𝑛=1
We can write
𝑘
𝑣𝑘 = ∑ (𝑎𝑛 + 𝑏𝑛 )
𝑛=1
= (𝑎1 + 𝑏1 ) + … + (𝑎𝑘 + 𝑏𝑘 )
= (𝑎1 + … + 𝑎𝑘 ) + (𝑏1 + … + 𝑏𝑘 )
= 𝑠 𝑘 + 𝑡𝑘 .
By assumption 𝑠𝑘 → 𝑎 and 𝑡𝑘 → 𝑏. Hence, by the Algebra of Limits in ℂ, we infer
∞
∑ (𝑎𝑛 + 𝑏𝑛 ) = lim 𝑣𝑘
𝑛=1 𝑘→∞

= lim (𝑠𝑘 + 𝑡𝑘 )
𝑘→∞
= lim 𝑠𝑘 + lim 𝑡𝑘
𝑘→∞ 𝑘→∞
= 𝑎+𝑏.
Part 2. Denote the partial sums by
𝑘 𝑘
𝑠𝑘 ∶= ∑ 𝑎𝑛 , 𝑡𝑘 ∶= ∑ 𝑐 ⋅ 𝑎𝑛 .
𝑛=1 𝑛=1
We can write
𝑘
𝑡𝑘 = ∑ 𝑐 ⋅ 𝑎 𝑘
𝑛=1
= 𝑐 ⋅ 𝑎1 + 𝑐 ⋅ 𝑎 2 + … + 𝑐 ⋅ 𝑎𝑘
= 𝑐 ⋅ (𝑎1 + 𝑎2 + … + 𝑎𝑘 ) =
= 𝑐 ⋅ 𝑠𝑘 .
By assumption 𝑠𝑘 → 𝑎, so that the Algebra of Limits in ℂ allows to conclude
∞
∑ 𝑐 ⋅ 𝑎𝑛 = lim 𝑡𝑘
𝑛=1 𝑘→∞

= lim 𝑐 ⋅ 𝑠𝑘
𝑘→∞
= 𝑐 ⋅𝑎.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 235

Let us apply the Algebra of Limits for Series to a concrete example.

Example 8.19

Consider the following series

∞
1 𝑛 2 𝑛
∑ (2 ( ) + ( ) ) .
𝑛=0 3 3
Note that
∞ ∞
1 𝑛 1 3 2 𝑛 1
∑( ) = = , ∑( ) = = 3,
𝑛=0 3
1 2 3 2
1− 𝑛=0 1−
3 3
by the Geometric Series Test of Theorem 0.311. Therefore we can apply the Algebra of Limit for Series
to conclude
∞ ∞ ∞
1 𝑛 2 𝑛 1 𝑘 2 𝑘
∑ (2 ( ) + ( ) ) = 2 ⋅ ∑ ( ) + ∑ ( )
𝑛=0 3 3 𝑛=0 3 𝑛=0 3
3
=2⋅ +3=6
2

Important

The Algebra of Limits Theorem 0.316 does not discuss product and quotient of series. The situation
becomes more complicated in this case: Indeed, we have

(𝑎1 + 𝑎2 ) ⋅ (𝑎2 + 𝑏2 ) = 𝑎1 𝑏1 + 𝑎2 𝑏2 + 𝑎1 𝑏2 + 𝑎2 𝑏1 .

Therefore, in general, we can expect

∞ ∞ ∞
(∑ 𝑎𝑛 ) ⋅ (∑ 𝑏𝑛 ) ≠ ∑ 𝑎𝑛 ⋅ 𝑏𝑛 . (8.6)
𝑛=0 𝑛=0 𝑛=0

A way to compute
∞ ∞
(∑ 𝑎𝑛 ) ⋅ (∑ 𝑏𝑛 )
𝑛=0 𝑛=0
is through the so-called Cauchy Product of two series. We do not cover the latter, and the interested
reader can refer to Page 82 in [1].
Similarly, we expect that
∞
∑ 𝑎𝑛
∞
𝑛=0 𝑎𝑛
≠∑ . (8.7)
𝑛=0 𝑏𝑛
∞
∑ 𝑏𝑛
𝑛=0

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 236

Let us give two examples to show that formulas (8.6) and (8.7) hold.

Example 8.20

• We know that
∞
1 𝑛 1
∑( ) = = 2. (8.8)
𝑛=0 2
1
1−
2
Therefore
∞ ∞
1 𝑛 1 𝑛
(∑ ( ) ) ⋅ (∑ ( ) ) = 2 ⋅ 2 = 4 .
𝑛=0 2 𝑛=0 2
However
∞ ∞
1 𝑛 1 𝑛 1 𝑛 1 4
∑( ) ⋅( ) = ∑( ) = = .
𝑛=0 2 2 𝑛=0 4
1 3
1−
4
Hence
∞ ∞ ∞
1 𝑛 1 𝑛 4 1 𝑛 1 𝑛
(∑ ( ) ) ⋅ (∑ ( ) ) = 4 ≠ = ∑ ( ) ⋅ ( ) .
𝑛=0 2 𝑛=0 2 3 𝑛=0 2 2

• Using (8.8) we have

∞
1 𝑛
∑( )
𝑛=0 2 2
= = 1.
∞
1 𝑛 2
∑( )
𝑛=0 2
On the other hand
1 𝑛
( )
∞ ∞
∑ 2
𝑛 = ∑1,
𝑛=0 ( 1 ) 𝑛=0
2
which does not converge by the Necessary Condition, since 1 does not converge to 0. Therefore
∞
1 𝑛
∑( ) 1 𝑛
𝑛=0 2
( )
∞
≠∑ 2 𝑛.
∞
1 𝑛 𝑛=0 ( 1 )
∑( ) 2
𝑛=0 2

8.4 Non-negative series

We now investigate series of which all terms are non-negative. To be precise:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 237

Definition 8.21: Non-negative series

Let (𝑎𝑛 ) be a sequence in ℝ. We call the series

∞
∑ 𝑎𝑛
𝑛=1

a non-negative series if
𝑎𝑛 ≥ 0 , ∀𝑛 ∈ ℕ.

The key remark for non-negative series is that the partial sums are increasing.

Lemma 8.22
Let (𝑎𝑛 ) be a sequence in ℝ with
𝑎𝑛 ≥ 0 , ∀𝑛 ∈ ℕ.
Define the partial sums as
𝑘
𝑠𝑘 ∶= ∑ 𝑎𝑛 .
𝑛=1
The sequence (𝑠𝑘 ) is increasing.

Proof
For all 𝑘 ∈ ℕ we have
𝑘+1
𝑠𝑘+1 = ∑ 𝑎𝑛 = 𝑠𝑘 + 𝑎𝑘+1 ≥ 𝑠𝑘 ,
𝑛=1
where we used that 𝑎𝑘+1 ≥ 0. Therefore (𝑠𝑘 ) is increasing.

Therefore, if we have a series with non-negative terms

∞
∑ 𝑎𝑛
𝑛=1

there are only 2 options:

∞
• ∑𝑛=1 𝑎𝑛 converges,
∞
• ∑𝑛=1 𝑎𝑛 diverges to +∞.

This is because the partial sums (𝑠𝑘 ) are increasing. Therefore we have either:

• (𝑠𝑘 ) is bounded above: Then 𝑠𝑘 converges by the Monotone Convergence Theorem

• (𝑠𝑘 ) is not bounded above: Therefore 𝑠𝑘 diverges to +∞.

We present several tests for the convergence of non-negative series.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 238

8.4.1 Cauchy Condensation Test

Let us start with the study of the two non-negative series:

∞ ∞
1 1
∑ 2, ∑ .
𝑛=1 𝑛 𝑛=1 𝑛

Question 8.23

Do the above series converge or diverge?

The answer is that the first series converges, while the second diverges. We prove it in the next two theo-
rems.

Theorem 8.24

The following series converges

∞
1
∑ 2
.
𝑛=1 𝑛

Proof
For 𝑘 ∈ ℕ define the sequence of partial sums

𝑘
1
𝑠𝑘 ∶= ∑ 2
.
𝑛=1 𝑛

Note that
1 1 1 1
𝑠𝑘 = 1 + 2
+ 2 + 2 +…+ 2
2 3 4 𝑘
1 1 1 1
=1+ + + +…+
2⋅2 3⋅3 4⋅4 𝑘⋅𝑘
1 1 1 1
<1+ + + +…+
2⋅1 3⋅2 4⋅3 𝑘 ⋅ (𝑘 − 1)
1 1 1 1 1 1
=1+( − )+( − )+…+( − )
2 3 3 4 𝑘−1 𝑘
1
=1+1−
𝑘
1
=2−
𝑘
< 2,

showing that 𝑠𝑘 is bounded above. Recall that 𝑠𝑘 is increasing, by Lemma 8.22. Therefore, by the Mono-

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 239

tone Convergence Theorem, we conclude that 𝑠𝑘 converges. Hence the series

∞
1
∑ 2
𝑛=1 𝑛

is convergent.

Theorem 8.25: Harmonic series

The harmonic series

∞
1
∑
𝑛=1 𝑛
is divergent.

Proof
For 𝑘 ∈ ℕ define the sequence of partial sums

𝑘
1
𝑠𝑘 ∶= ∑ .
𝑛=1 𝑛

Note that
1
𝑠2 = 1 +
2
while
1 1 1
𝑠4 = 1 + +( + )
2 3 4
1 1 1
>1+ +( + )
2 4 4
1 1
=1+ + 2( )
2 4
1 1
=1+ +
2 2
1
= 1 + 2( )
2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 240

Similarly
1 1 1 1 1 1 1
𝑠4 = 1 + +( + )+( + + + )
2 3 4 5 6 7 8
1 1 1 1 1 1 1
>1+ +( + )+( + + + )
2 4 4 8 8 8 8
1 1 1
=1+ + 2( ) + 4( )
2 4 8
1 1 1
=1+ + +
2 2 2
1
= 1 + 3( )
2

Proceeding in a similar way, for 𝑘 ∈ ℕ we have

1 1 1 1 1 1 1
𝑠2𝑘 = 1 + +( + )+( + + + )
2 3 4 5 6 7 8
1 1
+…+( +…+ )
2𝑘−1 +1 2𝑘
1 1 1 1 1 1 1
>1+ +( + )+( + + + )
2 4 4 8 8 8 8
1 1
+…+( +…+ )
2𝑘 2𝑘
1 1 1 1
= 1 + + 2 ( ) + 4 ( ) + … + 2𝑘−1 ( )
2 4 8 2𝑘
1 1 1 1
=1+ + + +…+
2 2 2 2
1
=1+𝑘( )
2

Hence
1
𝑠2𝑘 > 1 + 𝑘 ( ) , ∀ 𝑘 ∈ ℕ ,
2
showing that 𝑠2𝑘 is unbounded. Therefore 𝑠𝑘 is unbounded, and 𝑠𝑘 does not converge. Therefore the
series
∞
1
∑
𝑛=1 𝑛
is divergent.

The proofs of the above theorems inspire the Cauchy Condensation Test.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 241

Theorem 8.26: Cauchy Condensation Test

Let (𝑎𝑛 ) be a sequence in ℝ. Suppose that (𝑎𝑛 ) is non-negative and decreasing, that is,

𝑎𝑛 ≥ 𝑎𝑛+1 , ∀𝑛 ∈ ℕ.

They are equivalent:

1. The series
∞
∑ 𝑎𝑛
𝑛=1
converges.

2. The series
∞
∑ 2𝑛 𝑎2𝑛 = 𝑎1 + 2𝑎2 + 8𝑎8 + 16𝑎16 + …
𝑛=0
converges.

Proof
For 𝑘 ∈ ℕ denote the partial sums by

𝑘 𝑘
𝑠𝑘 ∶= ∑ 𝑎𝑛 , 𝑡𝑘 ∶= ∑ 2𝑛 𝑎2𝑛 .
𝑛=1 𝑛=0

Since 𝑎𝑛 ≥ 0, it is immediate to check that the sequences (𝑠𝑘 ) and (𝑡𝑘 ) are increasing.
Part 1. Assume that the series
∞
∑ 2𝑛 𝑎2𝑛
𝑛=0
diverges. Hence the sequence (𝑡𝑘 ) diverges and therefore (𝑡𝑘 ) is not bounded above.

Indeed, suppose (𝑡𝑘 ) was bounded above. Since (𝑡𝑘 ) is increasing, we would conclude that
(𝑡𝑘 ) is convergent by the Monotone Convergence Theorem. Contradiction.

We want to estimate 𝑠𝑘 from below. To this end, we notice that

𝑠2 = 𝑎 1 + 𝑎 2
1
≥ 𝑎1 + 𝑎2
2
1
= (𝑎1 + 2𝑎2 )
2
1
= 𝑡1 ,
2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 242

where we used that 𝑎𝑛 ≥ 0, and so 𝑎1 > 𝑎1 /2. Moreover

𝑠4 = 𝑎1 + 𝑎2 + (𝑎3 + 𝑎4 )
≥ 𝑎1 + 𝑎2 + (𝑎4 + 𝑎4 )
= 𝑎1 + 𝑎2 + 2𝑎4
1
≥ 𝑎1 + 𝑎2 + 2𝑎4
2
1
= (𝑎1 + 2𝑎2 + 4𝑎4 )
2
1
= 𝑡2 ,
2
where we used also that (𝑎𝑛 ) is decreasing. Similar reasoning yields
𝑠8 = 𝑎1 + 𝑎2 + (𝑎3 + 𝑎4 ) + (𝑎5 + 𝑎6 + 𝑎7 + 𝑎8 )
≥ 𝑎1 + 𝑎2 + (𝑎4 + 𝑎4 ) + (𝑎8 + 𝑎8 + 𝑎8 + 𝑎8 )
= 𝑎1 + 𝑎2 + 2𝑎4 + 4𝑎8
1
≥ 𝑎1 + 𝑎2 + 2𝑎4 + 4𝑎8
2
1
= (𝑎1 + 2𝑎2 + 4𝑎4 + 8𝑎8 )
2
1
= 𝑡3 .
2
where again we used that (𝑎𝑛 ) is decreasing, and 𝑎𝑛 ≥ 0. Iterating, we obtain that for all 𝑘 ∈ ℕ it holds:
𝑠2𝑘 = 𝑎1 + 𝑎2 + (𝑎3 + 𝑎4 ) + (𝑎5 + 𝑎6 + 𝑎7 + 𝑎8 )
+ … + (𝑎2𝑘−1 + … + 𝑎2𝑘 )
≥ 𝑎1 + 𝑎2 + (𝑎4 + 𝑎4 ) + (𝑎8 + 𝑎8 + 𝑎8 + 𝑎8 )
+ … + (𝑎2𝑘 + … + 𝑎2𝑘 )
= 𝑎1 + 𝑎2 + 2𝑎4 + 4𝑎8 + … + 2𝑘−1 𝑎2𝑘
1
≥ 𝑎1 + 𝑎2 + 2𝑎4 + 4𝑎8 + … + 2𝑘−1 𝑎2𝑘
2
1
= (𝑎1 + 2𝑎2 + 4𝑎4 + 8𝑎8 + … + 2𝑘 𝑎2𝑘 )
2
1
= 𝑡𝑘 .
2
We have shown that
1
𝑠2𝑘 ≥ 𝑡 , ∀𝑘 ∈ ℕ.
2 𝑘
Since (𝑡𝑘 ) is not bounded above, we infer that (𝑠2𝑘 ) is not bounded above. In particular (𝑠𝑘 ) is not bounded,
and hence divergent. Thus the series
∞
∑ 𝑎𝑛
𝑛=1

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 243

diverges.
Part2. Suppose that the series
∞
∑ 2 𝑛 𝑎2 𝑛
𝑛=0
converges. Hence the sequence (𝑡𝑘 ) converges, and therefore (𝑡𝑘 ) is bounded. This means that there
exists 𝑀 > 0 such that
|𝑡𝑘 | ≤ 𝑀 , ∀ 𝑛 ∈ ℕ .
Since 𝑡𝑘 ≥ 0, the above reads
𝑡𝑘 ≤ 𝑀 , ∀𝑛 ∈ ℕ.
Fix 𝑘 ∈ ℕ and let 𝑚 ∈ ℕ be such that
𝑘 ≤ 2𝑚+1 − 1 .
In this way
𝑠𝑘 ≤ 𝑠2𝑚+1 −1 .
We have

𝑠2𝑚+1 −1 = 𝑎1 + (𝑎2 + 𝑎3 ) + (𝑎4 + 𝑎5 + 𝑎6 + 𝑎7 )

+ … + (𝑎2𝑚 + … + 𝑎2𝑚+1 −1 )
≤ 𝑎1 + (𝑎2 + 𝑎2 ) + (𝑎4 + 𝑎4 + 𝑎4 + 𝑎4 )
+ … + (𝑎2𝑚 + … + 𝑎2𝑚 )
= 𝑎1 + 2𝑎2 + 4𝑎4 + … + 2𝑚 𝑎2𝑚
= 𝑡𝑚 ,

where we used that (𝑎𝑛 ) is decreasing. We have then shown

𝑠𝑘 ≤ 𝑠2𝑚+1 −1 ≤ 𝑡𝑚 ≤ 𝑀 .

Since 𝑀 does not depend on 𝑘, we conclude that

𝑠𝑘 ≤ 𝑀 , ∀𝑘 ∈ ℕ.

As 𝑠𝑘 ≥ 0, we conclude that 𝑠𝑘 is bounded. Recalling that (𝑠𝑘 ) is increasing, by the Monotone Convergence
Theorem we infer that (𝑠𝑘 ) converges. This proves that the series
∞
∑ 𝑎𝑛
𝑛=1

is convergent, ending the proof.

Thanks to the Cauchy Condensation Test of Theorem 0.324 we can prove the following result.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 244

Theorem 8.27: Convergence of 𝑝-series

Let 𝑝 ∈ ℝ. Consider the 𝑝-series

∞
1
∑ 𝑝 .
𝑛=1 𝑛
We have:

1. If 𝑝 > 1 the 𝑝-series converges.

2. If 𝑝 ≤ 1 the 𝑝-series diverges.

Proof
The series in question is
∞
1
∑ 𝑎𝑛 , 𝑎𝑛 ∶= .
𝑛=1 𝑛𝑝
Note that (𝑎𝑛 ) is decreasing and non-negative. Hence, by the Cauchy Condensation Test of Theorem
0.324, the 𝑝-series converges if and only if
∞
∑ 2 𝑛 𝑎2 𝑛
𝑛=0
converges. We have
∞ ∞ ∞
∑ 2𝑛 𝑎2𝑛 = ∑ 2𝑛−𝑛𝑝 = ∑(21−𝑝 )𝑛 ,
𝑛=0 𝑛=0 𝑛=0
and the latter is a Geometric Series of ratio

𝑥 ∶= 21−𝑝 .

By the Geometric Series Test, we have convergence of

∞
∑(21−𝑝 )𝑛
𝑛=0

if and only if
|𝑥| < 1 .
The above is equivalent to

21−𝑝 < 1 = 20 ⟺ 1−𝑝 <0 ⟺ 𝑝 > 1.

Therefore
∞
1
∑ 𝑝
𝑛=1 𝑛
converges if and only if 𝑝 > 1, ending the proof.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 245

The following is another notable application of the Cauchy Condensation Test.

Theorem 8.28
Let 𝑝 ∈ ℝ. Consider the series
∞
1
∑ 𝑝 .
𝑛=2 𝑛 (log 𝑛)
We have:

1. If 𝑝 > 1 the series converges.

2. If 𝑝 ≤ 1 the series diverges.

Proof
The series in question is
∞
1
∑ 𝑎𝑛 , 𝑎𝑛 ∶= 𝑝 .
𝑛=2 𝑛 (log 𝑛)
Note that (𝑎𝑛 ) is non-negative and decreasing. Therefore we can apply the Cauchy Condensation Test to
conclude that the above series is convergent if and only if the series
∞
∑ 2𝑛 𝑎2𝑛
𝑛=1

is convergent. We have
1 1
2𝑛 𝑎2𝑛 = 2𝑛 𝑝 =
2𝑛 (log 2𝑛 ) 𝑛𝑝 log 2
so that
∞ ∞
1 1
∑ 2 𝑛 𝑎2 𝑛 = ∑ .
𝑛=1 log 2 𝑛=1 𝑛𝑝
The latter is a 𝑝-series, which by Theorem 0.325 converges if and only if 𝑝 > 1. Hence
∞
1
∑ 𝑝
𝑛=2 𝑛 (log 𝑛)

converges if and only if 𝑝 > 1, and the proof is concluded.

8.4.2 Comparison Test

Another really useful result to study the convergence of non-negative series is the Comparison Test.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 246

Theorem 8.29: Comparison test

Let (𝑎𝑛 )𝑛∈ℕ and (𝑏𝑛 )𝑛∈ℕ be non-negative sequences. Suppose that there exists 𝑁 ∈ ℕ such that

𝑎𝑛 ≤ 𝑏 𝑛 , ∀𝑛 ≥ 𝑁 .

They hold:
∞ ∞
1. If ∑𝑛=1 𝑏𝑛 converges, then also ∑𝑛=1 𝑎𝑛 converges.
∞ ∞
2. If ∑𝑛=1 𝑎𝑛 diverges, then also ∑𝑛=1 𝑏𝑛 diverges.

Proof
Part 1. Define the partial sums starting at 𝑛 = 𝑁
𝑘 𝑘
𝑠𝑘 ∶= ∑ 𝑎𝑘 , 𝑡𝑘 ∶= ∑ 𝑏𝑘 .
𝑛=𝑁 𝑛=𝑁
Suppose that
∞
∑ 𝑏𝑛
𝑛=1
converges. Hence also the series
∞
∑ 𝑏𝑛
𝑛=𝑁
converges. Then (𝑡𝑘 ) is a convergent sequence, which implies that (𝑡𝑘 ) is bounded, and hence bounded
above. We have that 𝑠𝑘 is bounded above: Indeed, using the assumption, we have
𝑠𝑘 = 𝑎𝑁 + 𝑎𝑁 +1 + … + 𝑎𝑘
≤ 𝑏𝑁 + 𝑏𝑁 +1 + … + 𝑏𝑘
= 𝑡𝑘 ,
which reads
𝑠𝑘 ≤ 𝑡𝑘 , ∀𝑘 ≥ 𝑁 .
Therefore (𝑠𝑘 ) is bounded above, being (𝑡𝑘 ) bounded above. Recall that 𝑠𝑘 is increasing, by Lemma 8.22.
By the Monotone Convergence Theorem we conclude that 𝑠𝑘 is convergent, showing that the series
∞
∑ 𝑎𝑛
𝑛=𝑁
converges. Hence also the series
∞
∑ 𝑎𝑛
𝑛=1
converges, concluding the proof of Point 1.
Part 2. Note that Point 2 is the contrapositive of Point 1, and hence it holds.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 247

Let us give two applications of the Comparison Test.

Example 8.30

• Consider the series

∞
1
∑ . (8.9)
𝑛=1 𝑛2 + 3𝑛 − 1
Since 3𝑛 − 1 ≥ 0 for all 𝑛 ∈ ℕ, we get
1 1
≤ 2, ∀𝑛 ∈ ℕ.
𝑛2 + 3𝑛 − 1 𝑛
By Theorem 0.325 the 𝑝-series
∞
1
∑ 2
𝑛=1 𝑛
converges. Therefore also the series at (8.9) converges by the Comparison Test in Theorem 0.327.

• Consider the series

1
∞ 3𝑛 + 6𝑛 +
∑ 𝑛+1
. (8.10)
𝑛=0 2𝑛
Note that
1
3𝑛 + 6𝑛 +
𝑛+1 3𝑛 3 𝑛
≥ = ( ) , ∀𝑛 ∈ ℕ.
2𝑛 2𝑛 2
Since | 32 | = 3
2
> 1, the series
∞
3 𝑛
∑( )
𝑛=0 2
diverges by the Geometric Series Test in Theorem 0.311. Therefore, by the Comparison Test, also
the series at (8.10) diverges.

8.4.3 Limit Comparison Test

To apply the Comparison Test to the series

∞
∑ 𝑎𝑛 ,
𝑛=1
one needs to find another sequence (𝑏𝑛 ) such that
𝑎𝑛 ≤ 𝑏 𝑛 , ∀𝑛 ≥ 𝑁 ,
or
𝑏𝑛 ≤ 𝑎𝑛 , ∀𝑛 ≥ 𝑁 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 248

This is not always possible. However, one might be able to show that
𝑎
𝐿 = lim 𝑛
𝑛→∞ 𝑏𝑛
for some 𝐿 ∈ ℝ. In this case, the series of (𝑎𝑛 ) and (𝑏𝑛 ) can still be compared, in the sense specified in the
below theorem.

Theorem 8.31: Limit Comparison Test

Let (𝑎𝑛 ) and (𝑏𝑛 ) be sequences such that

𝑎𝑛 ≥ 0 , 𝑏𝑛 > 0 , ∀𝑛 ∈ ℕ.

Suppose there exists 𝐿 ∈ ℝ such that

𝑎𝑛
𝐿 = lim .
𝑛→∞ 𝑏𝑛

They hold:

1. If 0 < 𝐿 < ∞, then

∞ ∞
∑ 𝑎𝑛 converges ⟺ ∑ 𝑏𝑛 converges.
𝑛=1 𝑛=1

2. If 𝐿 = 0, then
∞ ∞
• If ∑𝑛=1 𝑏𝑛 converges also ∑𝑛=1 𝑎𝑛 converges,
∞ ∞
• If ∑𝑛=1 𝑎𝑛 diverges also ∑𝑛=1 𝑏𝑛 diverges.

Proof
Part 1. Suppose that 0 < 𝐿 < 1. Set
𝐿
.𝜀 ∶=
2
Since 𝜀 > 0 and 𝑎𝑛 /𝑏𝑛 → 𝐿, there exists 𝑁 ∈ ℕ such that
𝑎𝑛
| − 𝐿| < 𝜀 , ∀𝑛 ≥ 𝑁 .
𝑏𝑛
The above is equivalent to
𝑎𝑛
𝐿−𝜀 < < 𝜀 + 𝐿, ∀𝑛 ≥ 𝑁 .
𝑏𝑛
Since 𝜀 = 𝐿/2, we get
𝐿 𝑎𝑛 3𝐿
< < , ∀𝑛 ≥ 𝑁 ,
2 𝑏𝑛 2
or equivalently,
𝐿 3𝐿
𝑏 < 𝑎𝑛 < 𝑏 , ∀𝑛 ≥ 𝑁 .
2 𝑛 2 𝑛
We are now ready to prove the main claim:

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 249

• Suppose that
∞
∑ 𝑎𝑛
𝑛=1
converges. Then also
∞
∑ 𝑎𝑛 ,
𝑛=𝑁
since we are only discarding a finite number of terms. As
𝐿
𝑏 ≤ 𝑎𝑛 , ∀𝑛 ≥ 𝑁 ,
2 𝑛
it follows from the Comparison Test in Theorem 0.327 that the series
∞
𝐿
∑ 𝑏 .
𝑛=𝑁 2 𝑛

converges. Since 𝐿/2 is a constant, we also conclude that

∞
∑ 𝑏𝑛
𝑛=1

converges.

• Suppose that
∞
∑ 𝑏𝑛
𝑛=1
converges. Then also
∞
3𝐿
∑ 𝑏
𝑛=𝑁 2 𝑛
converges. Since
3𝐿
𝑎𝑛 < 𝑏 , ∀𝑛 ≥ 𝑁 ,
2 𝑛
by the Comparison Test we infer that
∞
∑ 𝑎𝑛
𝑛=𝑁
converges. Therefore, also
∞
∑ 𝑎𝑛
𝑛=1
converges.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 250

Part 2. Suppose that 𝐿 = 0. Note that the second bullet point is the contrapositive of the first. Hence we
only need to show the first bullet point. Let 𝜀 = 1. Since 𝑎𝑛 /𝑏𝑛 → 0, there exists 𝑁 ∈ ℕ such that
𝑎𝑛
| − 0| < 𝜀 = 1 , ∀𝑛 ≥ 𝑁 .
𝑏𝑛
Therefore
𝑎𝑛 < 𝑏 𝑛 , ∀𝑛 ≥ 𝑁 .
The thesis follows immediately by the Comparison Test in Theorem 0.327.

Important

It might happen that

𝑎𝑛
lim = ∞.
𝑛→∞ 𝑏𝑛
In this case it can be helpful to interchange the roles of 𝑎𝑛 and 𝑏𝑛 , since then

𝑏𝑛
lim = 0.
𝑛→∞ 𝑎𝑛

Let us give a few applications of the Limit Comparison Tets.

Example 8.32

The following series converges

∞
2𝑛3 + 5𝑛 + 1
∑ 6
.
𝑛=1 7𝑛 + 2𝑛 + 5

Proof . Set
2𝑛3 + 5𝑛 + 1 1
𝑎𝑛 ∶= , 𝑏𝑛 ∶= .
7𝑛6 + 2𝑛 + 5 𝑛3
We have
𝑎𝑛
𝐿 ∶= lim
𝑛→∞ 𝑏𝑛
2𝑛3 + 5𝑛 + 1 1
= lim 6
𝑛→∞ 7𝑛 + 2𝑛 + 5 / 𝑛3
2𝑛6 + 5𝑛4 + 𝑛3
= lim
𝑛→∞ 7𝑛6 + 2𝑛 + 5
5 1
2+ 2 + 3
𝑛 𝑛 2
= lim = .
𝑛→∞ 2 5 7
7+ 5 + 6
𝑛 𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 251

The series
∞
1
∑ 3
𝑛=1 𝑛
2
converges, being a 𝑝-series with 𝑝 = 3 > 1. Since 𝐿 = 7
> 0, also the series of interest
converges, by the Limit Comparison Test.

Example 8.33

Prove that the following series diverges

∞
𝑛 + cos(𝑛)
∑ 2
.
𝑛=1 𝑛

Proof. We expect the terms in the series to behave like 1/𝑛 for large 𝑛. Hence we set

𝑛 + cos(𝑛) 1
𝑎𝑛 ∶= , 𝑏𝑛 = .
𝑛2 𝑛
We compute
𝑎𝑛
𝐿 ∶=
𝑏𝑛
𝑛 + cos(𝑛) 1
= lim
𝑛→∞ 𝑛2 /𝑛
𝑛2 + 𝑛 cos(𝑛)
= lim
𝑛→∞ 𝑛2
cos(𝑛)
= lim (1 + )
𝑛→∞ 𝑛
Since
−1 ≤ cos(𝑛) ≤ 1 ,
we obtain
1 cos(𝑛) 1
− ≤ ≤ .
𝑛 𝑛 𝑛
As both − 𝑛1 → 0 and 1
𝑛
→ 0, by the Squeeze Theorem

cos(𝑛)
⟶ 0.
𝑛
Hence
cos(𝑛)
𝐿 = lim (1 + ) = 1.
𝑛→∞ 𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 252

The harmonic series

∞
1
∑
𝑛=1 𝑛
does not converge. Since 𝐿 = 1 > 0, the series of interest diverges by the Limit Comparison
Test.

Example 8.34

Prove that the series

∞
1
∑ (1 − cos ( ))
𝑛=1 𝑛
converges.

Solution. Since
1
cos ( ) ≤ 1 ,
𝑛
the above is a non-negative series. Recall the limit
1 − cos(𝑎𝑛 ) 1
lim = ,
𝑛→∞ (𝑎𝑛 )2 2
where (𝑎𝑛 ) is a sequence in ℝ such that 𝑎𝑛 → 0 and

𝑎𝑛 ≠ 0 ∀𝑛 ∈ ℕ.

In particular, for 𝑎𝑛 = 1/𝑛, we obtain

1 1
lim 𝑛2 (1 − cos ( )) = .
𝑛→∞ 𝑛 2
Set
1 1
𝑏𝑛 ∶= 1 − cos ( ) , 𝑐𝑛 ∶= .
𝑛 𝑛2
We have
𝑏𝑛
𝐿 ∶= lim
𝑛→∞ 𝑐𝑛
1
= lim 𝑛2 (1 − cos ( ))
𝑛→∞ 𝑛
1
= .
2
Note that the series
∞
1
∑ 2
𝑛=1 𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 253

converges, being a 𝑝-series with 𝑝 > 2. Therefore, since 𝐿 = 1/2 > 0, also the series
∞
1
∑ (1 − cos ( ))
𝑛=1 𝑛

converges, by the Limit Comparison Test.

Sometimes the Limit Comparison Test fails, but the Comparison Test works.

Example 8.35

Consider the series

∞
1 + sin(𝑛)
∑ .
𝑛=1 𝑛2
Since
sin(𝑛) ≥ −1 ,
the above is a non-negative series. We expect this series to behave similar to
∞
1
∑ 2
.
𝑛=1 𝑛

However
1 + sin(𝑛) 1
= 1 + sin(𝑛)
𝑛 2 / 𝑛2
does not converge. Hence, we cannot use the Limit Comparison Test. In alternative, we note that

1 + sin(𝑛) 2
2
≤ 2, ∀𝑛 ∈ ℕ.
𝑛 𝑛
The series
∞
2
∑ 2
𝑛=1 𝑛
converges, being a 𝑝-series with 𝑝 = 2 > 1. Therefore also
∞
1 + sin(𝑛)
∑
𝑛=1 𝑛2

converges, by the Comparison Test of Theorem 0.327.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 254

8.4.4 Ratio Test for positive series

The Ratio Test can be generalized to series. Notice that in this case the terms of the series need to be positive.

Theorem 8.36: Ratio Test for positive series

Let (𝑎𝑛 ) be a sequence in ℝ such that

𝑎𝑛 > 0 , ∀𝑛 ∈ ℕ.

1. Suppose that the following limit exists:

𝑎𝑛+1
𝐿 ∶= lim .
𝑛→∞ 𝑎𝑛

They hold:
∞
• If 𝐿 < 1 then ∑𝑛=1 𝑎𝑛 converges.
∞
• If 𝐿 > 1 then ∑𝑛=1 𝑎𝑛 diverges.

2. Suppose that there exists 𝑁 ∈ ℕ and 𝐿 > 1 such that

𝑎𝑛+1
≥ 𝐿, ∀𝑛 ≥ 𝑁 .
𝑎𝑛
∞
Then the series ∑𝑛=1 𝑎𝑛 diverges.

Proof
Part 1. Let 𝑎𝑛+1
𝐿 ∶= lim .
𝑛→∞ 𝑎𝑛

• Suppose that 𝐿 < 1. Therefore there exists 𝑟 such that

𝐿 < 𝑟 < 1.

Define
𝜀 ∶= 𝑟 − 𝐿 ,
so that 𝜀 > 0. By the convergence 𝑎𝑛+1 /𝑎𝑛 → 𝐿 there exists 𝑁 ∈ ℕ such that
𝑎𝑛+1
| − 𝐿| < 𝜀 = 𝑟 − 𝐿 , ∀𝑛 ≥ 𝑁 .
𝑎𝑛
In particular
𝑎𝑛+1
− 𝐿 < 𝑟 − 𝐿, ∀𝑛 ≥ 𝑁 ,
𝑎𝑛
which implies
𝑎𝑛+1 < 𝑟 𝑎𝑛 , ∀𝑛 ≥ 𝑁 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 255

Applying 𝑛 − 𝑁 times the above estimate we get

0 < 𝑎𝑛 < 𝑟 𝑎𝑛−1 < … < 𝑟 𝑛−𝑁 𝑎𝑁 , ∀𝑛 ≥ 𝑁 . (8.11)

Note that the series of 𝑟 𝑛−𝑁 𝑎𝑁 converges, since

∞ ∞
1
∑ 𝑟 𝑛−𝑁 𝑎𝑁 = 𝑎 𝑁 ∑ 𝑟 𝑘 = 𝑎 𝑁 ,
𝑛=𝑁 𝑘=0
1−𝑟

∞
where the last equality follows because ∑𝑘=0 𝑟 𝑘 is a geometric series and 0 < 𝑟 < 1. Since (8.11)
holds, by the Comparison Test in Theorem 0.327 we conclude that the series
∞
∑ 𝑎𝑛
𝑛=𝑁

converges. Therefore also the series

∞
∑ 𝑎𝑛
𝑛=1
converges, ending the proof in the case 𝐿 < 1.

• Suppose 𝐿 > 1: Then, by the Ratio Test for sequences, it follows that 𝑎𝑛 diverges. Therefore

lim 𝑎𝑛 ≠ 0 ,
𝑛→∞

∞
and the series ∑𝑛=1 diverges by the Necessary Condition, see Theorem 0.304.

Part 2. Suppose there exists 𝐿 > 1 and 𝑁 ∈ ℕ such that

𝑎𝑛+1
≥ 𝐿, ∀𝑛 ≥ 𝑁 .
𝑎𝑛
∞
By the Ratio Test for sequences it follows that 𝑎𝑛 diverges. Therefore ∑𝑛=1 diverges by the Necessary
Condition.

Example 8.37

Consider the series

∞
(𝑛!)2
∑
𝑛=1 (2𝑛)!
Let
(𝑛!)2
𝑎𝑛 =
(2𝑛)!

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 256

and compute

𝑎𝑛+1 ((𝑛 + 1)!)2 (𝑛!)2

=
𝑎𝑛 (2(𝑛 + 1))! / (2𝑛)!
((𝑛 + 1)!)2 (2𝑛)!
=
(𝑛!)2 (2(𝑛 + 1))!
2
(𝑛 + 1)! (2𝑛)!
=( )
𝑛! (2𝑛 + 2)!
(𝑛 + 1) 2
=
(2𝑛 + 2)(2𝑛 + 1)
1 2
(1 + )
𝑛
=
2 1
(2 + ) (2 + )
𝑛 𝑛
Therefore
1 2
𝑎𝑛+1 (1 + )
𝑛 1
𝐿 = lim = lim = .
𝑛→∞ 𝑎𝑛 2 1
(2 + ) (2 + ) 4
𝑛→∞
𝑛 𝑛
Since 𝐿 = 1/4 < 1, be the Ratio Test for Series we conclude that the series of interest converges.

Important

Like with the Ratio Test for sequences, the case 𝐿 = 1 is not covered by Theorem 0.334. This is because,
in this case, the series
∞
∑ 𝑎𝑛
𝑛=1
might be convergent or divergent, as shown in the next example.

Example 8.38

• Consider the series

∞
1
∑ .
𝑛=1 𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 257

Setting 𝑎𝑛 = 1/𝑛, we have

𝑎𝑛+1
𝐿 = lim
𝑛→∞ 𝑎𝑛
1 1
= lim
𝑛→∞ 𝑛 + 1 / 𝑛
𝑛
= lim = 1.
𝑛→∞ 𝑛 + 1

Therefore 𝐿 = 1 and we cannot apply the Ratio Test. However the series in question diverges,
being the harmonic series.

• Consider the series

∞
1
∑ 2
.
𝑛=1 𝑛

Setting 𝑎𝑛 = 1/𝑛2 , we have

𝑎𝑛+1
𝐿 = lim
𝑛→∞ 𝑎𝑛
1 1
= lim
𝑛→∞ (𝑛 + 1) / 𝑛2
2

𝑛2
= lim 2 = 1.
𝑛→∞ 𝑛 + 2𝑛 + 1

Therefore 𝐿 = 1 and we cannot apply the Ratio Test. However the series in question diverges,
being a 𝑝-series with 𝑝 = 2 > 1.

The Ratio Test can often be combined with other convergence tests, as seen in the following example.

Example 8.39

Using the Cauchy Condensation Test and the Ratio Test, prove that the series below converges
∞
log(𝑛)
∑ 2
.
𝑛=1 𝑛

Solution. With 𝑎𝑛 = log 𝑛/𝑛2 , we have

∞ ∞ ∞
log(2𝑛 ) 𝑛
∑ 2𝑛 𝑎2𝑛 = ∑ 2𝑛 2
= log(2) ∑ 𝑛
𝑛=0 𝑛=2 (2𝑛 ) 𝑛=2 2
This converges according to the Ratio Test: With 𝑏𝑛 = 𝑛/2𝑛 , we have
𝑏𝑛+1 𝑛 + 1 𝑛 𝑛+1 1
= 𝑛+1 𝑛 = ⟶ < 1.
𝑏𝑛 2 /2 2𝑛 2

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 258

∞ log 𝑛
Hence ∑𝑛=1 𝑛2
converges by the Cauchy Condensation Test.

8.5 General series

In the previous section we presented several tests for non-negative series. For non-negative series we shows
that the partial sums (𝑠𝑘 ) are increasing, see Lemma 8.22. This makes non-negative terms series relatively
easy to study.
When a series contains both positive and negative terms, the partial sums (𝑠𝑘 ) might oscillate, making the
series harder to study. In this section we present some tests for general series in ℂ.

8.5.1 Absolute Convergence Test

To study general series, we introduce a stronger notions of convergence, known as absolute convergence.

Definition 8.40: Absolute convergence

∞
Let (𝑎𝑛 ) be a sequence in ℂ. The series ∑𝑛=1 𝑎𝑛 is said to converge absolutely if the non-negative series
∞
∑ |𝑎𝑛 |
𝑛=1

converges.

Let us show that absolute convergence implies convergence.

Theorem 8.41: Absolute Convergence Test

∞
Let (𝑎𝑛 ) be a sequence in ℂ. If the series ∑𝑛=1 𝑎𝑛 converge absolutely, then the series converges.

Before proceeding with the proof, we introduce some notation.

Notation 8.42: Positive and negative part

For a number 𝑥 ∈ ℝ, we define

𝑥 if 𝑥 ≥ 0 −𝑥 if 𝑥 < 0
𝑥 + ∶= { 𝑥 − ∶= {
0 if 𝑥 < 0 0 if 𝑥 ≥ 0
These are called the positive part and negative part of 𝑥, respectively. Note that
𝑥+ ≥ 0 , 𝑥− ≥ 0 .

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 259

Moreover they hold

𝑥 = 𝑥+ − 𝑥− , |𝑥| = 𝑥 + + 𝑥 − .
The above relations are easy to check, and the proof is omitted. In particular, it holds that

𝑥 + ≤ |𝑥| , 𝑥 − ≤ |𝑥| .

Proof: Proof of Theorem 0.339

Part 1. Suppose first that (𝑎𝑛 ) is a sequence in ℝ such that

∞
∑ |𝑎𝑛 |
𝑛=1

converges. Since
0 ≤ 𝑎𝑛+ ≤ |𝑎𝑛 | , ∀𝑛 ∈ ℕ,
we can use the Comparison Test for non-negative series (Theorem 0.327) and conclude that the series
∞
∑ 𝑎𝑛+ .
𝑛=1

converges. Similarly, we have

0 ≤ 𝑎𝑛− ≤ |𝑎𝑛 | , ∀𝑛 ∈ ℕ.
Therefore also the series
∞
∑ 𝑎𝑛−
𝑛=1
converges by the Comparison Test. Since
𝑎𝑛 = 𝑎𝑛+ − 𝑎𝑛− , ∀𝑛 ∈ ℕ,
we can use the Algebra of Limits for series (Theorem 0.316) to conclude that
∞ ∞ ∞ ∞
∑ 𝑎𝑛 = ∑ (𝑎𝑛+ − 𝑎𝑛− ) = ∑ 𝑎𝑛+ − ∑ 𝑎𝑛−
𝑛=1 𝑛=1 𝑛=1 𝑛=1
converges.
Part 2. Suppose now that (𝑎𝑛 ) is a sequence in ℂ such that
∞
∑ |𝑎𝑛 |
𝑛=1

converges. Let 𝑥𝑛 , 𝑦𝑛 ∈ ℝ denote the real and imaginary part of 𝑎𝑛 . Therefore

|𝑥𝑛 | = √𝑥𝑛2 ≤ √𝑥𝑛2 + 𝑦𝑛2 = |𝑎𝑛 | , ∀𝑛 ∈ ℕ.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 260

Therefore the series

∞
∑ |𝑥𝑛 |
𝑛=1
converges by the Comparison Test for non-negative series (Theorem 0.327). Since (𝑥𝑛 ) is a real sequence,
from Part 1 of the proof we have that the series
∞
∑ 𝑥𝑛
𝑛=1

converges. Arguing in the same way for the imaginary part 𝑦𝑛 we conclude that also
∞
∑ 𝑦𝑛
𝑛=1

converges. Finally, by the Algebra of Limits in ℂ, we get

∞ ∞ ∞ ∞
∑ 𝑎𝑛 = ∑ (𝑥𝑛 + 𝑖𝑦𝑛 ) = ∑ 𝑥𝑛 + 𝑖 ∑ 𝑦𝑛 ,
𝑛=1 𝑛=1 𝑛=1 𝑛=1

∞
proving that ∑𝑛=1 𝑎𝑛 converges.

Example 8.43

The series
∞
1
∑(−1)𝑛
𝑛=1 𝑛
does not converge absolutely, since
∞ ∞
1 1
∑ |(−1)𝑛 | = ∑
𝑛=1 𝑛 𝑛=1 𝑛
doesn’t converge, being the harmonic series.

Example 8.44

Prove that the following series converges

∞
𝑛2 − 5𝑛 + 2
∑(−1)𝑛 .
𝑛=1 𝑛4

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 261

Solution. We have

𝑛2 − 5𝑛 + 2 |𝑛2 − 5𝑛 + 2| 𝑛2 + 5𝑛 + 2
|(−1)𝑛 | = ≤
𝑛4 𝑛4 𝑛4
where we used the triangle inequality. Note that

𝑛2 + 5𝑛 + 2 1 𝑛4 + 5𝑛3 + 2𝑛2
=
𝑛4 / 𝑛2 𝑛4
5 2
=1+ + 2 ⟶1
𝑛 𝑛
The series
∞
1
∑ 2
𝑛=1 𝑛
converges, being a 𝑝-series with 𝑝 = 2. Hence, also
∞
𝑛2 + 5𝑛 + 2
∑
𝑛=1 𝑛4

converges by the Limit Comparison Test for non-negative series (Theorem 0.329). Since

𝑛 2 − 5𝑛 + 2 |𝑛2 − 5𝑛 + 2|
0≤ |(−1)𝑛 |=
𝑛4 𝑛4
𝑛2 + 5𝑛 + 2
≤ ,
𝑛4
the series
∞
𝑛2 − 5𝑛 + 2
∑ |(−1)𝑛 |
𝑛=1 𝑛4
converges by the Comparison Test for non-negative series (Theorem 0.327). This shows the
series
∞
𝑛2 − 5𝑛 + 2
∑(−1)𝑛
𝑛=1 𝑛4
converges by the Absolute Convergence Test.

8.5.2 Ratio Test for general series

As an application of the Absolute Convergence Test we obtain the Ratio Test for general series.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 262

Theorem 8.45: Ratio Test for general series

Let (𝑎𝑛 ) be a sequence in ℂ, such that

𝑎𝑛 ≠ 0 ∀𝑛 ∈ ℕ.

1. Suppose that the following limit exists:

𝑎𝑛+1
𝐿 ∶= lim | |.
𝑛→∞ 𝑎𝑛
They hold:
∞
• If 𝐿 < 1 then ∑𝑛=1 𝑎𝑛 converges absolutely, and hence converges.
∞
• If 𝐿 > 1 then ∑𝑛=1 𝑎𝑛 diverges.

2. Suppose that there exists 𝑁 ∈ ℕ and 𝐿 > 1 such that

𝑎𝑛+1
| | ≥ 𝐿, ∀𝑛 ≥ 𝑁 .
𝑎𝑛
∞
Then the series ∑𝑛=1 𝑎𝑛 diverges.

Proof
Part 1. Let
𝑏𝑛 ∶= |𝑎𝑛 | ,
so that
𝑎𝑛+1 𝑏𝑛+1
𝐿 ∶= lim | | = lim .
𝑛→∞ 𝑎𝑛 𝑛→∞ 𝑏𝑛

• Suppose that 𝐿 < 1. Since (𝑏𝑛 ) is a sequence with non-negative terms, we have that
∞
∑ 𝑏𝑛
𝑛=1

converges by the Ratio Test for non-negative series, see Theorem 0.334. Since, by definition
∞ ∞
∑ 𝑏𝑛 = ∑ |𝑎𝑛 |
𝑛=1 𝑛=1
∞ ∞
also the latter series converges, i.e., ∑𝑛=1 𝑎𝑛 converges absolutely. In particular ∑𝑛=1 𝑎𝑛 converges,
by the Absolute Convergence Test in Theorem 0.339.
• Suppose that 𝐿 > 1. Then the sequence (𝑎𝑛 ) diverges by the Ratio Test for sequences. Hence the
series
∞
∑ 𝑎𝑛
𝑛=1

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 263

diverges by the Necessary Condition in Theorem 0.304.

Part 2. If there exists 𝑁 ∈ ℕ and 𝐿 > 1 such that

𝑎𝑛+1
| | ≥ 𝐿, ∀𝑛 ≥ 𝑁 ,
𝑎𝑛

then the sequence (𝑎𝑛 ) diverges by the Ratio Test for sequences, and we conclude as above.

Example 8.46

Prove that the below series converges

∞
(4 − 3𝑖)𝑛
∑
𝑛=1 (𝑛 + 1)!

Solution. Set
(4 − 3𝑖)𝑛
𝑎𝑛 ∶= .
(𝑛 + 1)!
Then
𝑎𝑛+1
𝐿 ∶= lim | |
𝑛→∞ 𝑎𝑛
𝑎𝑛+1
=| |
𝑎𝑛
(4 − 3𝑖)𝑛+1 (4 − 3𝑖)𝑛
=| |
((𝑛 + 1) + 1)! / (𝑛 + 1)!
|4 − 3𝑖|𝑛+1 (𝑛 + 1)!
=
|4 − 3𝑖|𝑛 (𝑛 + 2)!
5
=
𝑛+2
=0
∞
By the Ratio Test we conclude that the series ∑𝑛=1 𝑎𝑛 converges absolutely, and hence con-
verges.

8.5.3 Exponential function and Euler’s Number

We have already encountered the exponential function in the Euler’s identity

𝑒 𝑖𝜃 = cos(𝜃) + 𝑖 sin(𝜃) , 𝜃 ∈ ℝ.
Using the Ratio Test for general series, we can give a precise definition for 𝑒 𝑧 with 𝑧 ∈ ℂ. We start by studying
convergence of the exponential series.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 264

Theorem 8.47: Exponential series

Let 𝑧 ∈ ℂ. The exponential series

∞
𝑧𝑛
∑
𝑛=0 𝑛!
converges absolutely.

Proof
Set
𝑧𝑛
𝑎𝑛 = .
𝑛!
Then
𝑎𝑛+1
𝐿 = lim | |
𝑛→∞ 𝑎𝑛
𝑧 𝑛+1 𝑧𝑛
= lim | |
𝑛→∞ (𝑛 + 1)! / 𝑛!
|𝑧|𝑛+1 𝑛!
= lim
𝑛→∞ |𝑧|𝑛 (𝑛 + 1)!
|𝑧|
= lim
𝑛→∞ 𝑛 + 1
1
= |𝑧| ⋅ lim = 0.
𝑛→∞ 𝑛 + 1

Therefore the series converges absolutely by the Ratio Test in Theorem 0.343.

Definition 8.48: Exponential function

Define the exponential function

∞
𝑧𝑛
exp ∶ ℂ → ℂ , exp(𝑧) ∶= ∑
𝑛=0 𝑛!

for 𝑧 ∈ ℂ. We denote
𝑒 𝑧 ∶= exp(𝑧) , 𝑒 ∶= 𝑒 1 .

Remark 8.49

1. Using the definition of 𝑒 𝑧 , one can show the usual properties of exponentials, such that

d 𝑥
𝑒 𝑥+𝑦 = 𝑒 𝑥 𝑒 𝑦 , 𝑒 = 𝑒𝑥 .
d𝑥

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 265

2. We had defined
1 𝑛
𝑒 ∶= lim (1 + ) .
𝑛→∞ 𝑛
Using the binomial theorem one can prove that
∞
1 𝑛 1
lim (1 + ) = ∑ .
𝑛→∞ 𝑛 𝑛=0 𝑛!

8.5.4 Conditional convergence

Some series do not converge absolutely, but still converge. Such series are said to converge conditionally.

Definition 8.50: Conditional convergence

Let (𝑎𝑛 ) be a sequence in ℂ. We say that the series

∞
∑ 𝑎𝑛
𝑛=1

converges conditionally if it converges, but it does not converge absolutely.

In practice conditional convergence means that the convergence of the series depends on the order in which
we perform the summation. Changing the order of summation of a series is called rearrangement.

Definition 8.51: Rearrangement of a series

Let (𝑎𝑛 ) be a sequence in ℂ. We define:

• A permutation is a bijection 𝜎 ∶ ℕ → ℕ.
∞
• A rearrangement of the series ∑𝑛=1 𝑎𝑛 is a series
∞
∑ 𝑎𝜎(𝑛)
𝑛=1

for some permutation 𝜎.

If a series of complex numebers converges absolutely, then all its rearrangements converge to the same
limit.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 266

Theorem 8.52

Let (𝑎𝑛 ) be a sequence in ℂ such that

∞
∑ |𝑎𝑛 |
𝑛=1
converges. For any permutation 𝜎 we have
∞ ∞
∑ 𝑎𝜎(𝑛) = ∑ 𝑎𝑛 .
𝑛=1 𝑛=1

For a proof, see Theorem 3.55 in [3]. A very surprising result is the following: If a series of real numbers
converges conditionally, then the series can be rearranged to converge to any real number.

Theorem 8.53: Riemann rearrangement Theorem

Let (𝑎𝑛 ) be a real series such that

∞
∑ 𝑎𝑛
𝑛=1
converges conditionally. Let
𝐿 ∈ ℝ or 𝐿 = ±∞ .
∞
There exists a permutation 𝜎 such that the corresponding rearrangement ∑𝑛=1 𝑎𝜎(𝑛) converges condition-
ally to 𝐿, that is,
∞
∑ 𝑎𝜎(𝑛) = 𝐿 .
𝑛=1

For a proof, we refer the reader to Theorem 3.54 in [3].

8.5.5 Dirichlet and Alternating Series Tests

There are very few conditional convergence tests available. We present the Dirichlet Test and the Alternating
Series Test.

Theorem 8.54: Dirichlet Test

Let (𝑐𝑛 ) be a sequence in ℂ and (𝑞𝑛 ) a sequence in ℝ. Suppose that

• 𝑞𝑛 is decreasing,
• 𝑞𝑛 → 0,
• 𝑞𝑛 ≥ 0 for all 𝑛 ∈ ℕ.

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 267

• Suppose there exists 𝑀 > 0 such that

𝑘
|∑ 𝑐𝑛 | ≤ 𝑀 , ∀𝑘 ∈ ℕ.
𝑛=1

Then the following series converges

∞
∑ 𝑐𝑛 𝑞𝑛
𝑛=1

Proof
Define the partial sums
𝑘
𝑠𝑘 ∶= ∑ 𝑐𝑛 , ∀𝑘 ∈ ℕ.
𝑛=1
By assumption it holds
|𝑠𝑘 | ≤ 𝑀 , ∀𝑘 ∈ ℕ. (8.12)
Note that

𝑐1 = 𝑠 1
𝑐2 = 𝑠 2 − 𝑠 1
……
𝑐𝑛 = 𝑠𝑛 − 𝑠𝑛−1 .

Therefore
𝑘
∑ 𝑐𝑘 𝑞𝑘 = 𝑐1 𝑞1 + 𝑐2 𝑞2 + … + 𝑐𝑘 𝑞𝑘
𝑛=1
= 𝑠1 𝑞1 + (𝑠2 − 𝑠1 )𝑞2 + … + (𝑠𝑘 − 𝑠𝑘−1 )𝑞𝑘
= 𝑠1 (𝑞1 − 𝑞2 ) + 𝑠2 (𝑞2 − 𝑞3 ) + … + 𝑠𝑘−1 (𝑞𝑘−1 − 𝑞𝑘 ) + 𝑠𝑘 𝑞𝑘
𝑘−1
= (∑ 𝑠𝑛 (𝑞𝑛 − 𝑞𝑛+1 )) + 𝑠𝑘 𝑞𝑘
𝑛=1

Since (𝑠𝑘 ) is bounded and 𝑞𝑘 → 0, we conclude that

𝑠𝑘 𝑞𝑘 → 0 .

Further, notice that

𝑞𝑛 − 𝑞𝑛+1 ≥ 0 ,

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 268

since 𝑞𝑛 is decreasing. Therefore, using (8.12), we get

|𝑠𝑛 ||𝑞𝑛 − 𝑞𝑛+1 | = |𝑠𝑛 |(𝑞𝑛 − 𝑞𝑛+1 )

≤ 𝑀(𝑞𝑛 − 𝑞𝑛+1 )

for all 𝑛 ∈ ℕ. Note that

𝑘−1
∑ 𝑀(𝑞𝑛 − 𝑞𝑛+1 ) = 𝑀(𝑞1 − 𝑞𝑘 ) .
𝑛=1
Since 𝑞𝑘 → 0 as 𝑘 → ∞, we conclude that
∞
∑ 𝑀(𝑞𝑛 − 𝑞𝑛+1 ) = 𝑀𝑞1 .
𝑛=1

Hence, by the Comparison Test for non-negative series, we infer that

∞
∑ |𝑠𝑛 ||𝑞𝑛 − 𝑞𝑛+1 | .
𝑛=1

In particular the series

∞
∑ 𝑠𝑛 (𝑞𝑛 − 𝑞𝑛+1 )
𝑛=1
converges by the Absolute Convergence Test. Since we have shown

𝑘 𝑘−1
∑ 𝑐𝑘 𝑞𝑘 = (∑ 𝑠𝑛 (𝑞𝑛 − 𝑞𝑛+1 )) + 𝑠𝑘 𝑞𝑘
𝑛=1 𝑛=1

and 𝑠𝑘 𝑞𝑘 → 0, we conclude that

∞
∑ 𝑐𝑘 𝑞𝑘
𝑛=1
converges.

Example 8.55

Let 𝜃 ∈ ℝ, with
𝜃 ≠ 2𝑘𝜋 , ∀𝑘 ∈ ℤ.
Prove that the below series are conditionally convergent
∞ ∞ ∞
𝑒 𝑖𝜃𝑛 cos(𝜃𝑛) sin(𝜃𝑛)
∑ , ∑ , ∑ .
𝑛=1 𝑛 𝑛=1 𝑛 𝑛=1 𝑛

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 269

Solution. Clearly the series

∞
𝑒 𝑖𝜃𝑛
∑
𝑛=1 𝑛
does not converge absolutely, since
∞ ∞
𝑒 𝑖𝜃𝑛 1
∑| |=∑
𝑛=1 𝑛 𝑛=1 𝑛
∞ 𝑖𝜃𝑛
diverges, being the Harmonic Series. Let us prove convergence of ∑𝑛=1 𝑒 𝑛 . Define the
sequences
1
𝑐𝑛 ∶= 𝑒 𝑖𝜃𝑛 , 𝑞𝑛 ∶= .
𝑛
We have that 𝑞𝑛 is decreasing, 𝑞𝑛 → 0 and 𝑞𝑛 ≥ 0. Let us prove that there exists 𝑀 > 0 such
that
𝑘
|∑ 𝑒 𝑖𝜃𝑛 | ≤ 𝑀 , ∀𝑘 ∈ ℕ. (8.13)
𝑛=1
Note that
1 − 𝑒 𝑖𝜃 ≠ 0 ,
since 𝜃 ≠ 2𝑘𝜋 for all 𝑘 ∈ ℤ. Therefore we can use the Geometric Series (truncated) summa-
tion formula to get
𝑘 𝑘
∑ 𝑒 𝑖𝜃𝑛 = ∑ 𝑒 𝑖𝜃𝑛
𝑛=1 𝑛=1
1 − 𝑒 𝑖(𝑘+1)𝜃
= −1
1 − 𝑒 𝑖𝜃
1 − 𝑒 𝑖𝑘𝜃
= 𝑒 𝑖𝜃
1 − 𝑒 𝑖𝜃
Thus
𝑘
1 − 𝑒 𝑖𝑘𝜃
|∑ 𝑒 𝑖𝜃𝑛 | = |𝑒 𝑖𝜃 |
𝑛=1 1 − 𝑒 𝑖𝜃
1 − 𝑒 𝑖𝑘𝜃
= |𝑒 𝑖𝜃 | | |
1 − 𝑒 𝑖𝜃
|1 − 𝑒 𝑖𝑘𝜃 |
=
|1 − 𝑒 𝑖𝜃 |
|1| + |𝑒 𝑖𝑘𝜃 |
≤
|1 − 𝑒 𝑖𝜃 |
|1|
=
|1 − 𝑒 𝑖𝜃 |

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 270

where we used the triangle inequality. Since the right hand side does not depend on 𝑘, we
can set
|1|
𝑀 ∶= ,
|1 − 𝑒 𝑖𝜃 |
so that (8.13) holds. Therefore
∞
𝑒 𝑖𝜃𝑛
∑
𝑛=1 𝑛
converges by the Dirichlet Test. Recalling the Euler’s Identity

𝑒 𝑖𝜃 = cos(𝜃) + 𝑖 sin(𝜃) ,

we obtain that also the series

∞ ∞
cos(𝑛𝜃) sin(𝑛𝜃)
∑ , ∑
𝑛=1 𝑛 𝑛=1 𝑛

converge.

As a corollary of the Dirichlet Test we obtain the Alternate Convergence Test.

Theorem 8.56: Alternate Convergence Test

Let (𝑞𝑛 ) be a sequence in ℝ such that

• 𝑞𝑛 is decreasing,
• 𝑞𝑛 → 0,
• 𝑞𝑛 ≥ 0 for all 𝑛 ∈ ℕ.

The following series converges

∞
∑(−1)𝑛 𝑞𝑛
𝑛=1

Proof
Define the sequence
𝑐𝑛 ∶= (−1)𝑛 .
Then
𝑘
0 if 𝑘 even
∑ 𝑐𝑛 = {
𝑛=1 −1 if 𝑘 odd

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 271

Hence
𝑘
|∑ 𝑐𝑛 | ≤ 1 , ∀𝑘 ∈ ℕ.
𝑛=1
By the Dirichlet Test we have convergence of
∞ ∞
∑ 𝑐𝑛 𝑞𝑛 = ∑(−1)𝑛 𝑞𝑛 .
𝑛=1 𝑛=1

Example 8.57

Prove that the below series converges conditionally

∞
1
∑(−1)𝑛 .
𝑛=1 𝑛

Solution. The series does not converge absolutely, since

∞ ∞
1 1
∑ |(−1)𝑛 | =∑
𝑛=1 𝑛 𝑛=1 𝑛

diverges, being the Harmonic Series. We need to show convergence of

∞
1
∑(−1)𝑛 .
𝑛=1 𝑛

To this end, set

1
𝑞𝑛 ∶= .
𝑛
Clearly 𝑞𝑛 ≥ 0, 𝑞𝑛 → 0 and 𝑞𝑛 is decreasing. Hence the series converges by the Alternating
Series Test.

8.5.6 Abel’s Test

Another conditional convergence test is the Abel Test. This looks similar to the Dirichlet Test, however notice
that the Abel Test only deals with real sequences.

Theorem 8.58: Abel’s Test

Let (𝑎𝑛 ) and (𝑞𝑛 ) be sequences in ℝ. Suppose that

• 𝑞𝑛 is monotone,

Dr. Silvio Fanzon [email protected]

Numbers, Sequences and Series Page 272

• 𝑞𝑛 is bounded,
• The series below converges
∞
∑ 𝑎𝑛 .
𝑛=1

Then the following series converges

∞
∑ 𝑎𝑛 𝑞 𝑛
𝑛=1

The proof is similar to the one of the Dirichlet Test. We decided to omit it.

Example 8.59

Prove that the series below converges conditionally

∞
(−1)𝑛 1 𝑛
∑ (1 + ) .
𝑛=1 𝑛 𝑛

Solution. Set
(−1)𝑛 1 𝑛
𝑎𝑛 ∶= , 𝑞𝑛 ∶= (1 + ) .
𝑛 𝑛
∞
We have seen that 𝑞𝑛 is monotone increasing and bounded. Moreover the series ∑𝑛=1 𝑎𝑛
∞
converges by the Alternating Series Test, as seen in Example 8.57. Hence the series ∑𝑛=1 𝑎𝑛 𝑞𝑛
converges by the Abel Test.
However the series in question does not converge absolutely. Indeed,

(−1)𝑛 1 𝑛 1 1
| (1 + ) | = 𝑞𝑛 ≥ 𝑞1 ,
𝑛 𝑛 𝑛 𝑛
since (𝑞𝑛 ) is increasing. As the series
∞
1
∑ 𝑞1
𝑛=1 𝑛
diverges, by the Comparison Test we conclude that also
∞
(−1)𝑛 1 𝑛
∑| (1 + ) |
𝑛=1 𝑛 𝑛

diverges, proving that the series in the example converges conditionally.

Dr. Silvio Fanzon [email protected]

License

Reuse
This work is licensed under
CC-BY-NC-ND 4.0

Citation
For attribution, please cite this work as:

Fanzon, Silvio. (2024). Lecture Notes on Numbers, Sequences and Series.

https://www.silviofanzon.com/2024-NSS-Notes/

BibTex citation:

@electronic{Fanzon-NSS-2024,
author = {Fanzon, Silvio},
title = {Lecture Notes on Numbers, Sequences and Series},
url = {https://www.silviofanzon.com/2024-NSS-Notes/},
year = {2024}}

273
References
[1] S. Abbott. Understanding Analysis. Second Edition. Springer, 2015.
[2] Bartle, Robert G. and Sherbert, Donald R. Introduction to Real Analysis. Fourth Edition. Wiley, 2011.
[3] W. Rudin. Principles of Mathematical Analysis. Third Edition. McGraw Hill, 1976.

274