MATH255 - Honours Analysis 2

Basic point-set topology; metric spaces; Hölder-Minkowski Inequalities; compactness; series, series of functions, uniform
and pointwise convergence.

Based on lectures from Winter, 2024 by Prof. Dimitry Jakobson

Notes by Louis Meunier

Contents
1 Introduction 3
1.1 Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Point-Set Topology 7
2.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Continuous Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Product Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.6 Metrizability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.7 Compactness, Connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.8 Path Components, Connected Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.8.1 Cantor Staircase Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3 𝐿 𝑝 Spaces 23
3.1 Review of ℓ 𝑝 Norms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 ℓ 𝑝 Norms, Hölder-Minkowski Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Complete Metric Spaces, Completeness of ℓ 𝑝 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4 Contraction Mapping Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.5 Equivalent Notions of Compactness in Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4 Derivatives 35
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Chain Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3 Critical Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.4 Aside: Continued Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
 4.5 Back To Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.6 L’Hopital’s Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.7 Taylor’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.8 Convex Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

5 Riemann Integral 40
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5.2 Cauchy Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3 Squeeze Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.4 Fundamental Theorem of Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.5 Upper and Lower Riemann Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.6 Indefinite Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.7 Lebesgue Integrability Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.8 Integration by Parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6 Function Sequences, Series 48

6.1 Pointwise and Uniform Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.2 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.3 Tests for Absolute Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.4 Tests for Non-Absolute Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.5 Series of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6.6 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

7 Appendix 55
7.1 Notes from Tutorials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.2 Miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
7.3 Class Midterm Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

8 List of Theorems 63
1 Introduction

1.1 Metric Spaces

↩→ Definition 1.1: Metric Space

A set 𝑋 is a metric space with distance 𝑑 if

1. (symmetric) 𝑑(𝑥, 𝑦) = 𝑑(𝑦, 𝑥) ⩾ 0

2. 𝑑(𝑥, 𝑦) = 0 ⇐⇒ 𝑥 = 𝑦

3. (triangle inequality) 𝑑(𝑥, 𝑦) + 𝑑(𝑦, 𝑧) ⩾ 𝑑(𝑥, 𝑧)

Remark 1.1. If 1., 3. are satisfied but not 2., 𝑑 can be called a “pseudo-distance”.

↩→ Definition 1.2: Open Metric Space

Let (𝑋 , 𝑑) be a metric space. A subset 𝐴 ⊆ 𝑋 is open ⇐⇒ ∀ 𝑥 ∈ 𝐴, ∃𝑟 = 𝑟(𝑥) > 0 s.t. 𝐵(𝑥, 𝑟(𝑥)) ⊆ 𝐴.

↩→ Definition 1.3: Normed Space

Let 𝑋 be a vector space over R. The norm on 𝑋, denoted ||𝑥|| ∈ R, is a function that satisfies

1. ||𝑥|| ⩾ 0

2. ||𝑥|| = 0 ⇐⇒ 𝑥 = 0

3. ||𝑐 · 𝑥|| = |𝑐| · ||𝑥||

4. ||𝑥 + 𝑦|| ⩽ ||𝑥|| + ||𝑦||

If 𝑋 is a normed vector space over R, we can define a distance 𝑑 on 𝑋 by 𝑑(𝑥, 𝑦) = ||𝑥 − 𝑦||.

↩→Proposition 1.1
If 𝑋 is a normed vector space over R, a distance 𝑑 on 𝑋 by 𝑑(𝑥, 𝑦) = ||𝑥 − 𝑦|| makes (𝑋 , 𝑑) a metric space.

Proof
Proof.
Proof 1. 𝑑(𝑥, 𝑦) = ||𝑥 − 𝑦|| ⩾ 0

2. 𝑑(𝑥, 𝑦) = 0 ⇐⇒ ||𝑥 − 𝑦|| = 0 ⇐⇒ 𝑥 − 𝑦 = 0 ⇐⇒ 𝑥 = 𝑦

3. 𝑑(𝑥, 𝑦) + 𝑑(𝑦, 𝑧) = ||𝑥 − 𝑦|| + ||𝑦 − 𝑧|| ⩾ ||(𝑥 − 𝑦) + (𝑦 − 𝑧)|| = ||𝑥 − 𝑧|| := 𝑑(𝑥, 𝑧)
■

1.1 Introduction: Metric Spaces 3

 ⊛ Example 1.1: 𝐿 𝑝 distance in R𝑛
Let 𝑥 ∈ R𝑛 , 𝑥 = (𝑥 1 , 𝑥2 , . . . , 𝑥 𝑛 ). The 𝐿 𝑝 norm is defined

1
||𝑥|| 𝑝 := (|𝑥1 | 𝑝 + |𝑥2 | 𝑝 + · · · + |𝑥 𝑛 | 𝑝 ) 𝑝 .

In the case 𝑝 = 2, 𝑛 = 2, we simply have the standard Euclidean distance over R2 .

Unit Balls: consider when ||𝑥|| 𝑝 ⩽ 1, over R2 .

Unit Balls:

• 𝑝 = 1 : |𝑥1 | + |𝑥2 | ⩽ 1; this forms a “diamond ball” in the plane.

q
• 𝑝=2: |𝑥 1 | 2 + |𝑥2 | 2 ⩽ 1; this forms a circle of radius 1. Clearly, this surrounds a larger area
than in 𝑝 = 2.

A natural question that follows is what happens as 𝑝 → ∞? Assuming |𝑥1 | ⩾ |𝑥2 |:

𝑝1
||𝑥|| 𝑝 = |𝑥 1 | 𝑝 + |𝑥2 | 𝑝
𝑝  𝑝1
𝑥2
 
𝑝
= |𝑥1 | 1+
𝑥1
𝑝  𝑝1
𝑥2

= |𝑥1 | 1 +
𝑥1

1
If |𝑥1 | > |𝑥2 |, this goes to |𝑥 1 |. If they are instead equal, then ||𝑥|| 𝑝 = |𝑥 1 | · 2 𝑝 → |𝑥1 | · 1 as well.
Hence, lim𝑝→∞ ||𝑥|| 𝑝 = max{|𝑥1 | , |𝑥2 |}. Thus, the unit ball will approach max{|𝑥1 | , |𝑥2 |} ⩽ 1, that
is, the unit square.

↩→Proposition 1.2
Let 𝑥 ∈ R𝑛 . Then, ||𝑥|| 𝑝 → max{|𝑥1 | , . . . , |𝑥 𝑛 |} as 𝑝 → ∞.

Remark 1.2. This is an extension of the previous example to arbitrary real space; the proof follows nearly identically.

↩→ Definition 1.4: Convex Set

Let 𝑋 be a normed space, and take 𝑥, 𝑦 ∈ 𝑋. The line segment from 𝑥 to 𝑦 is the set

{𝑡 · 𝑥 + (1 − 𝑡) · 𝑦 : 0 ⩽ 𝑡 ⩽ 1}.

Let 𝐴 ⊆ 𝑋. 𝐴 is convex ⇐⇒ ∀ 𝑥, 𝑦 ∈ 𝐴, we have that

(𝑡 · 𝑥 + (1 − 𝑡) · 𝑦) ∈ 𝐴 ∀ 0 ⩽ 𝑡 ⩽ 1.

1.1 Introduction: Metric Spaces 4

 Figure 1: Regions of R2 where ||𝑥|| 𝑝 ⩽ 1 for various values of 𝑝.

Figure 2: Convex (left) versus not convex (right) sets.

Remark 1.3. Think of this as saying “a set is convex iff every point on a line segment connected any two points is in
the set”.

↩→ Definition 1.5: ℓ 𝑝
The space ℓ 𝑝 of sequences is defined as

∞
|𝑥 𝑛 | 𝑝 < +∞}
Õ
{𝑥 = (𝑥 1 , 𝑥2 , . . . , 𝑥 𝑛 , . . . ) : ∗.
𝑛=1

𝑝
𝑝1
Then, ∗ defines the ℓ 𝑝 norm on the space of sequences; that is, ||𝑥|| 𝑝 :=
Í∞
𝑛=1 |𝑥 𝑛 | .

1.1 Introduction: Metric Spaces 5

 1
⊛ Example 1.2: ℓ 𝑝 , 𝑥 𝑛 = 𝑛

. Let 𝑥 𝑛 = 𝑛1 . For which 𝑝 is 𝑥 ∈ ℓ 𝑝 ? We have, raising the norm to the power of 𝑝 for ease:

𝑝
||𝑥|| 𝑝 = |𝑥1 | 𝑝 + |𝑥2 | 𝑝 + · · · + |𝑥 𝑛 | 𝑝 + · · ·
 𝑝
𝑝 1
=1 + + · · · < ∞ ⇐⇒ 𝑝 > 1.
2

In the case that 𝑝 = 1, this becomes a harmonic sum, which diverges.

⊛ Example 1.3: 𝐿 𝑝 space of functions

Let 𝑓 (𝑥) be a continuous function. We define the norm of 𝑓 over an interval [𝑎, 𝑏]

∫ 𝑏  𝑝1
𝑝
|| 𝑓 || 𝑝 = | 𝑓 (𝑥)| 𝑑𝑥 .
𝑎

Remark 1.4. Triangle inequality for ||𝑥|| 𝑝 or || 𝑓 || 𝑝 is called Minkowski inequality; ||𝑥|| 𝑝 + ||𝑦|| 𝑝 ⩾ ||𝑥 + 𝑦|| 𝑝 . This
will be discussed further.

⊛ Example 1.4: Distances between sets in R2

Let 𝐴, 𝐵 be bounded, closed, “nice” sets in R2 . We define

𝑑(𝐴, 𝐵) := Area(𝐴△𝐵),

where
𝐴△𝐵 : (𝐴 \ 𝐵) ∪ (𝐵 \ 𝐴) = (𝐴 ∪ 𝐵) \ (𝐴 ∩ 𝐵).

It can be shown that this is a “valid” distance.

Remark 1.5. △ denotes the “symmetric difference” of two sets.

⊛ Example 1.5: 𝑝-adic distance

𝑎 𝑐
Let 𝑝 be a prime number. Let 𝑥 = ∈ Q, and write 𝑥 = 𝑝 𝑘 · , where 𝑐, 𝑑 are not divisible by 𝑝.


𝑏 𝑑
Then, the 𝑝-adic norm is defined ||𝑥|| 𝑝 := 𝑝 −𝑘 . It can be shown that this is a norm.

Suppose 𝑝 = 2, 𝑥 = 28 = 4 · 7 = 22 · 7. Then, ||28|| 2 = 2−2 = 14 ; similarly, ||1024|| 2 = ||210 || 2 = 2−10 .

More generally, we have that ||2 𝑘 || 2 = 2−𝑘 ; coversely, ||2−𝑘 || = 2 𝑘 . That is, the closer to 0, the
larger the distance, and vice versa, contrary to our notion of Euclidean distance.

↩→Proposition 1.3
||𝑥|| 𝑝 as defined above is a well-defined norm over Q.

1.1 Introduction: Metric Spaces 6

Proof. Left as a (homework) exercise.
Proof ■

2 Point-Set Topology

2.1 Definitions

↩→ Definition 2.1: Topological space

A set 𝑋 is a topological space if we have a collection of subsets 𝜏 of 𝑋 called open sets s.t.

1. ∅ ∈ 𝜏, 𝑋 ∈ 𝜏

2. Consider {𝐴 𝛼 } 𝛼∈𝐼 where 𝐴 𝛼 an open set for any 𝛼; then, 𝐴 𝛼 ∈ 𝜏, that is, it is also an open set.
Ð
𝛼∈𝐼

3. If 𝐽 is a finite set, and 𝐴𝛽 open for all 𝛽 ∈ 𝐽, then 𝐴𝛽 ∈ 𝜏 is also open.

Ñ
𝛽∈𝐽

In other words, 2.: arbitrary unions of open sets are open, and 3.: finite intersections of open sets are
open.

↩→ Definition 2.2: Closed sets

Closed sets are complements of open sets; hence, axioms for closed sets follow appropriately;

1.* 𝑋 , ∅ closed;

2.* 𝐵 𝛼 closed ∀ 𝛼 ∈ 𝐼 =⇒ 𝐵 𝛼 closed.

Ñ
𝛼∈𝐼

3.* 𝐵𝛽 closed ∀ 𝛽 ∈ 𝐽, 𝐽 finite, then 𝐵𝛽 also closed.

Ð
𝛽∈𝐽

↩→ Lecture 01; Last Updated: Tue Apr 9 14:45:17 EDT 2024

2.1 Point-Set Topology: Definitions 7

 ↩→ Definition 2.3: Equivalence of Metrics
Suppose we have a metric space 𝑋 with two distances 𝑑1 , 𝑑2 ; will these necessarily admit the same
topology?

A sufficient condition is that, if ∀ 𝑥 ≠ 𝑦 ∈ 𝑋, ∃1 < 𝐶 < +∞ s.t.

1 𝑑1 (𝑥, 𝑦)
< < 𝐶.
𝐶 𝑑2 (𝑥, 𝑦)

That is, the distances are equivalent, up to multiplication by a constant.

𝑑1
Indeed, this condition gives that 𝑑2 < 𝐶𝑑1 and 𝑑2 > 𝐶; this gives

𝑟
𝐵 𝑑1 (𝑥, ) ⊆ 𝐵 𝑑2 (𝑥, 𝑟) ⊆ 𝐵 𝑑1 (𝑥, 𝐶 · 𝑟).
𝑐

Hence, 𝑑1 , 𝑑2 define the same open/closed sets on 𝑋 thus admitting the same topologies. We write
𝑑1 ≍ 𝑑2 .

Remark 2.1. If 𝑑1 ≍ 𝑑2 and 𝑑2 ≍ 𝑑3 , then also 𝑑1 ≍ 𝑑3 . Moreover, clearly, 𝑑1 ≍ 𝑑1 and 𝑑1 ≍ 𝑑2 =⇒ 𝑑2 ≍ 𝑑1 , hence

this is a well-defined equivalence relation.
Hence, its enough to show that ∀ 1 < 𝑝 < +∞, we have ||𝑥|| 𝑝 ≍ ||𝑥|| ∞ to show that any ||𝑥|| 𝑞 norm are equivalent
for all 𝑞 on R𝑛 .

↩→ Definition 2.4: Interior, Boundary of a Topological Set

Let 𝑋 be a topological space, 𝐴 ⊆ 𝑋 and let 𝑥 ∈ 𝑋. We have the following possibilities

1. ∃𝑈-open : 𝑥 ∈ 𝑈 ⊆ 𝐴. In this case, we say 𝑥 ∈ the interior of 𝐴, denoted

𝑥 ∈ Int(𝐴).

2. ∃𝑉-open : 𝑥 ∈ 𝑉 ⊆ 𝑋 \ 𝐴 = 𝐴𝐶 . In this case, we write

𝑥 ∈ Int(𝐴𝐶 ).

3. ∀ 𝑈-open : 𝑥 ∈ 𝑈, 𝑈 ∩ 𝐴 ≠ ∅ AND 𝑈 ∩ 𝐴𝐶 ≠ ∅. In this case, we say 𝑥 is in the boundary of 𝐴, and

denote
𝑥 ∈ 𝜕𝐴.

↩→ Definition 2.5: Closure

𝑥 ∈ Int(𝐴) or 𝑥 ∈ 𝜕𝐴 (that is, 𝑥 ∈ Int(𝐴) ∪ 𝜕𝐴) ⇐⇒ every open set 𝑈 that contains 𝑥 intersects 𝐴.1 Such
points are called limit points of 𝐴. The set of all limits points of 𝐴 is called the closure of 𝐴, denoted 𝐴.

2.1 Point-Set Topology: Definitions 8

Remark 2.2. We have that
Int(𝐴) ⊆ 𝐴 ⊆ 𝐴 = Int(𝐴) ∪ 𝜕𝐴.

↩→Proposition 2.1: Properties of Int(𝐴)

Int(𝐴) is open, and it is the largest open set contained in 𝐴. It is the union of all 𝑈-open s.t. 𝑈 ⊆ 𝐴.
Moreover, we have that
Int(Int(𝐴)) = Int(𝐴).

↩→Proposition 2.2: Properties of 𝐴

𝐴 is closed; 𝐴 is the smallest closed set that contains 𝐴, that is, 𝐴 = 𝐵 where 𝐵 closed and 𝐴 ⊆ 𝐵. We
Ñ

have too that

(𝐴) = 𝐴.

↩→Proposition 2.3
1. 𝐴 is open ⇐⇒ 𝐴 = Int(𝐴)

2. 𝐴 is closed ⇐⇒ 𝐴 = 𝐴

2.2 Basis

↩→ Definition 2.6: Basis for a Toplogy

Let 𝜏 be a topology on 𝑋. Let ℬ ⊆ 𝜏 be a collection of open sets in 𝑋 such that every open set is a union
of open sets in ℬ.

⊛ Example 2.1: Example Basis

𝑋 = R, and ℬ = {all open intervals (𝑎, 𝑏) : −∞ < 𝑎 < 𝑏 < +∞}.

↩→Proposition 2.4
Let ℬ be a collection of open sets in 𝑋. Then, ℬ is a basis ⇐⇒

1. ∀ 𝑥 ∈ 𝑋 , ∃𝑈-open ∈ ℬ s.t. 𝑥 ∈ 𝑈.

2. If 𝑈1 ∈ ℬ and 𝑈2 ∈ ℬ, and 𝑥 ∈ 𝑈1 ∩ 𝑈2 , then ∃𝑈3 ∈ ℬ s.t. 𝑥 ∈ 𝑈3 ⊆ 𝑈1 ∩ 𝑈2 .

1“Requires” proof.

2.2 Point-Set Topology: Basis 9

 ⊛ Example 2.2
Consider 𝑋 = R. Requirement 1. follows from taking 𝑈 = (𝑥 − 𝜀, 𝑥 + 𝜀) for any 𝜀 > 0. For 2., suppose
𝑥 ∈ (𝑎, 𝑏) ∩ (𝑐, 𝑑) =: 𝑈1 ∩ 𝑈2 . Let 𝑈3 = (max{𝑎, 𝑐}, min{𝑏, 𝑑}); then, we have that 𝑈3 ⊆ 𝑈1 ∩ 𝑈2 ,
while clearly 𝑥 ∈ 𝑈3 .

↩→Proposition 2.5
In a metric space, a basis for a topology is a collection of open balls,

{𝐵(𝑥, 𝑟) : 𝑥 ∈ 𝑋 , 𝑟 > 0} = {{𝑦 ∈ 𝑋 : 𝑑(𝑥, 𝑦) < 𝑟} : 𝑥 ∈ 𝑋 , 𝑟 > 0}.

Proof We prove via proposition 2.4. Property 1. holds clearly; 𝑥 ∈ 𝐵(𝑥, 𝜀)-open ⊆ ℬ.
Proof
Proof.
For property 2., let 𝑥 ∈ 𝐵(𝑦1 , 𝑟1 ) ∩ 𝐵(𝑦2 , 𝑟2 ), that is, 𝑑(𝑥, 𝑦1 ) < 𝑟1 and 𝑑(𝑥, 𝑦2 ) < 𝑟2 . Let

𝛿 := min{𝑟1 − 𝑑(𝑥, 𝑦1 ), 𝑟2 − 𝑑(𝑥, 𝑦2 )}.

We claim that 𝐵(𝑥, 𝛿) ⊆ 𝑈1 ∩ 𝑈2 .

Let 𝑧 ∈ 𝐵(𝑥, 𝛿). Then,

△≠
𝑑(𝑧, 𝑦1 ) ⩽ 𝑑(𝑧, 𝑥) + 𝑑(𝑥, 𝑦1 ) < 𝛿 + 𝑑(𝑥, 𝑦1 ) ⩽ 𝑟1 − 𝑑(𝑥, 𝑦1 ) + 𝑑(𝑥, 𝑦1 ) = 𝑟1 ,

hence, as 𝑑(𝑧, 𝑦1 ) < 𝑟1 =⇒ 𝑧 ∈ 𝐵(𝑦1 , 𝑟1 ) = 𝑈1 . Replacing each occurrence of 𝑦1 , 𝑟1 with 𝑦2 , 𝑟2 respectively

gives identically that 𝑧 ∈ 𝐵(𝑦2 , 𝑟2 ) = 𝑈2 . Hence, we have that 𝐵(𝑥, 𝛿) ⊆ 𝑈1 ∩ 𝑈2 and 2. holds. ■

2.3 Subspaces

↩→ Definition 2.7
Let 𝑋 be a topological space and let 𝑌 ⊆ 𝑋. We define the subspace topology on 𝑌:

1. Open sets in 𝑌 = {𝑌 ∩ open sets in 𝑋}

↩→Proposition 2.6: Consequences of Subspace Topologies

Suppose ℬ is a basis for a topology in 𝑋. Then, {𝑈 ∩ 𝑌 : 𝑈 ∈ ℬ} forms a basis for the subspace 𝑌 ⊆ 𝑋.

Suppose 𝑋 a metric space. Then, 𝑌 is also a metric space, with the same distance.

↩→Proposition 2.7
Let 𝑌 ⊆ 𝑋- a metric space. Then, the metric space topology for (𝑌, 𝑑) is the same as the subspace
topology.

2.3 Point-Set Topology: Subspaces 10

Proof. (Sketch) A basis for the open sets in 𝑋 can be written
Proof 𝐵(𝑥 𝛼 , 𝑟 𝛼 ); hence
Ð
𝛼∈𝐼

Ø Ø
𝑌∩( 𝐵(𝑥 𝛼 , 𝑟 𝛼 )) = (𝑌 ∩ 𝐵(𝑥 𝛼 , 𝑟 𝛼 ))
𝛼∈𝐼 𝛼∈𝐼

is an open set topology for 𝑌. ■

↩→ Lemma 2.1
Let 𝐴 ⊆ 𝑋-open, 𝐵 ⊆ 𝐴; 𝐵-open in subspace topology for 𝐴 ⇐⇒ 𝐵-open in 𝑋.

↩→ Lemma 2.2

Let 𝑌 ⊆ 𝑋, 𝐴 ⊆ 𝑌. Then, 𝐴 in 𝑌 = 𝑌 ∩ 𝐴 in 𝑋. We can denote this

𝐴𝑌 = 𝐴𝑋 ∩ 𝑌.

2.4 Continuous Functions

↩→ Definition 2.8: Continuous Function

Let 𝑋 , 𝑌 be topological spaces. Let 𝑓 : 𝑋 → 𝑌. 𝑓 is continuous ⇐⇒ ∀ open 𝑉 ∈ 𝑌, 𝑓 −1 (𝑉)-open in 𝑋.

↩→Proposition 2.8
This definition is consistent with the normal 𝜀-𝛿 definition on the real line.

Proof Let 𝑓 : R → R, continuous; that is, ∀ 𝜀 > 0, ∀ 𝑥 ∈ R∃𝛿 > 0 s.t. |𝑥 1 − 𝑥| < 𝛿, then | 𝑓 (𝑥1 ) − 𝑓 (𝑥)| < 𝜀.
Proof
Proof.
Let 𝑉 ⊆ R open. Let 𝑦 ∈ 𝑉. Then, ∃𝜀 : (𝑦 − 𝜀, 𝑦 + 𝜀) ⊆ 𝑉. Let 𝑦 = 𝑓 (𝑥), hence 𝑦 ∈ 𝑓 −1 (𝑉). Now, if
𝑑(𝑥, 𝑥1 ) < 𝛿, we have that 𝑑( 𝑓 (𝑥1 ), 𝑓 (𝑥)) < 𝜀 (by continuity of 𝑓 ), hence 𝑓 (𝑥1 ) ∈ (𝑦 − 𝜀, 𝑦 + 𝜀) ⊆ 𝑉; moreover,
(𝑥 − 𝛿, 𝑥 + 𝛿) ⊆ 𝑓 −1(𝑉), thus 𝑓 −1 (𝑉) is open as required.
The inverse of this proof follows identically. ■

↩→ Lecture 02; Last Updated: Tue Apr 9 21:38:10 EDT 2024

↩→Proposition 2.9
Suppose ℬ forms a basis of topology for 𝑌. Then, 𝑓 : 𝑋 → 𝑌 is continuous if 𝑓 −1 (𝑈) open ∀ 𝑈 ∈ ℬ.

Proof If 𝑈-open set in 𝑌, then ∃𝐼-index set and a collection of open sets {𝐴 𝛼 } 𝛼∈𝐼 , 𝐴 𝛼 ∈ ℬ, s.t. 𝑈 = ∪𝛼∈𝐼 𝐴 𝛼 .
Proof
Proof.
Then, we have

𝑓 −1 (𝑈) = 𝑓 −1 (∪𝛼∈𝐼 (𝐴 𝛼 )) = ∪𝛼∈𝐼 𝑓 −1 (𝐴 𝛼 )

| {z }

2.4 Point-Set Topology: Continuous Functions 11

Hence, if each 𝑓 −1 (𝐴 𝛼 ) open, then ∪𝛼∈𝐼 𝑓 −1 (𝐴 𝛼 ) open; hence it suffices to check if 𝑓 −1 (𝑈) ∀ 𝑈-open in 𝑉 is open
to see if 𝑓 continuous. ■

↩→ Theorem 2.1: Continuity of Composition

If 𝑓 : 𝑋 → 𝑌 continuous and 𝑔 : 𝑌 → 𝑍 continuous, then 𝑔 ◦ 𝑓 continuous as well.

Proof Let 𝑈-open in 𝑍. Then

Proof
Proof.

(𝑔 ◦ 𝑓 )−1 (𝑈) = 𝑓 −1 ( 𝑔 −1 (𝑈) )

| {z }
open in 𝑌
| {z }
open in 𝑋

↩→Proposition 2.10
If 𝑓 : 𝑋 → 𝑌 continuous and 𝐴 ⊆ 𝑋, 𝐴 has subspace topology, then 𝑓 | 𝐴 : 𝐴 → 𝑌 is also continuous.2

Proof Let 𝑈-open in 𝑌. Then

Proof
Proof.

( 𝑓 | 𝐴 )−1 (𝑈) = 𝑓 −1 (𝑈) ∩ 𝐴

| {z } |{z}
open open

By the definition of subspace topology, this is an open set and hence 𝑓 | 𝐴 is continuous. ■

2.5 Product Spaces

↩→ Definition 2.9: Finite Product Spaces

Let 𝑋1 , . . . , 𝑋𝑛 be topological spaces. We define

(𝑋1 × 𝑋2 × · · · × 𝑋𝑛 ),

and aim to define a product topology; a basis of which consists of cylinder sets.

↩→ Definition 2.10: Cylinder Set

A cylinder set has the form
𝐴1 × 𝐴2 × · · · × 𝐴 𝑛

where each 𝐴 𝑗 -open in 𝑋 𝑗 .

2We denote 𝑓 | 𝐴 as the restriction of the domain of 𝑓 to 𝐴.

2.5 Point-Set Topology: Product Spaces 12

 ⊛ Example 2.3
Given an open interval (𝑎 1 , 𝑏1 ), (𝑎 2 , 𝑏2 ) ⊂ R, the set (𝑎 1 , 𝑏1 ) × (𝑎 2 , 𝑏2 ) ⊂ R2 is a basis for the topology
on R2 .

↩→ Definition 2.11: Projection

Let 𝑋1 × 𝑋2 × · · · 𝑋𝑛 =: 𝑋. The projection 𝜋 𝑗 : 𝑋 → 𝑋 𝑗 maps (𝑥1 , . . . , 𝑥 𝑛 ) → 𝑥 𝑗 ∈ 𝑋 𝑗 .

Remark 2.3. One can show 𝜋 𝑗 continuous.

↩→ Definition 2.12: Coordinate Function

Given a function 𝑓 : 𝑌 → 𝑋1 × · · · 𝑋𝑛 = (𝑥 1 (𝑦), 𝑥2 (𝑦), . . . , 𝑥 𝑛 (𝑦)). The coordinate function is

𝑓𝑗 = 𝜋𝑗 ◦ 𝑓 ; 𝑓 𝑗 = 𝑥 𝑗 (𝑦).

↩→Proposition 2.11
𝑓 : 𝑌 → 𝑋 = 𝑋1 × · · · × 𝑋𝑛 continuous ⇐⇒ 𝑓 𝑗 : 𝑌 → 𝑋 𝑗 continuous.

Proof Its enough to show that ∀ 𝑈 ∈ ℬ-basis for 𝑋-product space, 𝑓 −1 (𝑈)-open in 𝑌. Take 𝑈 = 𝐴1 × · · · 𝐴𝑛 -
Proof
Proof.
open. Then, we claim that

𝑓 −1 (𝑈) = 𝑓 −1 (𝐴1 × · · · × 𝐴𝑛 ) = 𝑓1−1 (𝐴1 ) ∩ 𝑓2−1 (𝐴2 ) ∩ · · · ∩ 𝑓𝑛−1 (𝐴𝑛 ). ★

If this holds, then as each 𝑓𝑖 continuous (being a composition of continuous functions) and each 𝐴 𝑖 open in
𝑋𝑖 , then each 𝑓𝑖−1 (𝐴 𝑖 ) open in 𝑌 and hence ★, being the finite intersection of open sets in 𝑌, is itself open in 𝑌.
■

2.5 Point-Set Topology: Product Spaces 13

 ⊛ Example 2.4: Fourier Transform: Motivation for Infinite Product Toplogies
Let 𝑓 ∈ 𝐶([0, 2𝜋]) is real-valued. We write the 𝑛th Fourier coefficients

2𝜋
1
∫
𝑓ˆ(𝑛) = 𝑓 (𝑥)𝑒 −𝑖𝑛𝑥 d𝑥
2𝜋 0
2𝜋 2𝜋
1 1
∫ ∫
= 𝑓 (𝑥) cos(𝑛𝑥) d𝑥 − 𝑖 𝑓 (𝑥) sin(𝑛𝑥) d𝑥 .
2𝜋 0 2𝜋 0

And the Fourier transform of 𝑓 as the infinite product

Ö
𝑓 (𝑥) ↦→ (. . . , 𝑓ˆ(−𝑛), 𝑓ˆ(−𝑛 + 1), . . . , 𝑓ˆ(−1), 𝑓ˆ(0), 𝑓ˆ(1), · · · , 𝑓ˆ(𝑛), . . . ) ∈ (C)𝑛 .
𝑛∈Z

Hence, this is an (countably, as indexed by integers) infinite product space.

Now, let 𝑓 : R → R. Suppose 𝑓 (𝑥) → 0 “fast enough” as |𝑥| → ∞ and 𝑓 continuous. Then, we
can define the Fourier coefficients
∞
1
∫
𝑓ˆ(𝑡) = 𝑓 (𝑥)𝑒 −𝑖𝑡𝑥 d𝑥 ,
2𝜋 −∞

where 𝑡 ∈ R. We then have the transform

𝑓 ↦→ { 𝑓ˆ(𝑡)} 𝑡∈R .

In this case, our index set is R is (uncountably) infinite.

↩→ Definition 2.13: Product Topology/Cylinder Sets for ∞ Products

Let 𝑋 = 𝑋𝛼 . Then, a basis for 𝑋 is given by cylinder sets of the form 𝐴 = 𝐴 𝛼 where 𝐴 𝛼 -open
Î Î
𝛼∈𝐼 𝛼∈𝐼
in 𝑋𝛼 , AND 𝐴 𝛼 = 𝑋𝛼 except for finitely many indices 𝛼.

That is, there exists a finite set 𝐽 = (𝛼1 , . . . , 𝛼 𝑘 ) ⊆ 𝐼, such that we can write 𝐴 = 𝐴𝛼 × 𝑋𝛼
Î Î
𝛼∈𝐽 𝛼∉𝐽
(where 𝐴 𝛼 open in 𝑋𝛼 ).

↩→Proposition 2.12
Given 𝑓 : 𝑌 → 𝑋𝛼 = 𝑋, then (taking 𝑓𝛼 = 𝜋 𝛼 ◦ 𝑓 as before) we have that 𝑓 is continuous in 𝑋 ⇐⇒
Î
𝛼∈𝐼
𝑓𝛼 : 𝑌 → 𝑋𝛼 continuous in 𝑋𝛼 ∀ 𝛼 ∈ 𝐼.

Remark 2.4. Extension of proposition 2.11 to infinite product space.

2.5 Point-Set Topology: Product Spaces 14

Proof. Write 𝑈 =
Proof 𝐴𝛼 × 𝑋𝛼 . Then,
Î Î
𝛼∈𝐽 𝛼∉𝐽

Ù
𝑓 −1 (𝑈) = 𝑓𝛼−1 (𝐴 𝛼 )
𝛼∈𝐽

which is open in 𝑌, hence 𝑓 continuous. ■

Remark 2.5. The intersection of the entire spaces give no restriction.

↩→ Lecture 03; Last Updated: Fri Jan 19 11:49:27 EST 2024

2.6 Metrizability

↩→Proposition 2.13
Different metrics can define the same topology.

⊛ Example 2.5
1. Different ℓ 𝑝 metrics in R𝑛 (PSET 1)

2. Let (𝑋 , 𝑑) be a metric space. Then,

˜ 𝑑(𝑥, 𝑦)
𝑑(𝑥, 𝑦) :=
𝑑(𝑥, 𝑦) + 1

is also a metric (the first two axioms are trivial), and defines the same topology. Note, moreover,
˜
that 𝑑(𝑥, 𝑦) ⩽ 1 ∀ 𝑥, 𝑦; this distance is bounded, and can often be more convenient to work
with in particular contexts.

↩→ Question 2.1
Suppose (𝑋 𝑘 , 𝑑 𝑘 ) are metric spaces ∀ 𝑘 ⩾ 1. Then, we can define the product topology 𝜏 on

∞
Ö
𝑋 := 𝑋𝑘 .
𝑘=1

Does the product topology 𝜏 come from a metric? That is, is 𝜏 metrizable?

Remark 2.6. There do indeed exist examples of non-metrizable topological spaces; this question is indeed well-founded.

Answer. Let xx = (𝑥 1 , 𝑥2 , . . . , 𝑥 𝑛 , . . . ), yy = (𝑦1 , 𝑦2 , . . . , 𝑦𝑛 , . . . ) ∈

Î∞
𝑘=1 (where 𝑥 𝑖 , 𝑦 𝑖 ∈ 𝑋𝑖 ) be infinite sequences
of elements. Then, for each metric space 𝑋 𝑘 take the metric

𝑑 𝑘 (𝑥 𝑘 , 𝑦 𝑘 )
𝑑˜ 𝑘 (𝑥 𝑘 , 𝑦 𝑘 ) =
1 + 𝑑 𝑘 (𝑥 𝑘 , 𝑦 𝑘 )

2.6 Point-Set Topology: Metrizability 15

(as in the example above). Then, we define

∞ ˜
Õ 𝑑 𝑘 (𝑥 𝑘 , 𝑦 𝑘 )
𝐷(xx, yy) = ,
𝑘=1
2𝑘

1
noting that 𝐷(xx, yy) ⩽ ∞𝑘=1 2 𝑘 = 1 (by our construction, “normalizing” each metric), hence this is a valid,
Í

converging metric (which wouldn’t otherwise be guaranteed if we didn’t normalize the metrics). It remains to
show whether this metric omits the same topology as 𝜏. ■

2.7 Compactness, Connectedness

↩→ Definition 2.14: Compact

A set 𝐴 in a topological space is said to be compact if every cover has a finite subcover. That is, if
Ø
𝐴⊆ 𝑈 𝛼 − open,
𝛼∈𝐼

Ð𝑛
then ∃{𝛼1 , . . . , 𝛼 𝑛 ∈ 𝐼} such that 𝐴 ⊆ 𝑖=1 𝑈 𝛼 𝑖 .

↩→Proposition 2.14
A closed interval [𝑎, 𝑏] is compact.

Proof If3 𝑎 = 𝑏, this is clear. Suppose 𝑎 < 𝑏, and let [𝑎, 𝑏] ⊆ 𝑈 𝑖 =: 𝒰 be an arbitrary cover. Then, we
Ð
Proof
Proof. 𝑖∈𝐼
proceed in the following steps:

1. Claim: Given 𝑥 ∈ [𝑎, 𝑏], 𝑥 ≠ 𝑏, ∃𝑦 ∈ [𝑎, 𝑏] s.t. [𝑥, 𝑦] has a finite subcover.
Let 𝑥 ∈ [𝑎, 𝑏], 𝑥 ≠ 𝑏. Then, ∃𝑈 𝛼 ∈ 𝒰 : 𝑥 ∈ 𝑈 𝛼 . Since 𝑈 𝛼 open, and 𝑥 ≠ 𝑏, we further have that
∃𝑐 ∈ [𝑎, 𝑏] s.t. [𝑥, 𝑐) ⊆ 𝑈 𝛼 .
Now, let 𝑦 ∈ (𝑥, 𝑐); then, the interval [𝑥, 𝑦] ⊆ [𝑥, 𝑐) ⊆ 𝑈 𝛼 , that is, [𝑥, 𝑦] has a finite subcover.

2. Define 𝐶 := {𝑦 ∈ [𝑎, 𝑏] : 𝑦 > 𝑎, [𝑎, 𝑦] has a finite subcover}. We note that

• 𝐶 ≠ ∅; taking 𝑥 = 𝑎 in Step 1. above, we have that ∃𝑦 ∈ [𝑎, 𝑏] such that [𝑎, 𝑦] has a finite step cover,
so this 𝑦 ∈ 𝐶.
• 𝐶 bounded; by construction, ∀ 𝑦 ∈ 𝐶, 𝑎 < 𝑦 ⩽ 𝑐.

Thus, we can validly define 𝑐 := sup 𝐶, noting that 𝑎 < 𝑐 ⩽ 𝑏. Ultimately, we wish to prove that 𝑐 = 𝑏,
completing the proof that [𝑎, 𝑏] has a finite subcover.

3. Claim: 𝑐 ∈ 𝐶.
Let 𝑈𝛽 ∈ 𝒰 : 𝑐 ∈ 𝑈𝛽 . Then, by the openness of 𝑈𝛽 , ∃𝑑 ∈ [𝑎, 𝑏] s.t. (𝑑, 𝑐] ⊆ 𝑈𝛽 .
3This proof is adapted from that of Theorem 27.1 in Munkre’s Topology, an identical theorem but applied to more general ordered
topologies.

2.7 Point-Set Topology: Compactness, Connectedness 16

 Supposing 𝑐 ∉ 𝐶, then ∃𝑧 ∈ 𝐶 such that 𝑧 ∈ (𝑑, 𝑐); if one did not exist, then this would imply that 𝑑 was
a smaller upper bound that 𝑐, a contradiction. Thus, [𝑧, 𝑐] ⊆ (𝑑, 𝑐] ⊆ 𝑈𝛽 .
Moreover, we have that, given 𝑧 ∈ 𝐶, [𝑎, 𝑧] has a finite subcover; call it 𝑈 𝑧 ⊆ 𝒰. This gives, then:

[𝑎, 𝑐] = [𝑎, 𝑧] ∪ [𝑧, 𝑐] ⊆ 𝑈 𝑧 ∪ 𝑈𝛽 .

But this is a finite subcover of [𝑎, 𝑐], contradicting the fact that 𝑐 ∉ 𝐶. We conclude, then, that 𝑐 ∈ 𝐶
after all.

4. Claim: 𝑐 = 𝑏.
Suppose not; then, since we have 𝑐 ⩽ 𝑏, then assume 𝑐 < 𝑏. Then, applying Step 1. with 𝑥 = 𝑐 (which
we can do, by our assumption of 𝑐 ≠ 𝑏!), then we have that ∃𝑦 > 𝑐 s.t. [𝑐, 𝑦] has a finite subcover, call
this 𝑈 𝑦 ⊆ 𝒰.
Moreover, we had 𝑐 ∈ 𝐶, hence [𝑎, 𝑐] has a finite subcover, call this 𝑈 𝑐 ⊆ 𝒰.
Then, this gives us that
[𝑎, 𝑦] = [𝑎, 𝑐] ∪ [𝑐, 𝑦] ⊆ 𝑈 𝑐 ∪ 𝑈 𝑦 ,
that is, [𝑎, 𝑦] has a finite subcover, and so 𝑦 ∈ 𝐶. But recall that 𝑦 > 𝑐; hence, this a contradiction to 𝑐
being the least upper bound of 𝐶. We conclude that 𝑐 = 𝑏, and thus [𝑎, 𝑏] has a finite subcover, and is
thus compact.

■
Remark 2.7. A similar proof shows that [𝑎, 𝑏] is connected; we cannot cover it by two disjoint open sets.

↩→ Theorem 2.2: On Compactness

Let 𝐴 ⊆ R𝑛 . Then, 𝐴 compact ⇐⇒ 𝐴 closed and bounded.

↩→Proposition 2.15
If 𝑋 , 𝑌 are compact topological spaces, then 𝑋 × 𝑌 is compact.

Î𝑛
Remark 2.8. By induction, if 𝑋1 , . . . , 𝑋𝑛 compact, so is 𝑖=1 𝑋𝑖 .

↩→Proposition 2.16
A closed subset of a compact topological space is compact in the subspace topology.

Proof (Of theorem 2.2)

Proof
Proof.
( ⇐= ) If 𝐴 ⊆ R𝑛 closed and bounded, then 𝐴 ⊆ [−𝑅, +𝑅]𝑛 for some 𝑅 > 0 (it is contained in some “𝑛-cube”).
Then, we have that [−𝑅, 𝑅] is compact, by proposition 2.14, proposition 2.15, and proposition 2.16, 𝐴 itself
compact.
( =⇒ ) Suppose 𝐴 ⊆ R𝑛 is compact. Then, 𝑥∈𝐴 𝐵(𝑥, 𝜀) for some 𝜀 > 0 is an open cover of 𝐴. As 𝐴
Ð
compact, there must exist a finite subcover of this cover, 𝐴 ⊆ 𝑁 𝑁
𝑖=1 𝐵(𝑥 𝑖 , 𝑟 𝑖 ). Let 𝑅 := max𝑖=1 (||𝑥 𝑖 || + 𝑟 𝑖 ). Then,
Ð

𝐴 ⊆ 𝐵(0, 𝑅), that is, 𝐴 is bounded.

2.7 Point-Set Topology: Compactness, Connectedness 17

 Now, suppose 𝑥 is a limit point of 𝐴. Then, any neighborhood of 𝑥 contains a point in 𝐴, so ∀ 𝑟 >
0, 𝐵(𝑥, 𝑟) ∩ 𝐴 ≠ ∅, and so 𝐵(𝑥, 𝑟) also contains a point of 𝐴 for any 𝑟 > 0.
Now, suppose 𝑥 ∉ 𝐴 (looking for a contradiction). Then,
Ø Ø
𝑈 := 𝑈𝑟 := (R𝑛 \ 𝐵(𝑥, 𝑟)) = R𝑛 \ {𝑥}
𝑟>0 𝑟>0

is an open cover for the set 𝐴. 𝐴 being compact implies that 𝑈 has an finite subcover such that 𝐴 ⊂
𝑈𝑟1 ∪ 𝑈𝑟2 ∪ · · · ∪ 𝑈𝑟𝑁 . Let 𝑟0 = min𝑁
𝑖=1 𝑟 𝑖 . Then, 𝐴 ⊆ 𝑈 𝑟0 , and 𝐴 ∩ 𝐵(𝑥, 𝑟0 ) = ∅; but this is a contradiction to the
definition of a limit point, hence any limit point 𝑥 is contained in 𝐴 and 𝐴 is thus closed by definition. ■

↩→Proposition 2.17
Compact =⇒ sequentially compact; that is, every sequence in a compact set has a convergent subse-
quence.

↩→ Lecture 04; Last Updated: Tue Apr 9 14:45:17 EDT 2024

↩→ Definition 2.15: Connected

A topological space 𝑋 is not connected if 𝑋 = 𝑈 ∪ 𝑉 for two open, nonempty, disjoint sets 𝑈 , 𝑉.

If this does not hold, 𝑋 is said to be connected.

A set 𝐴 ⊆ 𝑋 is not connected if 𝐴 is not connected in the subspace topology ⇐⇒ 𝐴 =⊆ 𝑈 ∪ 𝑉, for

𝑈 , 𝑉-open in 𝑋, (𝑈 ∩ 𝐴) ≠ ∅, (𝑉 ∩ 𝐴) ≠ ∅ and 𝑈 ∩ 𝑉 = ∅.

↩→ Theorem 2.3
Let 𝑋 be a connected topological space. Let 𝑓 : 𝑋 → 𝑌 be a continuous function. Then, 𝑓 (𝑋) is also
connected.

Proof Suppose, seeking a contradiction, that 𝑋 is connected, but 𝑓 (𝑋) is not. Then, we can write 𝑓 (𝑋) ⊆ 𝑌 as
Proof
Proof.
𝑓 (𝑋) ⊆ 𝑈 ∪ 𝑉, such that 𝑈 , 𝑉 open in 𝑌 and 𝑈 ∩ 𝑉 = ∅. Then,

(𝑈 ∩ 𝑓 (𝑋)) ∩ (𝑉 ∩ 𝑓 (𝑋)) = ∅.

We also have that

𝑋⊆ 𝑓 −1 (𝑈) ∪ 𝑓 −1 (𝑉) .
| {z } | {z }
open in 𝑋 , ≠ ∅ open in 𝑋 , ≠ ∅

𝑓 −1 (𝑈) ∩ 𝑓 −1 (𝑉) = ∅ (that is, they are disjoint) by our assumption; this is a contradiction to the connectedness
of 𝑋, as we are able to write it as a subset of two disjoint open sets. Hence, 𝑓 (𝑋) is indeed connected. ■

2.7 Point-Set Topology: Compactness, Connectedness 18

 ↩→ Lemma 2.3
Any interval (𝑎, 𝑏), [𝑎, 𝑏], [𝑎, 𝑏), . . . , ⊆ R is connected.

Proof
Proof.
Proof ■

↩→ Theorem 2.4: “Intermediate Value Theorem”

Suppose 𝑋 is connected and 𝑓 : 𝑋 → R is a continuous function. Then, 𝑓 takes intermediate values.

More precisely, let 𝑎 = 𝑓 (𝑥), 𝑏 = 𝑓 (𝑦) for 𝑥, 𝑦 ∈ 𝑋. Assume 𝑎 < 𝑏. Then, ∀ 𝑎 < 𝑐 < 𝑏, ∃𝑧 ∈
𝑋 s.t. 𝑓 (𝑧) = 𝑐.

Proof Suppose, seeking a contradiction, that ∃𝑐 : 𝑎 < 𝑐 < 𝑏 s.t. 𝑐 ∉ 𝑓 (𝑋) (that is, there exists an intermediate
Proof
Proof.
value that is “not reached” by the function).
Let 𝑈 = (−∞, 𝑐) and 𝑉 = (𝑐, +∞); note that these are disjoint open sets. Then, we have that

𝑋 = 𝑓 −1 (𝑈) ∪ 𝑓 −1 (𝑉),

by our assumption of 𝑐 ∉ 𝑓 (𝑋). But this gives that 𝑋 is not connected, as the union of two open (by
continuity), disjoint, nonempty ( 𝑓 (𝑥) = 𝑎 ∈ 𝑈 =⇒ 𝑥 ∈ 𝑓 −1 (𝑈), and 𝑓 (𝑦) = 𝑏 ∈ 𝑉 =⇒ 𝑦 ∈ 𝑓 −1 (𝑉)) sets, a
contradiction. ■

↩→ Theorem 2.5
Suppose 𝑋 is compact, 𝑌-topological space, 𝑓 : 𝑋 → 𝑌 is a continuous function. Then, 𝑓 (𝑋) is also
compact.

Proof Let {𝑈 𝛼 } 𝛼∈𝐼 be an open cover of 𝑓 (𝑋) ⊆ 𝑌, that is,

Proof
Proof.
Ø Ø Ø Ø
𝑓 (𝑋) ⊆ 𝑈 𝛼 =⇒ 𝑋 ⊆ 𝑓 −1 ( 𝑈𝛼 ) = 𝑓 −1 (𝑈 𝛼 ) =: 𝑉𝛼 − open.
𝛼∈𝐼 𝛼∈𝐼 𝛼∈𝐼 𝛼∈𝐼

Then, this is an open

Ð cover of 𝑋; 𝑋 is compact, thus there exists a finite subcover, that is, indices {𝛼1 , . . . , 𝛼 𝑛 } ⊆
𝐼 such that 𝑋 = 𝑛𝑖=1 𝑉𝛼 𝑖 . Thus,
𝑛
Ø
𝑓 (𝑋) ⊆ 𝑈𝛼𝑖 ,
𝑖=1

which is a finite subcover of 𝑓 (𝑋). Thus, 𝑓 (𝑋) is compact. ■

Remark 2.9. Recall the “extreme value theorem”: let 𝑓 : [𝑎, 𝑏] → R a continuous function; then, a minimum and
maximum is obtained for 𝑓 (𝑥) on this interval for values in this interval.

2.7 Point-Set Topology: Compactness, Connectedness 19

 ↩→ Theorem 2.6
Let 𝑋 compact, and 𝑓 : 𝑋 → R a continuous function. Then,

max 𝑓 (𝑥) and min 𝑓 (𝑥)

𝑥∈𝑋 𝑥∈𝑋

are both attained.

Proof. 𝑓 (𝑋) ⊆ R is compact by theorem 2.5, and so by theorem 2.2, 𝑓 (𝑋) is closed and bounded. Let, then,
Proof
𝑚 = inf 𝑓 (𝑋) and 𝑀 = sup 𝑓 (𝑋); these necessarily exist, since 𝑓 (𝑋) is bounded. Both 𝑚 and 𝑀 are limit
points of 𝑓 (𝑋). But 𝑓 (𝑋) is closed, and hence contains all of its limit points, and thus 𝑚 ∈ 𝑓 (𝑋) and 𝑀 ∈ 𝑓 (𝑋),
and thus ∃𝑦𝑚 : 𝑓 (𝑦𝑚 ) = 𝑚 and 𝑦 𝑀 : 𝑓 (𝑦 𝑀 ) = 𝑀. ■

↩→ Definition 2.16: Path Connected

A set 𝐴 ⊆ 𝑋 is called path connected if ∀ 𝑥, 𝑦 ∈ 𝐴, ∃ 𝑓 : [𝑎, 𝑏] → 𝑋, continuous, s.t. 𝑓 (𝑎) = 𝑥, 𝑓 (𝑏) = 𝑦
and 𝑓 ([𝑎, 𝑏]) ⊆ 𝐴.

The set { 𝑓 (𝑡) : 𝑎 ⩽ 𝑡 ⩽ 𝑏} is called a path from 𝑥 to 𝑦.

↩→ Theorem 2.7: Path connected =⇒ connected

If 𝐴 ⊆ 𝑋 is path connected, then 𝐴 is connected.

Proof Suppose, seeking a contradiction, that 𝐴 is path connected, but not connected. Then, we can write
Proof
Proof.
𝐴 ⊆ 𝑈 ∪ 𝑉 , for open, disjoint, nonempty subsets 𝑈 , 𝑉 ⊆ 𝑋.
Let 𝑥 ∈ 𝑈 ∩ 𝐴 and 𝑦 ∈ 𝑉 ∩ 𝐴. Then, ∃ 𝑓 : [𝑎, 𝑏] → 𝐴 s.t. 𝑓 (𝑎) = 𝑥, 𝑓 (𝑏) = 𝑦, and 𝑓 ([𝑎, 𝑏]) ⊆ 𝐴, by the path
connectedness of 𝐴. Then,

[𝑎, 𝑏] ⊆ 𝑓 −1 (𝐴) ⊆ 𝑓 −1 (𝑈 ∩ 𝐴) ∪ 𝑓 −1 (𝑉 ∩ 𝐴) =: 𝑈1 ∪ 𝑈2 ,
| {z } | {z } |{z} |{z}
open open 𝑎∈ 𝑏∈

that is, [𝑎, 𝑏] is contained in a union of open, nonempty, disjoint sets, contradicting [𝑎, 𝑏] the connectedness
of [𝑎, 𝑏] by lemma 2.3. Thus, 𝐴 is connected. ■
Remark 2.10. A counterexample to the opposite side of the implication is the Topologist’s sine curve, the set

1
 
{(𝑥, sin ) : 𝑥 ∈ (0, 1]} ∪ {0} × [−1, 1].
𝑥

This set is connected in R2 , but is not path connected.

↩→Proposition 2.18
For open sets in R𝑛 , path connected ⇐⇒ connected.

↩→ Lecture 05; Last Updated: Sun Feb 4 22:28:52 EST 2024

2.7 Point-Set Topology: Compactness, Connectedness 20

2.8 Path Components, Connected Components

Remark 2.11. Remark that if a metric space 𝑋 is not connected, then we can write 𝑋 = 𝑈 ∪ 𝑉 where 𝑈 , 𝑉 are open,
nonempty and disjoint. It follows, then, that 𝑈 = 𝑉 𝐶 (and vice versa) and hence 𝑈 , 𝑉 are both open and closed.

↩→ Definition 2.17: Connected Component

A connected component of 𝑥 ∈ 𝑋 is the largest connected subset of 𝑋 that contains 𝑥.

⊛ Example 2.6
Let 𝑋 = (0, 1) ∪ (1, 2). Here, we have two connected components, (0, 1) and (1, 2)

⊛ Example 2.7: Middle Thirds Cantor Set

1
Let 𝐶0 := [0, 1], and given 𝐶 𝑛 , define 𝐶 𝑛+1 := 3 (𝐶 𝑛 ∪ (2 + 𝐶 𝑛 )) for 𝑛 ⩾ 0. 𝐶∞ is totally disconnected.

↩→ Definition 2.18: Path Component

A path component 𝑃(𝑥) of 𝑥 ∈ 𝑋 is the largest path connected subset of 𝑋 that contains 𝑥.

↩→Proposition 2.19
𝑃(𝑥) = {𝑥 ∈ 𝑋 : ∃ conintuous path 𝛾 : [0, 1] → 𝑋 : 𝛾(0) = 𝑥, 𝛾(1) = 𝑦}.

Remark 2.12. Where we “start” a path does not matter. We write 𝑥 ∼ 𝑦 if ∃𝛾 from 𝑥 to 𝑦; this is an equivalence
relation on the elements of 𝑋.

Remark 2.13. The choice of [0, 1] here is arbitrary; any closed interval is homeomorphic.

↩→ Lemma 2.4
If 𝑃(𝑥) ∩ 𝑃(𝑦) ≠ ∅, then 𝑃(𝑥) = 𝑃(𝑦).

Proof 𝑃(𝑥) ∩ 𝑃(𝑦) ≠ ∅ =⇒ ∃𝑧 : 𝑥 ∼ 𝑧 ∧ 𝑦 ∼ 𝑧 =⇒ 𝑥 ∼ 𝑦.

Proof
Proof. ■

↩→ Lemma 2.5

If 𝐴 ⊆ 𝑋 is connected, then 𝐴 is also connected.

↩→ Lemma 2.6
Suppose 𝐴 ⊆ 𝑋 is both open and closed. Then, if 𝐶 ⊆ 𝑋 is connected and 𝐶 ∩ 𝐴 ≠ ∅, then 𝐶 ⊆ 𝐴.

2.8 Point-Set Topology: Path Components, Connected Components 21

Proof. If 𝐴 is both open and closed, then 𝐶 ∩ 𝐴 is both open and closed in 𝐶. If 𝐶 ∩ 𝐴𝐶 ≠ ∅, then this is also
Proof
open and closed in 𝐶. Hence, we can write 𝐶 = (𝐶 ∩ 𝐴) ∪ (𝐶 ∩ 𝐴𝐶 ), that is, a disjoint union of two nonempty
open sets, contradicting the connectedness of 𝐶. Hence, 𝐶 ∩ 𝐴𝐶 = ∅, and so 𝐶 ⊆ 𝐴. ■

↩→Proposition 2.20
Let {𝐶 𝛼 } 𝛼∈𝐼 be a collection of nonempty connected subspaces of 𝑋 s.t. ∀ 𝛼, 𝛽 ∈ 𝐼, 𝐶 𝛼 ∩ 𝐶𝛽 ≠ ∅. Then,
∪𝛼∈𝐼 𝐶 𝛼 is connected.

↩→Proposition 2.21
Suppose each 𝑥 ∈ 𝑋 has a path-connected neighborhood. Then, the path components in 𝑋 are the same
as the connected components in 𝑋.

2.8.1 Cantor Staircase Function

↩→ Definition 2.19: An Explicit Definition

0



Let 𝑥 ∈ 𝐶 : 𝑥 = 0.𝑎1 𝑎 2 𝑎 3 . . . (base 3), ie 𝑎 𝑗 = . Define
2



 𝑎 𝑗 /2 𝑥∈𝐶

 Í
2𝑗

𝑓 (𝑥) =
 extend by continuity 𝑥 ∉ 𝐶.



That is, if 𝑥 ∉ 𝐶, set 𝑓 (𝑦) = sup𝑥∈𝐶,𝑥<𝑦 𝑓 (𝑥) = inf𝑥∈𝐶,𝑥>𝑦 𝑓 (𝑥).

↩→ Definition 2.20: Complement Definition

To construct the complement of the Cantor set, begin with [0, 1] and at a step 𝑛, we remove 2𝑛 open
𝑘
intervals from this interval. 𝑓 (𝑥) will be constant on each of these intervals with values 2𝑛 where 𝑘 odd
and 0 < 𝑘 < 2𝑛 . Extend by continuity to all 𝑥 ∈ 𝐶.

Remark 2.14. Wikipedia’s explanation of this is far better than whatever this definition is trying to say.

↩→ Lecture 06; Last Updated: Tue Jan 23 11:03:35 EST 2024

2.8 Point-Set Topology: Path Components, Connected Components 22

3 𝐿 𝑝 Spaces

3.1 Review of ℓ 𝑝 Norms

Remark 3.1. Recall that for 1 ⩽ 𝑝 ⩽ +∞, we define for 𝑥 = (𝑥1 , . . . , 𝑥 𝑛 ) ∈ R𝑛 the norm

1 𝑛
||𝑥|| 𝑝 = (|𝑥1 | 𝑝 + · · · + |𝑥 𝑛 | 𝑝 ) 𝑝 , ||𝑥|| ∞ = max |𝑥 𝑖 | .
𝑖=1

Similarly, for infinite vector spaces, we had, for 𝑥 = (𝑥1 , . . . , 𝑥 𝑛 , . . . ), the norm

∞
! 𝑝1
|𝑥 𝑖 | 𝑝
Õ
||𝑥|| 𝑝 = , ||𝑥|| ∞ = sup |𝑥 𝑖 | .
𝑖=1 𝑖 ⩾1

Here, we define
ℓ 𝑝 := {𝑥 = (𝑥1 , . . . , 𝑥 𝑛 ) : ||𝑥|| 𝑝 < +∞}.

3.2 ℓ 𝑝 Norms, Hölder-Minkowski Inequalities

↩→ Definition 3.1: Hölder Conjugates

For 1 ⩽ 𝑝, 𝑞 ⩽ +∞, we say that 𝑝, 𝑞 are said to be Hölder conjugates if

1 1
+ = 1.
𝑝 𝑞

Remark 3.2. We refer to these simply as “conjugates” throughout as no other conception of conjugate numbers will be
discussed.
1
Further, we take by convention ∞ = 0.

↩→Proposition 3.1: Hölder’s Inequality

Let 𝑥 = (𝑥1 , . . . , 𝑥 𝑛 ), 𝑦 = (𝑦1 , . . . , 𝑦𝑛 ) ∈ R𝑛 . Suppose 𝑝, 𝑞 : 1 ⩽ 𝑝, 𝑞 ⩽ +∞ are conjugate. Then,

𝑛
Õ
⟨𝑥, 𝑦⟩R𝑛 := 𝑥 𝑖 𝑦 𝑖 ⩽ ||𝑥|| 𝑝 · ||𝑦|| 𝑞
𝑖=1

⊛ Example 3.1
For the case 𝑝 = 1 or ∞ (functionally, the same case):

3.2 𝐿 𝑝 Spaces: ℓ 𝑝 Norms, Hölder-Minkowski Inequalities 23

 ↩→ Lemma 3.1
Let 𝑝, 𝑞 be conjugates, and 𝑥, 𝑦 ⩾ 0. Then,

𝑥𝑝 𝑦𝑞
𝑥𝑦 ⩽ + .
𝑝 𝑞
1 1
Remark 3.3. If the inequality holds, then, for some 𝑡 > 0, let 𝑥˜ = 𝑡 𝑝 · 𝑥, 𝑦˜ = 𝑡 𝑞 𝑦. Substituting 𝑥 for 𝑥˜ and 𝑦 for 𝑦,
˜ we
have
1 1 1
+ 1𝑞
LHS: 𝑥˜ 𝑦˜ = 𝑡 𝑝 𝑥 · 𝑡 𝑞 𝑦 = 𝑡 𝑝 · 𝑥𝑦 = 𝑥𝑦
𝑥𝑝 𝑦𝑞
RHS: · · · = 𝑡( + .)
𝑝 𝑞

That is, we have

𝑥𝑝 𝑦𝑞
 
𝑡 · 𝑥𝑦 ⩽ 𝑡 + ,
𝑝 𝑞
hence, the inequality is preserved under multiplication by a positive scalar;
  𝑞 moreover, the original inequality holds iff
this “scaled” version holds. Hence, choosing 𝑡 such that 𝑦˜ = 1 (let 𝑡 = 𝑦1 ), it suffices to prove the lemma for 𝑦 = 1.

Proof If 𝑥 = 0 or 𝑦 = 0, then the entire LHS becomes 0 and we are done; assume 𝑥, 𝑦 > 0; by the previous
Proof
Proof.
remark, assume wlog 𝑦 = 1. Then, we have

𝑥𝑝 𝑦𝑞 𝑥𝑝 1
𝑥·𝑦⩽ + ⇐⇒ 𝑥 · 1 ⩽ +
𝑝 𝑞 𝑝 𝑞
𝑥 𝑝 1
⇐⇒ − 𝑥 + =: 𝑓 (𝑥) ⩾ 0.
𝑝 𝑞

Taking the derivative, we have

𝑝𝑥 𝑝−1
′
𝑓 (𝑥) =  − 1 = 𝑥 𝑝−1 − 1
𝑝


𝑓 ′(𝑥) > 0 ∀𝑥 > 1








𝑝 > 1 =⇒ 𝑝 − 1 > 0 =⇒ 𝑓 ′(𝑥) = 0 𝑥=0


 𝑓 ′(𝑥) < 0 ∀0 < 𝑥 < 1



1𝑝 1
Hence, 𝑥 = 1 is a local minimum of the function, and thus 𝑓 (𝑥) ⩾ 𝑓 (1) ∀ 0 < 𝑥 ⩽ 1. But 𝑓 (1) = 𝑝 −1+ 𝑞 =
1 − 1 = 0, hence 𝑓 (𝑥) ⩾ 0 ∀ 𝑥 ⩾ 0, as desired, and the inequality holds. ■

3.2 𝐿 𝑝 Spaces: ℓ 𝑝 Norms, Hölder-Minkowski Inequalities 24

Proof. Assume ||𝑥|| 𝑝 = ||𝑦|| 𝑞 = 1. Then,
Proof

𝑛
Õ 𝑛
Õ
𝑥 𝑖 𝑦𝑖 ⩽ |𝑥 𝑖 𝑦 𝑖 | (by triangle inequality)
𝑖=1 𝑖=1
𝑛 𝑝 𝑞
Õ 𝑥𝑖 𝑦𝑖
⩽ + (by lemma 3.1)
𝑝 𝑞
𝑖=1
𝑛 𝑛
! !
1 Õ 1 Õ
= |𝑥 𝑖 | 𝑝 + |𝑦 𝑖 | 𝑞
𝑝 𝑞
𝑖=1 𝑖=1
1 𝑝 1 𝑞
= ||𝑥|| 𝑝 + ||𝑦|| 𝑞 (by staring)
𝑝 𝑞
1 1 1 1
= · 1 𝑝 + · 11 = + = 1 (by assumption)
𝑝 𝑞 𝑝 𝑞
= ||𝑥|| 𝑝 · ||𝑦|| 𝑞 ,

and the proposition holds, in the special case ||𝑥|| 𝑝 = ||𝑦|| 𝑞 = 1.

If ||𝑥|| 𝑝 = 0 or ||𝑦|| 𝑞 = 0, then 𝑥 1 = · · · = 𝑥 𝑛 = 0 or 𝑦1 = · · · = 𝑦𝑛 = 0, resp, then we’d have (||𝑥|| 𝑝 = 0 case)

0 · 𝑦1 + · · · + 0 · 𝑦𝑛 ⩽ 0,

which clearly holds.

𝑥 𝑦
Assume, then, ||𝑥|| 𝑝 > 0, ||𝑦|| 𝑞 > 0. Let 𝑥˜ := ||𝑥|| 𝑝
, 𝑦˜ := ||𝑦|| 𝑞
. Then,

𝑝
|𝑥 𝑖 | 𝑝
Í𝑛

𝑝 𝑖=1
||𝑥|| 𝑝
|| 𝑥||
˜ 𝑝 = 𝑝 = 𝑝 = 1 =⇒ || 𝑥||
˜ 𝑝 = 1.
||𝑥|| 𝑝 ||𝑥|| 𝑝

The same case holds for 𝑦,

˜ hence || 𝑦||
˜ 𝑞 = 1; that is, we have “rescaled” both vectors. Hence, we can use the
case we proved above for when the norms were identically 1 on 𝑥, ˜ 𝑦.
˜ We have:

𝑛
Õ
𝑥˜ 𝑖 𝑦˜𝑖 ⩽ 1
𝑖=1

But by definition of 𝑥,
˜ 𝑦,
˜ we have

𝑛 𝑛 𝑛
Õ 1 Õ Õ
𝑥˜ 𝑖 𝑦˜𝑖 = 𝑥 𝑖 𝑦 𝑖 ⩽ 1 =⇒ 𝑥 𝑖 𝑦 𝑖 ⩽ ||𝑥|| 𝑝 · ||𝑦|| 𝑞 ,
||𝑥|| 𝑝 ||𝑦|| 𝑞
𝑖=1 𝑖=1 𝑖=1

and the proof is complete. ■

3.2 𝐿 𝑝 Spaces: ℓ 𝑝 Norms, Hölder-Minkowski Inequalities 25

 ↩→Proposition 3.2: Minkowski Inequality
Let 1 ⩽ 𝑝 ⩽ ∞, 𝑥, 𝑦 ∈ R𝑛 . Then,
||𝑥 + 𝑦|| 𝑝 ⩽ ||𝑥|| 𝑝 + ||𝑦|| 𝑝 .

Remark 3.4. This is just the triangle inequality for ℓ 𝑝 norms.

Proof The cases 𝑝 = 1, ∞ are left as an exercise.

Proof
Proof.
Assume 1 < 𝑝 < ∞. Then,

𝑛 𝑛
𝑝 𝑝 𝑝−1
Õ Õ
||𝑥 + 𝑦|| 𝑝 = 𝑥 𝑗 + 𝑦𝑗 = 𝑥 𝑗 + 𝑦𝑗 𝑥 𝑗 + 𝑦𝑗
𝑗=1 𝑗=1
∞
𝑝−1
Õ
𝑥 𝑗 + 𝑦𝑗 · 𝑥 𝑗 + 𝑦𝑗


⩽
𝑗=1
:=𝐵
z }| {
𝑛 𝑛
𝑝−1 𝑝−1
Õ Õ
= 𝑥 𝑗 · 𝑥 𝑗 + 𝑦𝑗 + 𝑦𝑗 · 𝑥 𝑗 + 𝑦𝑗 ⊛
𝑗=1 𝑗=1
| {z }
:=𝐴

Let 𝑢® = (|𝑥1 | , · · · , |𝑥 𝑛 |) and 𝑣® = (|𝑥1 + 𝑦1 | 𝑝−1 , · · · , |𝑥 𝑛 + 𝑦𝑛 | 𝑝−1 ), then, 𝐴 = 𝑢® · 𝑣® = ⟨®

𝑢 , 𝑣® ⟩R𝑛 . We have

𝑛
! 𝑝1
(|𝑥 𝑖 | 𝑝 )
Õ
|| 𝑢® || 𝑝 = = ||𝑥|| 𝑝
𝑖=1
𝑛 
! 1𝑞
𝑞
|𝑥 𝑖 + 𝑦 𝑖 | 𝑝−1
Õ
𝑣 || 𝑞 =
||®
𝑖=1
" 𝑛  𝑝
# 𝑝−1
𝑝
 𝑝−1
|𝑥 𝑖 + 𝑦 𝑖 | 𝑝−1
Õ
=
𝑖=1
" 𝑛
# 𝑝−1
𝑝

|𝑥 𝑖 + 𝑦 𝑖 | 𝑝
Õ
=
𝑖=1
𝑝−1
= ||𝑥 + 𝑦|| 𝑝

𝑝
where the second-to-last line follows from 𝑝, 𝑞 being conjugate, hence 𝑞 = 𝑝−1 . Thus, by Hölder’s Inequality,
we have that
𝑝−1
𝐴 = ⟨®
𝑢 , 𝑣® ⟩ ⩽ ||𝑢|| 𝑝 · ||𝑣|| 𝑞 = ||𝑥|| 𝑝 · ||𝑥 + 𝑦|| 𝑝 .
By a similar construction, we can show that

𝑝−1
𝐵 ⩽ ||𝑦|| 𝑝 · ||𝑥 + 𝑦|| 𝑝 .

3.2 𝐿 𝑝 Spaces: ℓ 𝑝 Norms, Hölder-Minkowski Inequalities 26

Thus, returning to our original inequality in ⊛ , we have

𝑝
||𝑥 + 𝑦|| 𝑝 ⩽ 𝐴 + 𝐵
𝑝−1 𝑝−1
⩽ ||𝑥|| 𝑝 · ||𝑥 + 𝑦|| 𝑝 + ||𝑦|| 𝑝 · ||𝑥 + 𝑦|| 𝑝
=⇒ ||𝑥 + 𝑦|| 𝑝 ⩽ ||𝑥|| 𝑝 + ||𝑦|| 𝑝 ,

and the proof is complete. ■

↩→ Lecture 07; Last Updated: Tue Jan 30 12:54:59 EST 2024

3.3 Complete Metric Spaces, Completeness of ℓ 𝑝

↩→ Theorem 3.1
The sequence of centers of balls with monotonically decreasing radii is a Cauchy sequence in 𝑋.

Proof Let 𝜀 > 0 and let 𝑁 : ∀ 𝑗 > 𝑁 , 𝑟 𝑗 < 𝜀. Then,

Proof
Proof.

𝑑(𝑥 𝑗 , 𝑥 𝑘 ) < 𝑟min(𝑗,𝑘) = 𝑟 𝑗

↩→ Definition 3.2: Complete Metric Space

A metric space is complete if every Cauchy sequence converges to a limit in that space.

⊛ Example 3.2: Examples of Complete Metric Spaces

1. R, 𝑝-adic integers (Z𝑝 )/rationals(Q𝑝 ).

|𝑥 𝑖 | 𝑝 < +∞}, 1 ⩽ 𝑝 ⩽ +∞
Í∞
2. ℓ 𝑝 = {𝑥 = (𝑥 𝑖 )∞
𝑖=1
: 𝑖=1

3. ℓ∞ = {𝑥 = (𝑥 𝑖 ) : sup∞
𝑖=1 |𝑥 𝑖 |
< +∞}.

↩→Proposition 3.3
Hölder’s Inequality and Minkowski Inequality inequalities hold for infinite sequences. that is,

1 1
1. if 𝑥 = (𝑥 𝑖 ) ∈ ℓ 𝑝 and 𝑦 = (𝑦 𝑖 ) ∈ ℓ 𝑞 with 𝑝 + 𝑞 = 1, then

∞
Õ
𝑥 𝑖 𝑦 𝑖 ⩽ ||𝑥 𝑖 ||ℓ 𝑝 ||𝑦 𝑖 ||ℓ 𝑞 .
𝑖=1

2. if 𝑥, 𝑦 ∈ ℓ 𝑝 , then
||𝑥 + 𝑦|| 𝑝 ⩽ ||𝑥|| 𝑝 + ||𝑦|| 𝑝 .

3.3 𝐿 𝑝 Spaces: Complete Metric Spaces, Completeness of ℓ 𝑝 27

Remark 3.5. 2. gives the triangle inequality for the ||𝑥|| 𝑝 norm on ℓ 𝑝 .
Moreover,

||𝑐 · 𝑥|| 𝑝 = ||(𝑐 1 𝑥1 , . . . , 𝑐 𝑛 𝑥 𝑛 , . . . )|| 𝑝

∞
! 𝑝1 ∞
! 𝑝1
|𝑐𝑥 𝑖 | 𝑝 |𝑐| 𝑝 |𝑥 𝑖 | 𝑝
Õ Õ
= =
𝑖=1 𝑖=1
1
𝑝
𝑝
= |𝑐| ||𝑥|| 𝑝 = 𝑐 · ||𝑥|| 𝑝

𝑝 𝑝 𝑝
Proof (of 2.) If 𝑥, 𝑦 ∈ ℓ 𝑝 , we have that ∞ ∞
𝑖=1 |𝑥 𝑖 | < +∞, 𝑖=1 |𝑦 𝑖 | < +∞, so ∃𝑁 > 0 :
∞
𝑖=𝑁+1 |𝑥 𝑖 | <
Í Í Í
Proof
Proof.
𝑝 (𝑛)
𝜀, ∞𝑖=𝑁+1 |𝑦 𝑖 | < 𝜀. Let 𝑥 𝑖 = (𝑥 1 , . . . , 𝑥 𝑛 , 0, 0, . . . ) be (𝑥) truncated after 𝑛 (finite) coordinates. This gives
Í

(𝑛) (𝑛)
||(𝑥 𝑖 + 𝑦 𝑖 )(𝑛) || 𝑝 ⩽ ||𝑥 𝑖 || 𝑝 + ||𝑦 𝑖 || 𝑝 ⩽ ||𝑥|| 𝑝 + ||𝑦|| 𝑝

by Minkowski on finite spaces. Taking 𝑛 → ∞ (ie, “detruncating”), we have (𝑥 + 𝑦) ∈ ℓ 𝑝 , and thus ||𝑥 + 𝑦|| 𝑝 ⩽
||𝑥|| 𝑝 + ||𝑦|| 𝑝 .
1. left as an exercise. ■

↩→Proposition 3.4
Let 1 ⩽ 𝑝 ⩽ +∞, and ||𝑥|| ∞ = sup∞
𝑖=1 |𝑥 𝑖 |
= 𝐴 < +∞, ||𝑦|| ∞ = sup∞
𝑖=1 |𝑦 𝑖 |
= 𝐵 < +∞. Then, the triangle
inequality ||𝑥 + 𝑦|| ∞ ⩽ ||𝑥|| ∞ + ||𝑦|| ∞ holds.

Proof We have
Proof
Proof.

∞ ∞ ∞ ∞
sup |𝑥 𝑖 + 𝑦 𝑖 | ⩽ sup (|𝑥 𝑖 | + |𝑦 𝑖 |) ⩽ sup |𝑥 𝑖 | + sup |𝑦 𝑖 | = ||𝑥|| ∞ + ||𝑦|| ∞ .
𝑖=1 𝑖=1 𝑖=1 𝑖=1

↩→Proposition 3.5
||𝑥|| ∞ := sup∞
𝑖=1 |𝑥 𝑖 |
is a well-defined norm on ℓ∞ .

Proof. The triangle inequality is prove in proposition 3.4. The remainder of the requirements are left as an
Proof
exercise. ■

↩→Proposition 3.6
ℓ 𝑝 ⊆ ℓ 𝑞 if 𝑝 < 𝑞.

3.3 𝐿 𝑝 Spaces: Complete Metric Spaces, Completeness of ℓ 𝑝 28

 |𝑥 𝑖 | 𝑝 < +∞, then ∃𝑁 : ∀ 𝑖 ⩾ 𝑁 , |𝑥 𝑖 | ⩽ 1. Then,
Í∞
Proof. Let 𝑥 ∈ ℓ 𝑝 . If
Proof 𝑖=1

|𝑥 𝑖 | 𝑞 ⩽ |𝑥 𝑖 | 𝑝 < ∞
Õ Õ

𝑖 ⩾𝑁 𝑖 ⩾𝑁
∞
|𝑥 𝑖 | 𝑞 < +∞ =⇒ 𝑥 ∈ ℓ 𝑞
Õ
=⇒
𝑖=1

=⇒ ℓ 𝑝 ⊆ ℓ 𝑞

■
↩→ Lecture 08; Last Updated: Thu Mar 28 09:13:10 EDT 2024

3.4 Contraction Mapping Theorem

↩→ Definition 3.3: Contraction Mapping

Let (𝑋 , 𝑑) be a metric space. A contraction mapping on 𝑋 is a function 𝑓 : 𝑋 → 𝑋 for which ∃ a constant
0 < 𝑐 < 1 such that
𝑑( 𝑓 (𝑥), 𝑓 (𝑦)) ⩽ 𝑐 · 𝑑(𝑥, 𝑦) ∀ 𝑥, 𝑦 ∈ 𝑋.

↩→ Theorem 3.2: Contraction Mapping Theorem

Let (𝑋 , 𝑑) be a complete metric space, and let 𝑓 : 𝑋 → 𝑋 be a contraction. Then, there exists a unique
fixed point 𝑧 of 𝑓 such that 𝑓 (𝑧) = 𝑧.

Moreover, 𝑓 [𝑛] (𝑥) := 𝑓 ◦ 𝑓 ◦ · · · ◦ 𝑓 (𝑥) → 𝑧 as 𝑛 → ∞ for any 𝑥 ∈ 𝑋.

Remark 3.6. The “functional construction” of the Cantor set is an example of a contraction mapping, with 𝑓1 (𝑥) =
𝑥 𝑥+2
3 , 𝑓2 (𝑥) = 3 . The first has a fixed point of 0, and the second a fixed point of 1.

Remark 3.7. This is a generalization of this proof done in Analysis I, an equivalent claim over the reals.

Proof Fix 𝑥 ∈ 𝑋. Consider the sequence {𝑥 0 , 𝑥1 , 𝑥2 , . . . , 𝑥 𝑛 , . . . } := {𝑥, 𝑓 (𝑥), 𝑓 ◦ 𝑓 (𝑥), · · · , 𝑓 [𝑛] (𝑥), . . . } (we call
Proof
Proof.
𝑓 [𝑛] the orbit of 𝑥 under iterations of 𝑓 ). We claim that this is a Cauchy sequence. Let 𝑛 ∈ N arbitrary, then
we have, by the property of the contraction mapping,

𝑑( 𝑓 [𝑛+1] (𝑥) − 𝑓 [𝑛] (𝑥)) ⩽ 𝑐 · 𝑑( 𝑓 [𝑛] (𝑥) − 𝑓 [𝑛−1] (𝑥)) ⩽ 𝑐 2 𝑑( 𝑓 [𝑛−1] (𝑥) − 𝑓 [𝑛−2] (𝑥)).

Arguing inductively, it follows that

𝑑( 𝑓 [𝑛+1] (𝑥) − 𝑓 [𝑛] (𝑥)) ⩽ 𝑐 𝑛 𝑑( 𝑓 (𝑥), 𝑥). ★

3.4 𝐿 𝑝 Spaces: Contraction Mapping Theorem 29

Let now 𝑚, 𝑘 ∈ N, 𝑚, 𝑘 > 0. It follows that

𝑑( 𝑓 [𝑚] , 𝑓 [𝑚+𝑘] (𝑥) ⩽ 𝑑( 𝑓 [𝑚] )(𝑥), 𝑓 [𝑚+1] (𝑥)) + 𝑑( 𝑓 [𝑚+1] (𝑥), 𝑓 [𝑚] (𝑥)) + · · · + 𝑑( 𝑓 [𝑚+𝑘−1] (𝑥), 𝑓 𝑚+𝑘 (𝑥))
★
⩽ 𝑑(𝑥, 𝑓 (𝑥))[𝑐 𝑚 + 𝑐 𝑚+1 + · · · + 𝑐 𝑚+𝑘−1 ]
𝑐 𝑚 𝑑(𝑥, 𝑓 (𝑥))
⩽ 𝑐 𝑚 𝑑(𝑥, 𝑓 (𝑥))[1 + 𝑐 + · · · + 𝑐 𝑘 + 𝑐 𝑘+1 + · · · ] =
1−𝑐

𝑐 𝑁 𝑑(𝑥, 𝑓 (𝑥))
Now, given 𝜀 > 0, choose 𝑁 such that 1−𝑐 < 𝜀. It follows, then, that { 𝑓 [𝑛] (𝑥)} 𝑛∈N a Cauchy sequence,
and thus converges, 𝑓 [𝑛] (𝑥) → 𝑧 as 𝑛 → ∞ for some 𝑧.
We further have to show that 𝑓 (𝑧) = 𝑧. It is easy to show that 𝑓 continuous due to the contraction mapping
(it is clearly Lipschitz with constant 𝑐), and it thus follows that

lim 𝑓 ( 𝑓 [𝑛] (𝑥)) = lim 𝑓 [𝑛] (𝑥) =⇒ 𝑓 (𝑧) = 𝑧,

𝑛→∞ 𝑛→∞

by sequential characterization of continuous functions.

Finally, we need to show that this limit is unique. Suppose ∃𝑦1 ≠ 𝑦2 , ie two fixed points with 𝑓 (𝑦1 ) = 𝑦1
and 𝑓 (𝑦2 ) = 𝑦2 . Then, by the property of the contraction mapping,

𝑑( 𝑓 (𝑦1 ), 𝑓 (𝑦2 )) ⩽ 𝑐 · 𝑑(𝑦1 , 𝑦2 ),

but by assumption of being fixed points,

𝑑( 𝑓 (𝑦1 ), 𝑓 (𝑦2 )) = 𝑑(𝑦1 , 𝑦2 ),

implying 𝑑(𝑦1 , 𝑦2 ) ⩽ 𝑐 · 𝑑(𝑦1 , 𝑦2 ). This is only possible if 𝑑(𝑦1 , 𝑦2 ) = 0, and thus 𝑦1 = 𝑦2 and the fixed point is
indeed unique. ■

↩→ Theorem 3.3: ℓ 𝑝 complete

The space ℓ 𝑝 is complete for all 1 ⩽ 𝑝 ⩽ +∞.

Equivalently, if (𝑥 1 ), (𝑥 2 ), . . . , (𝑥 𝑛 ) is a Cauchy sequence in ℓ 𝑝 , ∃𝑦 ∈ ℓ 𝑝 s.t. 𝑥 𝑛 → 𝑦 as 𝑛 → ∞.

Proof (Sketch) We suppose first 𝑝 < +∞. Consider an arbitrary number of Cauchy sequences in ℓ 𝑝 :
Proof
Proof.

(1) (1)
𝑥 (1) = (𝑥 1 , . . . , 𝑥 𝑛 , . . . )
(2) (2)
𝑥 (2) = (𝑥 1 , . . . , 𝑥𝑛 , . . . )
.. .. ..
. . .
(𝑘) (𝑘)
𝑥 (𝑘) = (𝑥1 , . . . , 𝑥 𝑛 , . . . ) ∈ ℓ 𝑝

(𝑛)
We claim that, for any 𝑘 ∈ N, the (𝑥 𝑘 )𝑛∈N is a Cauchy sequence; note that in this definition we are taking a
fixed-index (namely, the 𝑘th) element from different sequences (namely, the 𝑛th sequence).

3.4 𝐿 𝑝 Spaces: Contraction Mapping Theorem 30

 Since 𝑥 (1) , 𝑥 (2) , . . . , 𝑥 (𝑛) , . . . are Cauchy sequences in ℓ 𝑝 , we have for a fixed 𝜀 > 0, ∃𝑁 ∈ N : ∀ 𝑚, 𝑛 > 𝑁,
𝑑 𝑝 (𝑥 (𝑚) , 𝑥 (𝑛) ) < 𝜀:

∞ 𝑝
𝑝
Õ
(𝑛) 𝑝
< 𝜀𝑝
(𝑚) (𝑚) (𝑚) (𝑛)
𝑑 𝑝 (𝑥 ,𝑥 ) = ||𝑥 − 𝑥 (𝑛) || 𝑝 = 𝑥𝑖 − 𝑥𝑖
𝑖=1
𝑝 ∞
Õ 𝑝 𝑝
− 𝑥 𝑛𝑘 − 𝑥 𝑛𝑘 < 𝜀𝑝
(𝑚) (𝑚) (𝑛) (𝑚)
𝑥𝑘 ⩽ 𝑥𝑖 − 𝑥𝑖 =⇒ 𝑥 𝑘
𝑖=1
(𝑚) (𝑛)
=⇒ 𝑥 𝑘 − 𝑥𝑘 < 𝜀,

(𝑛)
since we are taking “less of the summands in the second line”. It follows, then, that for each 𝑘, ∃𝑧 𝑘 : 𝑥 𝑘 → 𝑧 𝑘
as 𝑛 → ∞. Let 𝑧 = (𝑧 1 , . . . , 𝑧 𝑛 , . . . ). We claim that 𝑥 (𝑛) → 𝑧 ∈ ℓ 𝑝 as 𝑛 → ∞.
First, we show that 𝑑 𝑝 (𝑥 (𝑛) , 𝑧) → 0 as 𝑛 → 0 (that is, 𝑥 (𝑛) → 𝑧 as 𝑛 → ∞). Fix 𝜀 > 0, and choose 𝑁 ∈ N for
which 𝑑 𝑝 (𝑥 (𝑚) , 𝑥 (𝑛) ) < 𝜀 ∀ 𝑚, 𝑛 ⩾ 𝑁 (by Cauchy). Fix 𝐾 ∈ N, 𝐾 > 0.

∞ 𝑝
𝑝 𝑝
Õ
(𝑛)
𝑑 𝑝 (𝑥 (𝑛) , 𝑧) = ||𝑥 (𝑛) − 𝑧|| 𝑝 = 𝑥𝑖 − 𝑧𝑖
𝑖=1
𝐾 𝑝
𝑝
Õ
||𝑥 (𝑚) − 𝑥 (𝑛) || 𝑝 < 𝜀 𝑝 =⇒ ⩽ 𝜀𝑝
(𝑚) (𝑛)
𝑥𝑖 − 𝑥𝑖
𝑖=1

(𝑚)
Let 𝑚 → ∞; then 𝑥 𝑖 → 𝑧 𝑖 (note that 𝑖 fixed!), and we have

𝐾
Õ 𝑝
𝑧𝑖 − 𝑥𝑖
(𝑛)
⩽ 𝜀𝑝 .
𝑖=1

Let 𝐾 → ∞; then,
∞
Õ 𝑝
𝑧𝑖 − 𝑥𝑖
(𝑛)
⩽ 𝜀 𝑝 =⇒ ||𝑧 − 𝑥|| 𝑝 ⩽ 𝜀 =⇒ 𝑑 𝑝 (𝑧, 𝑥 𝑛 ) ⩽ 𝜀,
𝑖=1

and thus 𝑥 𝑛 → 𝑧 as 𝑛 → ∞.
It remains to show that 𝑧 ∈ ℓ 𝑝 , ie ||𝑧|| 𝑝 < +∞. We have:

||𝑧|| 𝑝 ⩽ ||𝑧 − 𝑥 (𝑛) || 𝑝 +||𝑥 (𝑛) || 𝑝 .

| {z }
→0

For sufficiently large 𝑛, ||𝑧 − 𝑥 (𝑛) || ⩽ 1 (for instance); 𝑥 (𝑛) ∈ ℓ 𝑝 , hence ||𝑥 (𝑛) || 𝑝 < +∞ (say, ||𝑥 (𝑛) || 𝑝 ⩽ 𝑀). Thus:

||𝑧|| 𝑝 ⩽ 1 + 𝑀 < +∞ =⇒ 𝑧 ∈ ℓ 𝑝 ,

and the proof is complete. ■

3.4 𝐿 𝑝 Spaces: Contraction Mapping Theorem 31

3.5 Equivalent Notions of Compactness in Metric Spaces

↩→ Definition 3.4: Totally Bounded

Ð𝑛
Let (𝑋 , 𝑑) be a metric space. If for every 𝜀 > 0, ∃𝑥1 , . . . , 𝑥 𝑛 ∈ 𝑋 , 𝑛 = 𝑛(𝜀) : 𝑖=1 𝐵(𝑥 𝑖 , 𝜀) = 𝑋, we say 𝑋 is
totally bounded.

↩→ Lecture 09; Last Updated: Tue Apr 9 22:27:24 EDT 2024

↩→ Theorem 3.4
Let (𝑋 , 𝑑) be a metric space. TFAE:

1. 𝑋 is complete and totally bounded;

2. 𝑋 is compact;

3. 𝑋 is sequentially compact (every sequence has a convergent subsequence).

Proof (1. =⇒ 2.) Suppose 𝑋 complete and totally bounded. Assume towards a contradiction that 𝑋 not
Proof
Proof.
compact, ie there exists an open cover {𝑈 𝛼 } 𝛼∈𝐼 of 𝑋 with no finite subcover.
𝑋 being totally bounded gives that it can be covered by finitely many open balls of radius 12 . It must be
that at least one of these open balls cannot be finitely covered, otherwise we would have a finite subcover.
Let 𝐹1 be the closure of this ball. 𝐹1 closed, with diameter diam(𝐹1 ) ⩽ 1. 𝑋.
We also have that 𝑋 can be covered by finitely many balls of radius 14 ; again, there must be at least one
ball 𝐵1 such that 𝐵1 ∩ 𝐹1 cannot be covered by finitely many open sets from the cover. Let 𝐹2 = 𝐵1 ∩ 𝐹1 -closed,
with diam(𝐹2 ) ⩽ 41 + 14 = 12 .4
Arguing inductively, at some step 𝑛, 𝑋 can can be covered by finitely many balls of radius 21𝑛 ; at least one
of these balls 𝐵 cannot be covered by a finite subcover hence 𝐵 ∩ 𝐹𝑛−1 cannot be covered by finitely many 𝑈 𝛼 ’s.
1
Let 𝐹𝑛 = 𝐵 ∩ 𝐹𝑛−1 -closed, with diam(𝐹𝑛 ) ⩽ 2𝑛−1 .
1
As such, we have a nested sequence 𝐹1 ⊇ 𝐹2 ⊇ · · · ⊇ 𝐹𝑛 ⊇ · · · of closed sets, where diam(𝐹 𝑘 ) ⩽ 2 𝑘−1
→0
as 𝑘 → ∞.
Ñ∞
↩→ Lemma 3.1 (Cantor Intersection Theorem). 𝑘=1 𝐹 𝑘 ≠ ∅.

Proof (Of Lemma) Let 𝑥 𝑘 ∈ 𝐹 𝑘 . Then, {𝑥 𝑘 } 𝑘∈N is a Cauchy sequence, since

Proof
Proof.

1
𝑑(𝑥 𝑛 , 𝑥 𝑛+𝑘 ) ⩽ diam(𝐹𝑛 ) + · · · + diam(𝐹𝑛+𝑘 ) ⩽ ,
2𝑛−1

by the nested property, which can be made arbitrarily small for sufficiently large 𝑛, 𝑘. Hence, 𝑥 𝑛 → 𝑦 ∈ 𝑋
for some 𝑦, as 𝑋 complete. The tail of 𝑥 𝑛 lies in 𝐹𝑛 for all sufficiently large 𝑛, and as each 𝐹𝑛 closed, the limit
must lie in 𝐹𝑛 for all sufficiently large 𝑛. We conclude the intersection nonempty. ■
1
4𝐵1 has radius 4 and hence diameter 12 . The intersection of 𝐵1 with a set with a larger diameter must have diameter leq 1
2

3.5 𝐿 𝑝 Spaces: Equivalent Notions of Compactness in Metric Spaces 32

 This 𝑦 from the lemma is covered by some 𝑈 𝛼0 -open for some 𝛼0 ∈ 𝐼. Being open, ∃𝜀 > 0 : 𝐵(𝑦, 𝜀) ⊆ 𝑈 𝛼0 .
1
Let 𝑛 : 2𝑛1−1 < 𝜀. Then, 𝑦 ∈ 𝐹𝑛 , and as diam(𝐹𝑛 ) ⩽ 2𝑛−1 1
, we have that 𝐹𝑛 ⊆ 𝐵(𝑦, 2𝑛−1 ) ⊆ 𝐵(𝑦, 𝜀) ⊆ 𝑈 𝛼0 . But
then, we have that 𝐹𝑛 covered by a single open set 𝑈 𝛼0 , a contradiction to our inductive construction of 𝐹𝑛 .
We conclude 𝑋 compact.

(2. =⇒ 3.) Suppose 𝑋 compact. Let {𝑥 𝑛 } 𝑛∈N ∈ 𝑋. Let 𝐹𝑛 = we have too that
Ð
𝑘 ⩾ 𝑛 {𝑥 𝑘 }-closed;
𝐹1 ⊇ 𝐹2 ⊇ · · · ⊇ 𝐹𝑛 ⊇ · · · .

↩→ Definition 3.5: Finite Intersection Property

ℱ has finite intersection property provided any finite subcollection of sets in ℱ has a non-empty
intersection.

↩→ Lemma 3.2 (Finite Interesection Formulation of Compactness). 𝑋-compact ⇐⇒ every collection ℱ

of closed subsets of 𝑋 with finite intersection property has non-empty intersection.

Proof
Proof.
Proof ■

This lemma directly gives that ∞ 𝑛=1 𝐹𝑛 ≠ ∅, {𝐹𝑛 } 𝑛∈N being a collection of closed subsets with any subset
Ñ
having nonempty intersection (by the nestedness). Let 𝑦 ∈ ∩∞ 𝐹 . Take 𝐵(𝑦, 1𝑘 ), which thus has nonempty
𝑛=1 𝑛
intersection with {𝑥 𝑘 } 𝑘 ⩾ 𝑛 ∀ 𝑛, ie ∃𝑛1 : 𝑑(𝑦, 𝑥 𝑛1 ) < 1 and ∃𝑛2 > 𝑛1 : 𝑑(𝑦, 𝑥 𝑛2 ) < 12 . Arguing inductively,
∃𝑛 𝑗 > 𝑛 𝑗−1 : 𝑑(𝑦, 𝑥 𝑛 𝑗 ) < 1𝑗 for any given 𝑛 𝑗−1 . It follows that lim 𝑗→∞ 𝑥 𝑛 𝑗 = 𝑦, and thus {𝑥 𝑛 𝑗 } is a convergent
subsequence of {𝑥 𝑛 } that converges within 𝑋, and thus 𝑋 is sequentially compact.
(3. =⇒ 1.) Suppose 𝑋 sequentially compact. Let {𝑥 𝑛 } ∈ 𝑋 be a Cauchy sequence in 𝑋, which thus have a
convergent subsequence {𝑥 𝑛 𝑘 } → 𝑦.

↩→ Lemma 3.3. Let {𝑥 𝑛 } be a Cauchy sequence in 𝑋 where 𝑋 sequentially compact. Then, if {𝑥 𝑛 𝑘 } → 𝑦, so

does {𝑥 𝑛 } → 𝑦

Proof
Proof.
Proof ■

Then, {𝑥 𝑛 } 𝑛 → 𝑦 and so 𝑋 complete.

Suppose 𝑋 not totally bounded, ie ∃𝜀 > 0 : 𝑋 cannot be covered by a finite union of balls of 𝐵(𝑥 𝑗 , 𝜀).
Let 𝑥 1 ∈ 𝑋 s.t. 𝐵(𝑥1 , 𝜀) ⊉ 𝑋; ∃𝑥2 ∈ 𝑋 \ 𝐵(𝑥1 , 𝜀), and so 𝑋 ⊈ 𝐵(𝑥1 , 𝜀) ∪ 𝐵(𝑥2 , 𝜀) Ð by assumption. Then,
choose 𝑥 3 ∈ 𝑋 \ (𝐵(𝑥 1 , 𝜀) ∪ 𝐵(𝑥 2 , 𝜀)). Arguing inductively, we have that ∃𝑥 𝑛 ∈ 𝑋 \ ( 𝑛𝑖=1 𝐵(𝑥 𝑖 , 𝜀)), noting that
𝑑(𝑥 𝑛 , 𝑥 𝑗 ) ⩾ 𝜀 ∀ 1 ⩽ 𝑗 ⩽ 𝑛.
Consider the sequence {𝑥 𝑗 } 𝑗∈N :

↩→ Lemma 3.4. {𝑥 𝑗 } cannot have a convergent subsequence.

Proof Follows by 𝑑(𝑥 𝑚 , 𝑥 𝑛 ) ⩾ 𝜀 ∀ 𝑚, 𝑛.

Proof
Proof. ■

This contradicts our assumption that 𝑋 sequentially compact, and we conclude 𝑋 must be totally bounded. ■

↩→ Lecture 10; Last Updated: Tue Feb 6 09:50:59 EST 2024

3.5 𝐿 𝑝 Spaces: Equivalent Notions of Compactness in Metric Spaces 33

 ⊛ Example 3.3: Complete Metric Space Example: 𝐿 𝑝 norm
Let 𝑓 ∈ 𝐶([𝑎, 𝑏]). We define the norm

∫ 𝑏  𝑝1
|| 𝑓 || 𝑝 := | 𝑓 (𝑥)| 𝑝 d𝑥 .
𝑎

As desired, || 𝑓 || 𝑝 ⩾ 0; || 𝑓 || 𝑝 = 0 ⇐⇒ 𝑓 ≡ 0; ||𝑐 · 𝑓 || 𝑝 = 𝑐 · || 𝑓 || 𝑝 .
1 1
Hölder’s and Minkowski’s inequalities for functions also hold; for 𝑝 + 𝑞 = 1, 1 ⩽ 𝑝, 𝑞 ⩽ ∞,
∫
| 𝑓 𝑔| ⩽ || 𝑓 || 𝑝 · || 𝑔|| 𝑞 ; || 𝑓 + 𝑔|| 𝑝 ⩽ || 𝑓 || 𝑝 + || 𝑔|| 𝑞 ,

respectively.

We similarly have the 𝐿∞ norm, namely, for a function 𝑓 : [𝑎, 𝑏] → R,

|| 𝑓 || ∞ = sup | 𝑓 (𝑥)| ,
𝑥∈[𝑎,𝑏]

which obeys all the necessary properties as well.

Let 𝑓𝑛 → 𝑓 in 𝐶([𝑎, 𝑏]), wrt || · · · || ∞ , where { 𝑓𝑛 } 𝑛∈N a sequence of functions. Namely, we say that

∀ 𝜀 > 0, ∃𝑁 ∈ N : ∀ 𝑛 ⩾ 𝑁 , sup | 𝑓𝑛 (𝑥) − 𝑓 (𝑥)| < 𝜀.

𝑥∈[𝑎,𝑏]

If this holds, we say that 𝑓𝑛 uniformly converges.

We say that 𝑓𝑛 (𝑥) → 𝑓 (𝑥) pointwise on [𝑎, 𝑏] if ∀ 𝑥 ∈ [𝑎, 𝑏], 𝑓𝑛 (𝑥) → 𝑓 (𝑥). Note that uniform
convergence implies pointwise convergence, but not the converse.

↩→ Theorem 3.5
Suppose 𝑓𝑛 (𝑥) continuous, and 𝑓𝑛 (𝑥) → 𝑓 (𝑥) uniformly on [𝑎, 𝑏]. Then, 𝑓 (𝑥) also continuous on [𝑎, 𝑏].

Proof Fix 𝜀 > 0, 𝑥0 ∈ [𝑎, 𝑏]. We have that ∃𝑁 : 𝑛 ⩾ 𝑁 , | 𝑓𝑛 (𝑥) − 𝑓 (𝑥)| < 3𝜀 , ∀ 𝑥 ∈ [𝑎, 𝑏].
Proof
Proof.
Let 𝑛 ⩾ 𝑁. 𝑓𝑛 (𝑥) continuous at 𝑥0 , hence ∃𝛿(𝑥0 ) > 0 : |𝑦 − 𝑥 0 | =⇒ | 𝑓𝑛 (𝑦) − 𝑓𝑛 (𝑥0 )| < 3𝜀 . We have

| 𝑓 (𝑥0 ) − 𝑓 (𝑦)| ⩽ | 𝑓 (𝑥 0 ) − 𝑓𝑛 (𝑥 0 )| + | 𝑓𝑛 (𝑥 0 ) − 𝑓𝑛 (𝑦)| + | 𝑓𝑛 (𝑦) − 𝑓 (𝑦)|

𝜀 𝜀 𝜀
⩽ + + = 𝜀,
3 3 3

completing the proof. ■

Remark 3.8. This does not hold with pointwise convergence.

3.5 𝐿 𝑝 Spaces: Equivalent Notions of Compactness in Metric Spaces 34

Remark 3.9. We will prove later that 𝐶([𝑎, 𝑏]) is complete for || 𝑓 || ∞ , but not for arbitrary || 𝑓 || 𝑝 , 1 ⩽ 𝑝 < +∞. To
“complete” 𝐶([𝑎, 𝑏]) for 𝑝 ≠ ∞, we will need to consider measurable functions and redefine our notion of integration.

↩→ Lecture 11; Last Updated: Thu Feb 8 09:51:13 EST 2024

4 Derivatives

4.1 Introduction

↩→ Definition 4.1: Differentiable

𝑓 (𝑥)− 𝑓 (𝑐)
We say 𝑓 (𝑥) differentiable at 𝑐 if ∃ lim𝑥→𝑐 𝑥−𝑐 . If so, we denote the limit 𝑓 ′(𝑐).

Remark 4.1. For 𝑥 close to 𝑐, then 𝑓 (𝑥) ≈ 𝑓 (𝑐) + 𝑓 ′(𝑐)(𝑥 − 𝑐); this is a linear approximation of 𝑓 at 𝑐.

⊛ Example 4.1: Weierstrass

Í∞ cos(3𝑛 )𝑥
𝑓 (𝑥) = 𝑛=1 2𝑛 is continuous in R, but nowhere differentiable.

↩→ Definition 4.2
The derivative, d𝑥, is a linear map 𝐶 1 ([𝑎, 𝑏]) → 𝐶 0 ([𝑎, 𝑏]).

4.2 Chain Rule

Remark 4.2. See Analysis I notes as well.

↩→ Theorem 4.1: Caratheodory’s Theorem

Let 𝑓 : 𝐼 → R, 𝑐 ∈ 𝐼. 𝑓 is differentiable at 𝑥 = 𝑐 iff ∃𝜑(𝑥) : 𝐼 → R s.t. 𝜑 continuous at 𝑐 and
𝑓 (𝑥) − 𝑓 (𝑐) = 𝜑(𝑥)(𝑥 − 𝑐).5

Proof If 𝑓 ′(𝑐) exists, let

Proof
Proof.
 𝑓 (𝑥)− 𝑓 (𝑐)
𝑥≠𝑐


𝑥−𝑐

𝜑(𝑥) = ,
 𝑓 ′(𝑐)
 𝑥 = 𝑐.

𝑓 (𝑥)− 𝑓 (𝑐)
which is well defined. Moreover, for 𝑥 ≠ 𝑐, 𝜑(𝑥)(𝑥 − 𝑐) = 𝑥−𝑐 (𝑥 − 𝑐) = 𝑓 (𝑥) − 𝑓 (𝑐) as desired; the case for
𝑥 = 𝑐 is clear. Continuity at 𝑐:

𝑓 (𝑥) − 𝑓 (𝑐)
lim 𝜑(𝑥) = lim = 𝑓 ′(𝑐) = 𝜑(𝑐).
𝑥→𝑐 𝑥→𝑐 𝑥−𝑐
5If not stated otherwise, sets named 𝐼 or 𝐽 are intervals.

4.2 Derivatives: Chain Rule 35

 Conversely, suppose such a 𝜑 exists. Then, by continuity,

𝑓 (𝑥) − 𝑓 (𝑐)
∃𝜑(𝑐) = lim 𝜑(𝑥) = lim
𝑥→𝑐 𝑥→𝑐 𝑥−𝑐

which gives directly that 𝑓 differentiable at 𝑐. ■

↩→ Theorem 4.2: Chain Rule

Let 𝑓 : 𝐽 → R, 𝑔 : 𝐼 → R, 𝑓 (𝐽) ⊆ 𝐼. If 𝑓 (𝑥) differentiable at 𝑐 and 𝑔(𝑦) is differentiable at 𝑦 = 𝑓 (𝑐), then
𝑔 ◦ 𝑓 (𝑥) is also differentiable at 𝑐, and

(𝑔 ◦ 𝑓 )′(𝑐) = 𝑔 ′( 𝑓 (𝑐)) 𝑓 ′(𝑐).

Proof Using Caratheodory’s Theorem, ∃𝜑 : 𝑓 (𝑥) − 𝑓 (𝑐) = 𝜑(𝑥)(𝑥 − 𝑐) with 𝜑(𝑐) = 𝑓 ′(𝑐). Let 𝑑 = 𝑓 (𝑐), then
Proof
Proof.
similarly ∃𝜓 : 𝑔(𝑦) − 𝑔(𝑑) = 𝜓(𝑦)(𝑦 − 𝑑) with 𝜓(𝑑) = 𝑔 ′(𝑑), with 𝜑, 𝜓 continuous at 𝑐, 𝑑 resp. Then,

𝑔( 𝑓 (𝑥)) − 𝑔( 𝑓 (𝑐)) = 𝜓( 𝑓 (𝑥))( 𝑓 (𝑥) − 𝑓 (𝑐)) = (𝜓 ◦ 𝑓 )(𝑥) · (𝜙(𝑥)(𝑥 − 𝑐))

𝜓 ◦ 𝑓 is continuous at 𝑐, as a composition of continuous functions (𝜓, 𝜙 continuous by construction, 𝑓

differentiable and thus continuous). It follows, then, that

𝑔( 𝑓 (𝑥)) − 𝑔( 𝑓 (𝑐))
lim = lim (𝜓 ◦ 𝑓 )(𝑥) · 𝜑(𝑥) = 𝜓( 𝑓 (𝑐))𝜑(𝑐) = 𝑔 ′( 𝑓 (𝑐)) · 𝑓 ′(𝑐),
𝑥→𝑐 𝑥−𝑐 𝑥→𝑐

by construction. ■

4.3 Critical Points

↩→ Definition 4.3
𝑓 : 𝐼 → R has a max/min 𝑐 if ∃𝐽 ⊆ 𝐼 : 𝑥 ∈ 𝐽 s.t. max𝑥∈𝐽 𝑓 (𝑥)/min𝑥∈𝐽 𝑓 (𝑥) = 𝑓 (𝑐).

↩→ Theorem 4.3: Rolle’s

Let 𝑓 : [𝑎, 𝑏] → R continuous. Suppose 𝑓 ′(𝑥) exists for all 𝑥 ∈ (𝑎, 𝑏) and 𝑓 (𝑎) = 𝑓 (𝑏) = 0. Then,
∃𝑐 ∈ (𝑎, 𝑏) : 𝑓 ′(𝑐) = 0.

Remark 4.3. A “complex-version” of Rolle’s:

↩→ Theorem 4.4: Gauss-Lucas

Let 𝑃(𝑧) be a complex-valued polynomial. Then, the roots of 𝑃 ′(𝑧) lie inside the convex hull of roots of
𝑃(𝑧), where a convex hull is the smallest polygon with vertices at the roots of 𝑃(𝑧).

4.3 Derivatives: Critical Points 36

 ↩→ Definition 4.4
Consider 𝑃(𝑧) = 𝑧 𝑛 − 1 for some 𝑛 ∈ N. If 𝑧 a root, we can show that (|𝑧|)𝑛 = 1, hence all roots lie on
the unit circle in the complex plane at multiples of the same angle. This gives us a regular 𝑛-gon in the
complex plane. We then have that 𝑃 ′(𝑧) = 𝑛𝑧 𝑛−1 , with has root 𝑧 = 0, which clearly lies within the 𝑛-gon
hull.

↩→ Theorem 4.5: Mean Value

Let 𝑓 be continuous on [𝑎, 𝑏] and differentiable on (𝑎, 𝑏). Then, ∃𝑐 ∈ (𝑎, 𝑏) s.t. 𝑓 (𝑏) − 𝑓 (𝑎) = 𝑓 ′(𝑐)(𝑏 − 𝑎).

𝑓 (𝑏)− 𝑓 (𝑎)
Proof Let 𝜑(𝑥) = 𝑓 (𝑥) − 𝑓 (𝑎) =
Proof
Proof. (𝑏−𝑎)
(𝑥 − 𝑎), where 𝜑(𝑎) = 0 = 𝜑(𝑏). By Rolle’s theorem, ∃𝑐 ∈ (𝑎, 𝑏) :
𝑓 (𝑏)− 𝑓 (𝑎)
𝜑′(𝑐) = 0 = 𝑓 ′(𝑐) − (𝑏−𝑎)
, as desired. ■

↩→ Lecture 12; Last Updated: Tue Apr 9 22:31:27 EDT 2024

4.4 Aside: Continued Fractions

We have that, for any 𝑥 ∈ R, 𝑥 = ⌊𝑥⌋ + {𝑥}, with {𝑥} ∈ (0, 1); ⌊𝑥⌋ and {𝑥} are the integral and fractional parts
of 𝑥 respectively.
Fix 𝑥 ∈ R, assuming 𝑥 ≠ 0.
1
Let 𝑥1 := {𝑥}
. We can write
1
𝑥 = ⌊𝑥⌋ + .
𝑥1
1
If {𝑥1 } ≠ 0, let 𝑥 2 := {𝑥 1 }
and write
1
𝑥 = ⌊𝑥⌋ + 1
.
⌊𝑥1 ⌋ + 𝑥2

Continuing in this manner, this process stops if {𝑥 𝑖 } = 0 for

√
some 𝑖; if 𝑥 ∈ Q, this process will stop, else, it
5±1
will continue infinitely. For instance, the Golden Ratio 𝑥 = 2 has continued fraction expansion

1
𝑥= 1
.
1+ 1
1+ 1+···

More succinctly, we can denote 𝑎 0 := ⌊𝑥⌋ and 𝑎 𝑖 = ⌊𝑥 𝑖 ⌋, 𝑖 ⩾ 1, and write

1
𝑥 = 𝑎0 + 1
.
𝑎1 + 1
𝑎2 +
..
𝑎3 + .

We notate, accordingly, 𝑥 := (𝑎1 , 𝑎2 , 𝑎3 , . . . ); in this case, the Golden Ratio can be notated (1, 1, 1, . . . ).
𝑝
We denote 𝑞 𝑛𝑛 as the 𝑛th continued fraction of a given 𝑥. It turns out that this is the best possible rational
approximation for 𝑥 ∉ Q.

4.4 Derivatives: Aside: Continued Fractions 37

4.5 Back To Derivatives

↩→ Theorem 4.6
𝑓 : 𝐼 → R, differentiable. 𝑓 is increasing (resp decreasing) iff 𝑓 ′(𝑥) ⩾ 0 ∀ 𝑥 ∈ 𝐼 (resp 𝑓 ′(𝑥) ⩽ 0 ∀ 𝑥 ∈ 𝐼).

↩→Proposition 4.1
Let 𝑓 continuous on 𝐼 = [𝑎, 𝑏]. Let 𝑎 < 𝑐 < 𝑏 and suppose 𝑓 differentiable on (𝑎, 𝑐) and (𝑐, 𝑏). Suppose
𝑓 ′(𝑥) ⩾ 0 on (𝑐 − 𝛿, 𝑐) and 𝑓 ′(𝑥) ⩽ 0 on (𝑐, 𝑐 + 𝛿) for some 𝛿 > 0. Then, 𝑓 has local max at 𝑥 = 𝑐.

↩→ Lemma 4.1
Let 𝐼 ⊆ R, and assume 𝑓 : 𝐼 → R is differentiable at 𝑥 = 𝑐 ∈ 𝐼.

1. If 𝑓 ′(𝑐) > 0, then ∃𝛿 > 0 : 𝑓 (𝑥) > 𝑓 (𝑐) ∀ 𝑥 ∈ 𝐼, 𝑥 ∈ (𝑐, 𝑐 + 𝛿).

2. (Reverse statement for 𝑓 ′(𝑐) < 0)

↩→ Theorem 4.7: Darboux

Suppose 𝑓 differentiable on 𝐼 := [𝑎, 𝑏] and 𝑓 ′(𝑎) < 𝑘 < 𝑓 ′(𝑏). Then, ∃𝑐 ∈ (𝑎, 𝑏) such that 𝑓 ′(𝑐) = 𝑘.

↩→ Lecture 13; Last Updated: Thu Feb 15 09:49:55 EST 2024

4.6 L’Hopital’s Rules

↩→Proposition 4.2
Suppose 𝑓 (𝑥), 𝑔(𝑥) : [𝑎, 𝑏] → R with 𝑓 (𝑎) = 𝑔(𝑎) = 0, and 𝑔(𝑥) ≠ 0 ∀ 𝑎 < 𝑥 < 𝑏. Suppose 𝑓 , 𝑔 are
𝑓 (𝑥) 𝑓 ′ (𝑎)
differentiable at 𝑥 = 𝑎 and 𝑔 ′(𝑎) ≠ 0. Then, lim𝑥→𝑎 + 𝑔(𝑥)
exists, and moreover, it is equal to 𝑔 ′ (𝑎)
.

Proof
Proof.
Proof

𝑓 (𝑥) − 𝑓 (𝑎) 𝑔(𝑥) − 𝑔(𝑎) 𝑓 (𝑥) 𝑥 − 𝑎 𝑓 (𝑥)

lim+ ( )/( ) = lim+ = lim+ ,
𝑥→𝑎 𝑥−𝑎 𝑥−𝑎 𝑥→𝑎 𝑥 − 𝑎 𝑔(𝑥) 𝑥→𝑎 𝑔(𝑥)

𝑓 ′ (𝑎)
but the original line is simply 𝑔 ′ (𝑎)
. ■

⊛ Example 4.2
sin 𝑥 cos(0)
lim𝑥→0 𝑥 = 1 = 1.

4.6 Derivatives: L’Hopital’s Rules 38

 ↩→ Theorem 4.8: Cauchy Mean Value
Let 𝑓 (𝑥), 𝑔(𝑥) : [𝑎, 𝑏] → R where 𝑓 , 𝑔 continuous on [𝑎, 𝑏] and differentiable on (𝑎, 𝑏). Assuming
𝑔 ′(𝑥) ≠ 0, ∀ 𝑥 ∈ (𝑎, 𝑏), then ∃𝑐 ∈ (𝑎, 𝑏) such that

𝑓 (𝑏) − 𝑓 (𝑎) 𝑓 ′(𝑐)

= ′ .
𝑔(𝑏) − 𝑔(𝑎) 𝑔 (𝑐)

↩→Proposition 4.3: More General L’Hopital

let −∞ ⩽ 𝑎 < 𝑏 ⩽ +∞ and 𝑓 , 𝑔 differentiable on (𝑎, 𝑏). Suppose lim𝑥→𝑎 + 𝑓 (𝑥) = 0 = lim𝑥→𝑎 + 𝑔(𝑥).

𝑓 ′ (𝑥) 𝑓 (𝑥)
1. If ∃𝐿 := lim𝑥→𝑎 + 𝑔 ′ (𝑥)
where 𝐿 some real number, then lim𝑥→𝑎 + 𝑔(𝑥)
= 𝐿 as well.

𝑓 ′ (𝑥) 𝑓 (𝑥)
2. If ∃𝐿 := lim𝑥→𝑎 + 𝑔 ′ (𝑥)
where 𝐿 = +∞ or −∞, then lim𝑥→𝑎 + 𝑔(𝑥)
= 𝐿 as well.

↩→Proposition 4.4
Let −∞ ⩽ 𝑎 < 𝑏 ⩽ +∞, 𝑓 , 𝑔 differentiable on (𝑎, 𝑏) and 𝑔 ′(𝑥) ≠ 0 ∀ 𝑥 ∈ (𝑎, 𝑏). Suppose lim𝑥→𝑎 + 𝑔(𝑥) = ±∞.

𝑓 ′ (𝑥) 𝑓 (𝑥)
1. If lim𝑥→𝑎 + 𝑔 ′ (𝑥)
=: 𝐿 exists and is some finite real number, then lim𝑥→𝑎 + 𝑔(𝑥)
= 𝐿 as well.

𝑓 ′ (𝑥) 𝑓 (𝑥)
2. If lim𝑥→𝑎 + 𝑔 ′ (𝑥)
=: 𝐿 exists and is ±∞, then lim𝑥→𝑎 + 𝑔(𝑥)
= 𝐿 as well.

4.7 Taylor’s Theorem

↩→ Theorem 4.9: Taylor’s Theorem

Let 𝐼 = [𝑎, 𝑏] ⊆ R, 𝑓 : 𝐼 → R, 𝑓 ∈ 𝐶 𝑛 (𝐼) and suppose 𝑓 (𝑛+1) (𝑥) exists on (𝑎, 𝑏). Let 𝑥0 ∈ [𝑎, 𝑏]. Then, for
any 𝑥 ∈ [𝑎, 𝑏], ∃𝑐 between 𝑥, 𝑥 0 such that

𝑓 ′′(𝑥0 ) 2 𝑓 (𝑛) (𝑥0 ) 𝑛 𝑓 (𝑛+1) (𝑐)

′
𝑓 (𝑥) = 𝑓 (𝑥0 ) + 𝑓 (𝑥0 )(𝑥 − 𝑥0 ) + (𝑥 − 𝑥 0 ) + · · · + (𝑥 − 𝑥 0 ) + (𝑥 − 𝑥0 )𝑛+1
2 𝑛! (𝑛 + 1)!

↩→ Lecture 14; Last Updated: Wed Apr 10 08:27:48 EDT 2024

4.7 Derivatives: Taylor’s Theorem 39

 ↩→ Theorem 4.10: Relative Extrema
6 Let 𝐼 ⊆ R be an open interval, 𝑥0 ∈ 𝐼, and 𝑛 ⩾ 2. Suppose 𝑓 ′ , 𝑓 ′′ , . . . , 𝑓 (𝑛) are continuous in a
neighborhood of 𝑥0 , and 𝑓 ′(𝑥 0 ) = 𝑓 ′′(𝑥0 ) = · · · = 𝑓 (𝑛−1) (𝑥 0 ) = 0 and 𝑓 (𝑛) (𝑥0 ) ≠ 0. Then:

1. if 𝑛 is even and 𝑓 (𝑛) (𝑥0 ) > 0, then 𝑓 has a local minimum at 𝑥0 ;

2. if 𝑛 is even and 𝑓 (𝑛) (𝑥0 ) < 0, then 𝑓 has a local maximum at 𝑥 0 ;

3. if 𝑛 is odd, then 𝑓 has neither a local minimum nor maximum at 𝑥0 .

Proof If 𝑛 := 2𝑚-even and 𝑓 (2𝑚) (𝑥0 ) > 0, then 𝑓 (𝑛) (𝑐) > 0 so 𝑓 (𝑥) − 𝑓 (𝑥 0 ) = 𝑓 (2𝑚) (𝑐)(𝑥 − 𝑥0 ) > 0.
Proof
Proof. ■

4.8 Convex Sets

↩→ Definition 4.5: Convex Set

𝐴 ⊆ 𝑉-vector space over R is convex if for any 𝑥, 𝑦 ∈ 𝐴 and any 0 ⩽ 𝑡 ⩽ 1, 𝑡 · 𝑥 + (1 − 𝑡) · 𝑦 ∈ 𝐴.

↩→ Definition 4.6: Convex Function

Let 𝑓 : 𝐼 → R. 𝑓 is convex if ∀ 𝑥1 , 𝑥2 ∈ 𝐼 and 0 ⩽ 𝑡 ⩽ 1,

𝑓 ((1 − 𝑡)𝑥1 + 𝑡𝑥2 ) ⩽ (1 − 𝑡) 𝑓 (𝑥 1 ) + 𝑡 𝑓 (𝑥2 ).

↩→ Lecture 15; Last Updated: Thu Feb 22 21:53:23 EST 2024

5 Riemann Integral

5.1 Introduction

↩→ Definition 5.1: Partitions

A partition is a division of an interval (𝑎, 𝑏), denoted

𝒫 = {𝑎 = 𝑥0 , 𝑥1 , . . . , 𝑥 𝑛−1 , 𝑥 𝑛 = 𝑏}.

We define diam(𝒫) := max𝑛 |𝑥 𝑖 − 𝑥 𝑖−1 |.

¤ is one in which, for each interval we choose some 𝑡 𝑖 ∈ (𝑥 𝑖 , 𝑥 𝑖+1 ].
A marked partition, denoted 𝒫,

6Bartle-Sherbert, Theorem 6.4.4

5.1 Riemann Integral: Introduction 40

 ↩→ Definition 5.2: Riemann Sum
We denote
𝑛
Õ
¤ =
𝑆( 𝑓 , 𝒫) 𝑓 (𝑡 𝑖 )(𝑥 𝑖 − 𝑥 𝑖−1 ).
𝑖=1

↩→ Definition 5.3: Riemann Integrable

¤ → 𝐿 as diam(𝒫)
A function 𝑓 is Riemann Integrable on [𝑎, 𝑏] if 𝑆( 𝑓 , 𝒫) ¤ → 0 for any choice of 𝑡 𝑖 ∈ [𝑥 𝑖 , 𝑥 𝑖+1 ].

¤ < 𝜀.
That is, ∀ 𝜀 > 0, ∃𝛿 : if diam(𝒫) < 𝛿, then for any choice of 𝑡 𝑖 ∈ [𝑥 𝑖 , 𝑥 𝑖+1 ] we have 𝐿 − 𝑆( 𝑓 , 𝒫)

↩→Proposition 5.1
1. If 𝐿 exists, it is unique.
∫𝑏 ∫𝑏 ∫𝑏 ∫𝑏
2. The integral is linear in 𝑓 (𝑥); if 𝑎
𝑓 (𝑥) d𝑥 and 𝑎
𝑔(𝑥) d𝑥 exist, then 𝑎
(𝑐1 𝑓 + 𝑐 2 𝑔) d𝑥 = 𝑐1 𝑎
𝑓 d𝑥 +
∫𝑏
𝑐2 𝑎
𝑔 d𝑥.
∫𝑏 ∫𝑏
3. If 𝑓 ⩽ 𝑔 are Riemann integrable on [𝑎, 𝑏], then 𝑎
𝑓 (𝑥) d𝑥 ⩽ 𝑎
𝑔(𝑥) d𝑥.

↩→Proposition 5.2
If 𝑓 (𝑥) integrable on [𝑎, 𝑏], the 𝑓 (𝑥) is bounded on [𝑎, 𝑏].

∫𝑏
¤ < 𝛿 then 𝐿 − 𝑆( 𝑓 , 𝒫)
Proof Suppose 𝑎 𝑓 exists. Let 𝜀 > 0, and 𝛿 such that if diam(𝒫)
Proof.
Proof ¤ . Let 𝜀 = 1. Then,
¤ ⩽ |𝐿| + 1.
𝑆( 𝑓 , 𝒫)
Let 𝑄 = {𝑎 = 𝑥0 , . . . , 𝑥 𝑛 = 𝑏} be a partition of [𝑎, 𝑏] such that diam(𝑄) < 𝛿. Suppose towards a
contradiction that 𝑓 is not bounded on [𝑎, 𝑏]. Then, 𝑓 is unbounded on at least one interval [𝑥 𝑖 , 𝑥 𝑖+1 ],
say,
Í on [𝑥 𝑘 , 𝑥 𝑘+1 ]. Let 𝑡 𝑖 = 𝑥 𝑖 for 𝑖 ≠ 𝑘 and choose 𝑡 𝑘 ∈ [𝑥 𝑘 , 𝑥 𝑘+1 ] such that | 𝑓 (𝑡 𝑘 )| (𝑥 𝑘+1 − 𝑥 𝑘 ) > |𝐿| + 1 +
𝑖≠𝑘 𝑓 (𝑡 𝑖 )(𝑥 𝑖+1 − 𝑥 𝑖 ) (which we can do by assumption of 𝑓 being unbounded).

¤ ⩽ |𝐿| + 1, but we have that

By assumption, 𝑆( 𝑓 , 𝑄)
Õ
¤ =
𝑆( 𝑓 , 𝑄) 𝑓 (𝑡 𝑖 )(𝑥 𝑖+1 − 𝑥 𝑖 ) + | 𝑓 (𝑡 𝑘 )| (𝑥 𝑘+1 − 𝑥 𝑘 ) > 2𝑁 + |𝐿| + 1,
𝑖≠𝑘
| {z }
:=𝑁

contradiction. ■

5.2 Cauchy Criterion

5.2 Riemann Integral: Cauchy Criterion 41

 ↩→Proposition 5.3: Cauchy Criterion for Integrability
𝑓 ∈ ℛ[𝑎, 𝑏] ⇐⇒ ∀ 𝜀 > 0, ∃𝛿 > 0 : if 𝑃¤ and 𝑄¤ are tagged partitions of [𝑎, 𝑏] s.t. diam 𝑃¤ < 𝛿 and
¤ < 𝜀.7
¤ − 𝑆( 𝑓 , 𝑄)
diam 𝑄¤ < 𝛿, then 𝑆( 𝑓 , 𝑃)

5.3 Squeeze Theorem

↩→ Theorem 5.1
∫𝑏
Let 𝑓 : [𝑎, 𝑏] → R. Then 𝑎
𝑓 exists ⇐⇒ ∀ 𝜀 > 0, ∃𝛼 𝜀 (𝑥), 𝜔 𝜀 (𝑥) ∈ ℛ[𝑎, 𝑏], 𝛼 𝜀 ⩽ 𝑓 ⩽ 𝜔 𝜀 , and

∫ 𝑏
(𝜔 𝜀 − 𝛼 𝜀 ) < 𝜀
𝑎

Proof If 𝑓 ∈ ℛ[𝑎, 𝑏] then take 𝛼 𝜀 = 𝑓 = 𝜔 𝜀 .

Proof
Proof.
Conversely, let 𝜀 > 0. Since 𝛼 𝜀 , 𝜔 𝜀 ∈ ℛ[𝑎, 𝑏], then, ∃𝛿 > 0 such that for any tagged partition with
¤ − 𝑏 𝛼 𝜀 < 𝜀 and 𝑆(𝜔 𝜀 , 𝑃)
dim 𝑃¤ < 𝛿, then 𝑆(𝛼 𝜀 , 𝑃) ¤ − 𝑏 𝜔 𝜀 < 𝜀, thus
∫ ∫
𝑎 𝑎

∫ 𝑏 ∫ 𝑏
¤ ⩽ 𝑆( 𝑓 , 𝑃)
𝛼 𝜀 − 𝜀 < 𝑆(𝛼 𝜀 , 𝑃) ¤ ⩽ 𝑆(𝜔 𝜀 , 𝑃)
¤ < 𝜔 𝜀 + 𝜀.
𝑎 𝑎

∫𝑏
Let 𝑄¤ be any other tagged partition with diam 𝑄¤ < 𝛿; then, the same inequality holds ie 𝑎
¤ <
𝛼 𝜀 − 𝜀 < 𝑆( 𝑓 , 𝑄)
∫𝑏
𝑎
𝜔 𝜀 + 𝜀. Subtracting one from the other, we see that

∫ 𝑏 ∫ 𝑏
¤ − 𝑆( 𝑓 , 𝑄)
𝑆( 𝑓 , 𝑃) ¤ < 𝜔𝜀 − 𝛼 𝜀 + 2𝜀 < 3𝜀,
𝑎 𝑎

and thus 𝑓 ∈ ℛ[𝑎, 𝑏] by Cauchy Criterion. ■

↩→ Lecture 16; Last Updated: Wed Apr 10 08:39:04 EDT 2024

↩→ Lemma 5.1: BS-7.2.4

Let 𝐽 ⊆ [𝑎, 𝑏] an interval with endpoints 𝑐 < 𝑑. If

1 𝑥 ∈ 𝐽



𝜑 𝐽 (𝑥) := .
0 𝑥 ∉ 𝐽


∫𝑏
Then, 𝜑 𝐽 ∈ ℛ[𝑎, 𝑏] and 𝑎
𝜑 𝐽 = 𝑑 − 𝑐.

7Note that ℛ[𝑎, 𝑏] is the set of all real-valued functions integrable on the interval [𝑎, 𝑏].

5.3 Riemann Integral: Squeeze Theorem 42

 ↩→ Theorem 5.2
Let 𝜑 : [𝑎, 𝑏] → R, 𝜑 ∈ ℛ[𝑎, 𝑏]; that is, step functions are integrable.

↩→ Theorem 5.3
𝑓 continuous on [𝑎, 𝑏] implies 𝑓 ∈ ℛ[𝑎, 𝑏].

Proof (Sketch) 𝑓 uniform continuous; use this to construct step functions that “bound” 𝑓 from above and
Proof
Proof.
below. Apply the squeeze theorem. ■

↩→ Lecture 17; Last Updated: Thu Mar 28 14:33:47 EDT 2024

↩→ Theorem 5.4: BS-7.2.7

Monotone functions on [𝑎, 𝑏] are integrable.

Proof We show only for increasing. Let 𝑓 : [𝑎, 𝑏] → R be monotone increasing. If 𝑓 constant, then it is a step
Proof
Proof.
function and we are done.
𝑓 (𝑏)− 𝑓 (𝑎) 𝜀
Otherwise, 𝑓 (𝑏) − 𝑓 (𝑎) > 0. Let 𝜀 > 0 and 𝑞 ∈ N such that ℎ ..= 𝑞 < 𝑏−𝑎 , effectively subdividing the
𝑦-axis into 𝑞 equal-sized parts. Then, let

𝑦 𝑖 ..= 𝑓 (𝑎) + 𝑖 ℎ, 0 ⩽ 𝑖 ⩽ 𝑞,

and take



 ∅



𝐴 𝑘 ..= 𝑓 −1 ([𝑦 𝑘+1 , 𝑦 𝑘 )) = {𝑥} .


 𝐼𝑖



We disregard each 𝐴 𝑘 : 𝐴 𝑘 = ∅, and adjoin the isolated points {𝑥} to the 𝐼 𝑖 ’s, and hence have a partition
∪ 𝑘 𝐴 𝑘 = [𝑎, 𝑏]. Letting 𝛼(𝑥) = 𝑦 𝑘−1 and 𝜔(𝑥) = 𝑦 𝑘 for 𝑥 ∈ 𝐴 𝑘 , then 𝛼(𝑥) ⩽ 𝑓 (𝑥) ⩽ 𝜔(𝑥) ∀ 𝑥 ∈ [𝑎, 𝑏] (effectively,
we are created a “series of squeezes”). Then,

∫ 𝑏 𝑞
Õ
𝜔(𝑥) − 𝛼(𝑥) d𝑥 = (𝑦 𝑘 − 𝑦 𝑘−1 )(𝑥 𝑘 − 𝑥 𝑘−1 ) = ℎ(𝑏 − 𝑎) < 𝜀,
𝑎 𝑘=1

and the proof is completed by applying the squeeze theorem. ■

↩→ Theorem 5.5: Additivity; BS-7.2.8

Let 𝑓 : [𝑎, 𝑏] → R and 𝑎 < 𝑐 < 𝑏. Then, 𝑓 ∈ ℛ[𝑎, 𝑏] ⇐⇒ 𝑓 ∈ ℛ[𝑎, 𝑐] and 𝑓 ∈ ℛ[𝑐, 𝑏]. Moreover,
∫𝑏 ∫ 𝑐 ∫𝑏
𝑎
𝑓 (𝑥) d𝑥 = 𝑎
𝑓 (𝑥) d𝑥 + 𝑐
𝑓 (𝑥) d𝑥.

Proof See book. Remark that this holds for finite summations of integrals as such by induction.
Proof
Proof. ■

5.3 Riemann Integral: Squeeze Theorem 43

5.4 Fundamental Theorem of Calculus

↩→ Definition 5.4
Call 𝐹(𝑥) a primitive of 𝑓 (𝑥) if 𝐹 differential and 𝐹′(𝑥) = 𝑓 (𝑥).

↩→ Theorem 5.6: Fundamental Theorem of Calculus

Let 𝐹, 𝑓 : [𝑎, 𝑏] → R and 𝐸 ⊆ [𝑎, 𝑏] a finite set s.t.

1. 𝐹 continuous on [𝑎, 𝑏]

2. 𝐹′(𝑥) = 𝑓 (𝑥) ∀ 𝑥 ∈ [𝑎, 𝑏] \ 𝐸; ie they agree for all but finitely many points

3. 𝑓 ∈ ℛ[𝑎, 𝑏]

∫𝑏
Then, 𝑎
𝑓 (𝑥) = 𝐹(𝑏) − 𝐹(𝑎).

Proof (Sketch) Remark first that it suffices to prove for 𝐸 ..= {𝑎, 𝑏}; using additivity, we can subdivide any
Proof
Proof.
other such 𝐸 into such subsets of 1 or 2 elements.
∫𝑏
Fix 𝜀 > 0 and take 𝛿 > 0 such that for any 𝑃¤ of [𝑎, 𝑏] s.t. diam 𝑃¤ < 𝛿, 𝑆( 𝑓 , 𝑃)
¤ −
𝑎
𝑓 (𝑥) < 𝜀. Applying the
¤
mean value theorem to 𝐹 on each [𝑥 𝑖−1 , 𝑥 𝑖 ] of 𝑃:

𝐹(𝑥 𝑖 ) − 𝐹(𝑥 𝑖−1 ) = 𝐹′(𝑢𝑖 )(𝑥 𝑖 − 𝑥 𝑖−1 ), 𝑢𝑖 ∈ [𝑥 𝑖−1 , 𝑥 𝑖 ]

= 𝑓 (𝑢𝑖 )(𝑥 𝑖 − 𝑥 𝑖−1 )

Hence, summing over each of these,

𝐹(𝑎) 𝐹(𝑏)
𝐹(𝑥 1 ) − 
𝐹(𝑥 ) + 𝐹(𝑥2 ) − 𝐹(𝑥1 ) + · · · + 
𝐹(𝑥
𝑛) − 𝐹(𝑥 𝑛−1 ) = 𝑓 (𝑢1 )(𝑥 1 − 𝑎) + · · · + 𝑓 (𝑢𝑛 )(𝑥 𝑛 − 𝑥 𝑛−1 )
*
 *

0
𝑛
Õ
=⇒ 𝐹(𝑏) − 𝐹(𝑎) = 𝑓 (𝑢𝑖 )(𝑥 𝑖 − 𝑥 𝑖+1 ) =: 𝑆( 𝑓 , 𝑃¤1 )
𝑖=1

∫𝑏
by construction, diam(𝑃¤1 ) < 𝛿 since the only change we have made from 𝑃¤ is the tags, hence 𝑆( 𝑓 , 𝑃¤1 ) − 𝑎
𝑓 (𝑥) <
𝜀. Thus,
∫ 𝑏 ∫ 𝑏
𝑆( 𝑓 , 𝑃¤1 ) − 𝑓 (𝑥) = 𝐹(𝑏) − 𝐹(𝑎) − 𝑓 (𝑥) < 𝜀
𝑎 𝑎
∫ 𝑏
=⇒ 𝐹(𝑏) − 𝐹(𝑎) = 𝑓 (𝑥) as 𝜀 → 0.
𝑎

■
↩→ Lecture 18; Last Updated: Thu Mar 28 09:07:47 EDT 2024

5.4 Riemann Integral: Fundamental Theorem of Calculus 44

5.5 Upper and Lower Riemann Sums

↩→ Definition 5.5: Upper/Lower Riemann Sums

For a partition 𝑃,

Í𝑛
• 𝑆( 𝑓 , 𝑃) ..= 𝑖=1 (sup𝑡∈[𝑥 𝑖−1 ,𝑥 𝑖 ] 𝑓 (𝑡)) · (𝑥 𝑖 − 𝑥 𝑖−1 )
Í𝑛
• SS( 𝑓 , 𝑃) ..= 𝑖=1 (inf𝑡∈[𝑥 𝑖−1 ,𝑥 𝑖 ] 𝑓 (𝑡)) · (𝑥 𝑖 − 𝑥 𝑖−1 )

↩→Proposition 5.4
¤
For any tagged partition 𝑃,
SS( 𝑓 , 𝑃) ⩽ 𝑆( 𝑓 , 𝑃)
¤ ⩽ 𝑆( 𝑓 , 𝑃).

Moreover, 𝑓 ∈ ℛ[𝑎, 𝑏] if ∀ 𝜀 > 0, ∃𝛿 > 0 s.t. diam(𝑃) < 𝛿 =⇒ 𝑆( 𝑓 , 𝑃) − SS( 𝑓 , 𝑃) < 𝜀.

Proof (Sketch) Remark that this is a similar idea to saying that inf = sup =⇒ limit exists.
Proof
Proof. ■

↩→Proposition 5.5
Let 𝑃1 , 𝑃2 be partitions of [𝑎, 𝑏], and let 𝑃3 be the common refinement of 𝑃1 , 𝑃2 . Then

SS( 𝑓 , 𝑃𝑖 ) ⩽ SS( 𝑓 , 𝑃3 ) ⩽ 𝑆( 𝑓 , 𝑃3 ) ⩽ 𝑆( 𝑓 , 𝑃𝑖 ), 𝑖 = 1, 2,

that is, the finer refinement always gives a better approximation.

5.6 Indefinite Integral

↩→ Definition 5.6
For 𝑓 ∈ ℛ[𝑎, 𝑏] and any 𝑧 ∈ [𝑎, 𝑏], define
∫ 𝑧
..
𝐹(𝑧) = 𝑓 (𝑥) d𝑥 .
𝑎

↩→ Theorem 5.7
𝐹(𝑧) continuous on [𝑎, 𝑏].

Proof 𝑓 ∈ ℛ[𝑎, 𝑏] =⇒ 𝑓 bounded =⇒ ∃𝑀 s.t. | 𝑓 (𝑥)| ⩽ 𝑀 ∀ 𝑥 ∈ [𝑎, 𝑏], so (assuming 𝑧 < 𝑤),
Proof
Proof.
∫ 𝑧 ∫ 𝑤 ∫ 𝑤
|𝐹(𝑧) − 𝐹(𝑤)| = 𝑓 (𝑥) d𝑥 − 𝑓 (𝑥) d𝑥 = 𝑓 (𝑥) d𝑥 ⩽ 𝑀 · |𝑧 − 𝑤| ,
𝑎 𝑎 𝑧

5.6 Riemann Integral: Indefinite Integral 45

so taking 𝑤 → 𝑧, |𝐹(𝑧) − 𝐹(𝑤)| → 0. ■

↩→ Theorem 5.8: Another Fundamental Theorem of Calculus

Let 𝑓 ∈ ℛ[𝑎, 𝑏], 𝑓 -continuous at 𝑐 ∈ [𝑎, 𝑏]. Then 𝐹(𝑧) differentiable at 𝑐 and 𝐹′(𝑐) = 𝑓 (𝑐).

↩→ Corollary 5.1
If 𝑓 (𝑥) continuous on [𝑎, 𝑏] 𝐹′(𝑥) = 𝑓 (𝑥) ∀ 𝑥 ∈ [𝑎, 𝑏].

↩→ Theorem 5.9: Substitution/Change of Variables

Let 𝐽 ..= [𝛼, 𝛽], 𝜑 : 𝐽 → R, 𝜑 ∈ 𝐶 1 ([𝑎, 𝑏]). Suppose 𝜑(𝐽) ⊆ 𝐼 ⊆ R, and let 𝑓 : 𝐼 → R be continuous on 𝐼.
Then,
∫ 𝜑(𝛽) ∫ 𝛽
𝑓 (𝑥) d𝑥 = 𝑓 (𝜑(𝑡)) · 𝜑′(𝑡) d𝑡 .
𝜑(𝛼) 𝛼

Proof Left as a (homework) exercise; make use of the chain rule!

Proof.
Proof ■

⊛ Example 5.1
∫ 4 sin(√𝑡 )
Compute 1
√ d𝑡 using the previous theorem.
𝑡

5.7 Lebesgue Integrability Criterion

↩→ Definition 5.7: Lebesgue Measure 0

𝐴 ⊆ R has Lebesgue measure 0 iff ∀ 𝜀 > 0, 𝐴 can be covered by a countable union of intervals 𝐽𝑘 ..= [𝑎 𝑘 , 𝑏 𝑘 ]
Í∞
such that 𝑘=1 |𝐽𝑘 | ⩽ 𝜀. We also call such an 𝐴 a null set.

For some set 𝑆 ⊆ R and statement 𝑃, we say “𝑃 holds for almost every 𝑥 ∈ 𝑆” if {𝑥 ∈ 𝑆 : 𝑃 false } has
Lebesgue measure 0.

⊛ Example 5.2
1. Any countable set is a null set.

2. The Cantor set is a null set.

5.7 Riemann Integral: Lebesgue Integrability Criterion 46

 ↩→ Theorem 5.10: Lebesgue Integrability Criterion
Let 𝑓 : [𝑎, 𝑏] → R be a bounded function. Then

𝑓 ∈ ℛ[𝑎, 𝑏] ⇐⇒ 𝑓 − continuous for almost every 𝑥 ∈ [𝑎, 𝑏]

⇐⇒ {𝑧 ∈ [𝑎, 𝑏] : 𝑓 discontinuous} has Lebesgue measure 0.

Remark 5.1. The proof is rather involved, but is in the appendix of Bartle. Its important to remark that this is a necessary
and sufficient condition.

⊛ Example 5.3

1 𝑥 ∈ Q



..
1. Let 𝑓 : [0, 1] → R, 𝑓 (𝑥) = . 𝑓 discontinuous everywhere, so 𝑓 ∉ ℛ[𝑎, 𝑏].
0 𝑥 ∉ Q



 𝑏1 𝑥= 𝑎
∈ Q s.t. (𝑎, 𝑏) = 1


𝑏

..
2. Let 𝑓 (𝑥) = . One can show that 𝑓 continuous on 𝑥 ∈ R \ Q and
𝑥∉Q 0

only discontinuous on Q. But this is a countable set so certainly has Lebesgue measure 0 and


so 𝑓 ∈ ℛ[0, 1].

↩→ Lecture 19; Last Updated: Tue Apr 9 14:45:17 EDT 2024

↩→ Theorem 5.11: Composition

𝑓 ∈ ℛ[𝑎, 𝑏], 𝑓 ([𝑎, 𝑏]) ⊆ [𝑐, 𝑑], 𝜑 : [𝑐, 𝑑] → R continuous, then 𝜑 ◦ 𝑓 ∈ ℛ[𝑎, 𝑏].

Proof
Proof
Proof.
{𝑥 s.t. 𝜑 ◦ 𝑓 discontinuous at 𝑥} ⊆ {𝑥 : 𝑓 discontinuous at 𝑥}
since 𝜑 continuous. The RHS has Lebesgue measure 0, and thus so does the LHS, hence the proof. ■

↩→ Theorem 5.12: Product Theorem

𝑓 , 𝑔 ∈ ℛ[𝑎, 𝑏] =⇒ 𝑓 · 𝑔 ∈ ℛ[𝑎, 𝑏].

Proof 𝑓 · 𝑔 = 41 ( 𝑓 + 𝑔)2 − ( 𝑓 − 𝑔)2 . 𝑓 ± 𝑔 ∈ ℛ[𝑎, 𝑏] and so so is ( 𝑓 ± 𝑔)2 by taking 𝜑(𝑥) ..= 𝑥 2 as in the
 
Proof
Proof.
previous theorem. It follows that 𝑓 · 𝑔 ∈ ℛ[𝑎, 𝑏]. ■

5.8 Integration by Parts

5.8 Riemann Integral: Integration by Parts 47

 ↩→ Theorem 5.13
Let 𝐹, 𝐺 be differentiable on [𝑎, 𝑏], with 𝑓 ..= 𝐹′ , 𝑔 ..= 𝐺′. Suppose 𝑓 , 𝑔 ∈ ℛ[𝑎, 𝑏], then
∫ 𝑏 ∫ 𝑏
𝑓 (𝑥)𝐺(𝑥) d𝑥 = 𝐹(𝑥)𝐺(𝑥)| 𝑏𝑎 − 𝐹(𝑥)𝑔(𝑥) d𝑥 .
𝑎 𝑎

Proof Remark that (𝐹𝐺)′ = 𝐹′ 𝐺 + 𝐹𝐺′ = 𝑓 𝐺 + 𝐹 𝑔, so on the one hand

Proof
Proof.
∫ 𝑏 ∫ 𝑏 ∫ 𝑏 ∫ 𝑏
′
(𝐹𝐺) d𝑥 = ( 𝑓 𝐺 + 𝐹 𝑔) d𝑥 = 𝑓 𝐺 d𝑥 + 𝐹 𝑔 d𝑥 ,
𝑎 𝑎 𝑎 𝑎

but on the other hand, by the fundamental theorem of calculus,

∫ 𝑏
(𝐹𝐺)′ d𝑥 = [𝐹 · 𝐺](𝑎) − [𝐹 · 𝐺](𝑏) = 𝐹(𝑥)𝐺(𝑥)| 𝑏𝑎 ,
𝑎

and so
∫ 𝑏 ∫ 𝑏
𝐹(𝑥)𝐺(𝑥)| 𝑏𝑎 = 𝑓 𝐺 d𝑥 + 𝐹 𝑔 d𝑥
𝑎 𝑎
∫ 𝑏 ∫ 𝑏
=⇒ 𝑓 (𝑥)𝐺(𝑥) d𝑥 = 𝐹(𝑥)𝐺(𝑥)| 𝑏𝑎 − 𝐹(𝑥)𝑔(𝑥) d𝑥 ,
𝑎 𝑎

and hence the result. ■

↩→ Theorem 5.14: Taylor’s Theorem, Remainder’s Version

Suppose 𝑓 ′ , 𝑓 ′′ , . . . , 𝑓 (𝑛) exist on [𝑎, 𝑏] and 𝑓 (𝑛+1) ∈ ℛ[𝑎, 𝑏].8 Then,

𝑓 ′(𝑎) 𝑓 ′′(𝑎) 𝑓 (𝑛) (𝑎)

𝑓 (𝑏) = 𝑓 (𝑎) + (𝑏 − 𝑎) + +···+ (𝑏 − 𝑎)𝑛 + 𝑅 𝑛 ,
1! 2! 𝑛!
∫𝑏
1
with 𝑅 𝑛 ..= 𝑛! 𝑎 𝑓 (𝑛+1) (𝑡)(𝑏 − 𝑡)𝑛 d𝑡.

Proof See Bartle; makes use of integration by parts.

Proof
Proof. ■

6 Function Sequences, Series

6.1 Pointwise and Uniform Convergence

8Remark that this is a weaker condition than continuity as was used in our previous statement of Taylor’s theorem.

6.1 Function Sequences, Series: Pointwise and Uniform Convergence 48

 ↩→ Definition 6.1: Pointwise vs Uniform Convergence
We say a sequence of functions 𝑓𝑛 → 𝑓 pointwise on a set 𝐸 if ∀ 𝑥 ∈ 𝐸, 𝑓𝑛 (𝑥) → 𝑓 (𝑥) as 𝑛 → ∞.

On the other hand, 𝑓𝑛 → 𝑓 uniformly on 𝐸 if ∀ 𝜀 > 0, ∃𝑁 ∈ N such that ∀ 𝑛 ⩾ 𝑁 and 𝑥 ∈ 𝐸,

| 𝑓𝑛 (𝑥) − 𝑓 (𝑥)| < 𝜀.

Remark 6.1. Notice that uniformly implies pointwise convergence.

⊛ Example 6.1

 2𝑛𝑥 0 ⩽ 𝑥 ⩽

 1
2𝑛
. Show that 𝑓𝑛 → 0 pointwise but not uniformly (hint: 𝑓𝑛 ( 21𝑛 ) = 1 ∀ 𝑛).

..
Let 𝑓𝑛 =
0 1
 𝑥 > 2𝑛


↩→ Theorem 6.1
The space of functions 𝐶([𝑎, 𝑏]) equipped with the sup norm is complete.

Proof Proven in tutorials.

Proof.
Proof ■

↩→ Theorem 6.2: Interchange of Limits

Let 𝐽 ⊆ R be a bounded interval such that ∃𝑥 0 ∈ 𝐽 : 𝑓𝑛 (𝑥0 ) → 𝑓 (𝑥0 ). Suppose 𝑓𝑛′ (𝑥) → 𝑔(𝑥) uniformly
∀ 𝑥 ∈ 𝐽. Then, ∃ 𝑓 : 𝑓𝑛 (𝑥) → 𝑓 (𝑥) uniformly on 𝐽, 𝑓 (𝑥) differentiable on 𝐽, and 𝑓 ′(𝑥) = 𝑔(𝑥) ∀ 𝑥 ∈ 𝐽.

Proof This is a rather painful proof; one needs to make use of the “multiple epsilons” from each given
Proof
Proof.
continuity/convergence/differentiability statement. ■
↩→ Lecture 20; Last Updated: Wed Apr 10 13:31:53 EDT 2024

↩→ Theorem 6.3
∫𝑏 ∫𝑏
Let 𝑓𝑛 ∈ ℛ[𝑎, 𝑏], 𝑓𝑛 → 𝑓 uniformly on [𝑎, 𝑏]. Then, 𝑓 ∈ ℛ[𝑎, 𝑏] and 𝑎
𝑓𝑛 (𝑥) d𝑥 → 𝑎
𝑓 (𝑥) d𝑥.

↩→ Theorem 6.4: Bounded Convergence Theorem

𝑓𝑛 ∈ ℛ[𝑎, 𝑏], 𝑓𝑛 → 𝑓 ∈ ℛ[𝑎, 𝑏], not necessarily uniformly. Suppose ∃𝐵 > 0 s.t. | 𝑓𝑛 (𝑥)| ⩽ 𝐵 ∀ 𝑥 ∈ [𝑎, 𝑏].
∫𝑏 ∫𝑏
Then, 𝑎
𝑓𝑛 → 𝑎
𝑓 as 𝑛 → ∞.

↩→ Theorem 6.5: Dimi’s Theorem/Monotone Convergence

𝑓𝑛 ∈ 𝐶([𝑎, 𝑏]), 𝑓𝑛 (𝑥) monotone (as a sequence). Suppose 𝑓𝑛 → 𝑓 ∈ 𝐶([𝑎, 𝑏]). Then, 𝑓𝑛 → 𝑓 uniformly on
[𝑎, 𝑏].

↩→ Lecture 21; Last Updated: Thu Mar 28 11:58:26 EDT 2024

6.1 Function Sequences, Series: Pointwise and Uniform Convergence 49

6.2 Series

↩→ Definition 6.2: Absolute Convergence

Let {𝑥 𝑗 } ∈ 𝑋 where 𝑋 a normed vector space (say, R). We say

∞
Õ ∞
Õ
𝑥 𝑗 converges absolutely ⇐⇒ ||𝑥 𝑗 || < +∞.
𝑗=1 𝑗=1

↩→ Theorem 6.6
Any rearrangement of absolutely convergent series given the same sum.

↩→ Definition 6.3: Conditional Convergence

Í∞ Í∞ Í∞
𝑗=1 𝑥®(𝑗) conditionally convergent if 𝑗=1 𝑥 (𝑗) converges (ie each component converges) but 𝑗=1 || 𝑥®(𝑗) || =
∞.

↩→ Theorem 6.7
Í∞
If 𝑖=1 𝑎 𝑖 ∈ R conditionally convergent, you can change the order of summation such that ∀ 𝑥 ∈ R,
Í∞
∃𝜎-permutation such that 𝑖=1 𝑎 𝜋(𝑖) = 𝑥.

Proof (Sketch) Separate 𝑎 𝑖 into positive, negative parts. Since conditionally convergent, 𝑎 𝑗 >0 𝑎 𝑗 = +∞ and
Í
Proof.
Proof
𝑎 𝑗 <0 𝑎 𝑗 = −∞. Add positive 𝑎 𝑖 ’s until the partial sum ⩾ 𝑥, then add negative 𝑎 𝑖 ’s until the partial sum ⩽ 𝑥,
Í
and repeat. The final rearrangement will converge as desired. ■
↩→ Lecture 22; Last Updated: Thu Mar 28 12:14:46 EDT 2024

↩→ Theorem 6.8
𝑣® 𝑖 ∈ R𝑛 , converges, but
Í∞ Í∞ Í∞
Suppose 𝑖=1 𝑣
®𝑖 , 𝑖=1 𝑣 𝑖 || = +∞. Then, the set of rearranged sums
||® 𝑖=1 𝑣
® 𝜎(𝑖) for
each 𝜎 : N ↔ N permutation form an affine subspace of R𝑛 .

6.3 Tests for Absolute Convergence

↩→Proposition 6.1
𝑥𝑛
Let 𝑥 𝑛 , 𝑦𝑛 be sequences and 𝑟 ..= lim𝑛→∞ 𝑦𝑛 .

Í∞ Í∞
1. If 𝑟 ≠ 0, 𝑛=1 𝑥 𝑛 converges absolutely iff 𝑛=1 𝑦𝑛 converges absolutely. In addition, if 0 < 𝑟1 ..=
𝑥𝑛 𝑥𝑛
lim inf 𝑦𝑛 ⩽ lim sup 𝑦𝑛 =: 𝑟2 < +∞, this still holds.

2. If 𝑟 = 0, and if 𝑦𝑛 converges absolutely, so does 𝑥𝑛 .

Í Í

6.3 Function Sequences, Series: Tests for Absolute Convergence 50

 ↩→Proposition 6.2: Root Test

If there ∃𝑟 < 1 such that |𝑥 𝑛 | 1/𝑛 ⩽ 𝑟 for sufficiently large 𝑛 ⩾ 𝐾, then

Í∞ Í∞
𝑛=𝐾 |𝑥 𝑛 | ⩽ 𝑛=𝐾 𝑟 −𝑛 converges.

If |𝑥 𝑛 | 1/𝑛 ⩾ 𝑛 for 𝑛 ⩾ 𝐾, 𝑥 𝑛 does not converge absolutely.

Í

↩→Proposition 6.3: Ratio Test

𝑥 𝑛+1
Let 𝑥 𝑛 ≠ 0. If ∃0 < 𝑟 < 1, ⩽ 𝑟 for sufficiently large 𝑛, 𝑥 𝑛 absolutely convergent.
Í
𝑥𝑛

𝑥 𝑛+1
If ⩾ 1 for 𝑛 ⩾ 𝐾, 𝑥 𝑛 diverges.
Í
𝑥𝑛

↩→Proposition 6.4: Integral Test

Í∞
Let 𝑓 (𝑥) ⩾ 0 be non-increasing/non-decreasing function of 𝑥 ⩾ 1. Then 𝑘=1 𝑓 (𝑘) converges ⇐⇒
∫ 𝑘
lim 𝑘→∞ 1
𝑓 (𝑥) d𝑥 finite.

↩→Proposition 6.5: Raube’s Test

Let 𝑥 𝑛 ≠ 0.

𝑥 𝑛+1
1. Suppose ∃𝑎 > 1 s.t. ⩽ 1 − 𝑛1 , 𝑛 ⩾ 𝐾. Then 𝑥 𝑛 converges absolutely.
Í
𝑥𝑛

𝑥 𝑛+1
2. If ∃𝑎 ⩽ 1 s.t. ⩾ 1 − 𝑛1 , 𝑛 ⩾ 𝐾. Then 𝑥 𝑛 does not converge absolutely.
Í
𝑥𝑛

↩→ Corollary 6.1
𝑥 𝑛+1
Let 𝑎 ..= lim 𝑛(1 − ), if such a limit exists. Then, if 𝑎 > 1, 𝑥 𝑛 converges absolutely, and if 𝑎 < 1,
Í
𝑥𝑛
𝑥 𝑛 does not.
Í

6.4 Tests for Non-Absolute Convergence

↩→Proposition 6.6: Alternating Series

If 𝑥 𝑛 > 0, 𝑥 𝑛+1 ⩽ 𝑥 𝑛 , lim𝑛→∞ 𝑥 𝑛 = 0 =⇒ (−1)𝑛 𝑥 𝑛 converges.
Í

↩→ Lemma 6.1: Abel’s Lemma

Í𝑛
Let 𝑥 𝑛 , 𝑦𝑛 ∈ R. Let 𝑠0 ..= 0, 𝑠 𝑛 ..= 𝑘=1 𝑦 𝑘 . Then, for 𝑚 > 𝑛,

𝑚
Õ 𝑚
Õ
𝑥 𝑘 𝑦 𝑘 = 𝑥 𝑚 𝑠 𝑚 − 𝑥 𝑛+1 𝑠 𝑛+1 + (𝑥 𝑘 − 𝑥 𝑘+1 )𝑠 𝑘
𝑘=𝑛+1 𝑘=𝑛+1

6.4 Function Sequences, Series: Tests for Non-Absolute Convergence 51

 ↩→ Lecture 23; Last Updated: Thu Mar 28 12:36:05 EDT 2024

↩→ Theorem 6.9: Dirichlet’s Test

Í∞
Suppose 𝑥 𝑛 decreasing and lim𝑛→∞ 𝑥 𝑛 = 0. If 𝑠 𝑛 ..= 𝑦1 + · · · + 𝑦𝑛 bounded, then 𝑛=1 𝑥 𝑛 𝑦𝑛 converges.

Proof Fix 𝐵 > 0 such that |𝑠 𝑛 | ⩽ 𝐵. 𝑥 𝑛 decreasing so 𝑥 𝑘 − 𝑥 𝑘+1 ⩾ 0. By Abel’s lemma,

Proof
Proof.

𝑚
Õ 𝑚−1
Õ
𝑥 𝑘 𝑦 𝑘 ⩽ |𝑥 𝑚 𝑠 𝑚 − 𝑥 𝑛+1 𝑠 𝑛 | + 𝑠 𝑘 (𝑥 𝑘 − 𝑥 𝑘+1 )
𝑘=𝑛+1 𝑘=𝑛+1
𝑚−1
Õ
⩽ 𝑥 𝑚 𝐵 + 𝑥 𝑛+1 𝐵 + (𝑥 𝑘 − 𝑥 𝑘+1 )𝐵
𝑘=𝑛+1
| {z }
telescopes

= 2𝑥 𝑛+1 𝐵 → 0.
𝑛→∞

■
Remark 6.2. What is red and commutes? An abelian grape!

⊛ Example 6.2: Improving Convergence

1
𝑦 𝑗 = (−1) 𝑗+1 does not converge, but |𝑠 𝑛 | ⩽ 1, so taking 𝑥 𝑛 = gives 𝑥 𝑛 𝑦𝑛 finite.
Í
𝑛 𝑛

This is an example of “improving convergence”, ie making a nearly-convergent series conver-

gence. Another example isÍby taking successive arithmetic means of a given sequence; ie let
𝑛
𝑦1 +𝑦2 𝑦𝑖
𝑎 1 = 𝑦1 , 𝑎 2 = 2 , . . . , 𝑎𝑛 = 𝑖=1
𝑛 . Then, in this case again 𝑦𝑛 = (−1)𝑛+1 , which does not converge,
has 𝑎 𝑛 → 0.

Moreover, if 𝑦𝑛 → 𝐴, then 𝑎 𝑛 → 𝐴 as well (converse does not hold in general, as above).

↩→ Theorem 6.10: Abel’s Test

Let 𝑥 𝑛 -convergent and monotone, and suppose 𝑦𝑛 converges. Then 𝑥 𝑛 𝑦𝑛 also converges.
Í Í
𝑛 𝑛

6.5 Series of Functions

↩→ Definition 6.4: Convergence

We say a series of functions 𝑛 𝑓𝑛 (𝑥) absolutely convergent on 𝐸 if | 𝑓𝑛 (𝑥)| converges for all 𝑥 ∈ 𝐸.
Í Í
𝑛

We say the convergence 𝑛 𝑓𝑛 (𝑥) → 𝑔(𝑥) is uniform if the convergence is uniform for any 𝑥 ∈ 𝐸; that
Í

is, ∀ 𝜀 > 0, ∃𝑁 ∈ N such that ∀ 𝑛 ⩾ 𝑁 and 𝑥 ∈ 𝐸, | 𝑔(𝑥) − 𝑛 𝑓𝑛 (𝑥)| < 𝜀.

Í

6.5 Function Sequences, Series: Series of Functions 52

 ↩→Proposition 6.7
Suppose for 𝑓𝑛 : 𝐸 ..= [𝑎, 𝑏] → R, 𝑛 𝑓𝑛 (𝑥) → 𝑔(𝑥) uniform for 𝑥 ∈ 𝐸, and 𝑓𝑛 ∈ ℛ[𝑎, 𝑏]. Then
Í

∫ 𝑏 ∞ ∫
Õ 𝑏
𝑔(𝑥) d𝑥 = 𝑓𝑛 (𝑥) d𝑥 .
𝑎 𝑛=1 𝑎

That is, the integral of the limit is equal to the limit of the integral.

↩→Proposition 6.8
Let 𝑓𝑛 : [𝑎, 𝑏] → R where 𝑓𝑛′ ∃ on [𝑎, 𝑏]. Suppose 𝑛 𝑓𝑛 (𝑥) converges for some 𝑥 ∈ [𝑎, 𝑏] and 𝑛 𝑓𝑛 (𝑥)
′
Í Í

converges uniformly on [𝑎, 𝑏]. Then there exists some 𝑔 : [𝑎, 𝑏] → R such that 𝑛 𝑓𝑛 → 𝑔 uniformly on
Í

[𝑎, 𝑏], 𝑔 differentiable on [𝑎, 𝑏], and 𝑔 ′(𝑥) = 𝑛 𝑓𝑛 (𝑥).

′ That is, the derivative of the limit equals the limit
Í

of the derivatives.

↩→ Lecture 24; Last Updated: Wed Apr 10 15:09:45 EDT 2024

↩→ Theorem 6.11: Cauchy Criterion

Í𝑚
𝑓𝑛 (𝑥) : 𝐷 ⊆ R → R converges uniformly on 𝐷 iff ∀ 𝜀 > 0∃𝑁 s.t. ∀ 𝑚, 𝑛 ⩾ 𝑁, 𝑖=𝑛+1 𝑓𝑖 (𝑥) < 𝜀 ∀ 𝑥 ∈ 𝐷.

↩→Proposition 6.9: Weierstrass M-Test

If | 𝑓𝑛 (𝑥)| ⩽ 𝑀𝑛 ∀ 𝑥 ∈ 𝐷 ⊆ R and 𝑀𝑛 < +∞, then 𝑛 𝑓𝑛 (𝑥) converges uniformly on 𝐷.
Í Í
𝑛

Í𝑚 Í𝑚
Proof. Suffices to look at the tail:
Proof 𝑗=𝑛+1 | 𝑓𝑛 (𝑥)| ⩽ 𝑗=𝑛+1 𝑀𝑗. ■

6.6 Power Series

↩→ Definition 6.5: Power Series

A function, series of the form
∞
Õ
𝑓 (𝑥) = 𝑎 𝑛 (𝑥 − 𝑐)𝑛 ⊛
𝑛=0

is said to be a power series centered at 𝑐 ∈ R.

↩→ Definition 6.6: Radius of Convergence

1
For 𝑎 𝑛 as in ⊛ , let 𝜌 ..= lim sup𝑛→∞ |𝑎 𝑛 |. Then, 𝑅 ..=
p
𝑛
𝜌 the radius of convergence of 𝑓 (taking 𝑅 = 0 if
𝜌 = ∞, 𝑅 = ∞ if 𝜌 = 0).

6.6 Function Sequences, Series: Power Series 53

 ↩→ Theorem 6.12: Cauchy-Hadamard
Let 𝑅 be the radius of convergence of ⊛ . Then, 𝑎 𝑛 (𝑥 − 𝑐)𝑛 converges if |𝑥 − 𝑐| < 𝑅, and diverges if
Í
𝑛
|𝑥 − 𝑐| > 𝑅. If precisely equal, either case could happen (and needs to be treated individually).

1/𝑛
Proof Directly apply the root test; assume 𝑐 = 0. Then |𝑎 𝑛 𝑥 𝑛 |
Proof
Proof. = |𝑎 𝑛 | 1/𝑛 |𝑥|. If |𝑥| < 𝜌, then lim sup |𝑎 𝑛 | 1/𝑛 |𝑥| <
𝜌 𝜌1 = 1 so ∀ 𝜀 > 0, ∃𝑁 s.t. ∀ 𝑛 ⩾ 𝑁, |𝑎 𝑛 | 1/𝑛 |𝑥| < 1 − 𝜀 =⇒ |𝑎 𝑛 | |𝑥| 𝑛 < (1 − 𝜀)𝑛 . It follows that
|𝑎 𝑛 𝑥 𝑛 | < 𝑗 ⩾ 𝑁 (1 − 𝜀) 𝑗 . But this RHS is a geometric series with 𝑟 < 1 so thus converges. The con-
Í Í
𝑗 ⩾𝑁
verse follows similarly (well, backwards). ■

⊛ Example 6.3
1. 1 + 𝑥 + 𝑥 2 + · · · converges absolutely for |𝑥| < 1.

𝑥2 𝑥3
2. 1 + 𝑥 + 2 + 3 + · · · converges for −1 ⩽ 𝑥 < 1.

𝑥𝑛
3. with 𝑘 ⩾ 2 converges for −1 ⩽ 𝑥 ⩽ 1 (check the 𝑥 = 1 case by comparison test, then the
Í
𝑛 𝑛𝑘
𝑥 = −1 test follows by alternating series test.)

↩→ Theorem 6.13
Let 𝐽 be a closed and bounded interval strictly contained in the interval of convergence of ⊛ . Then 𝑓 (𝑥)
converges uniformly in 𝐽.

↩→ Lecture 25; Last Updated: Tue Apr 9 15:14:47 EDT 2024

Remark 6.3. In-class review. Good luck!

↩→ Lecture 26; Last Updated: Tue Apr 9 15:16:14 EDT 2024

6.6 Function Sequences, Series: Power Series 54

 7 Appendix

7.1 Notes from Tutorials

↩→ Theorem 7.1
Let (𝑋 , 𝑑) be a compact metric space.9 Let 𝐶(𝑋) := { 𝑓 : 𝑋 → R : 𝑓 continuous} be a vector space. Take
the uniform norm || 𝑓 || := sup𝑥∈𝑋 | 𝑓 (𝑥)| on 𝐶(𝑥). Then, (𝐶(𝑥), || • ||) is complete.10

Proof Denote the “canonical norm” 𝜌( 𝑓 , 𝑔) := || 𝑓 − 𝑔||.

Proof
Proof.
Let ( 𝑓𝑛 ) ∈ 𝐶(𝑋) be a Cauchy sequence. Then, ∀ 𝜀 > 0, ∃𝑁 ∈ N : ∀ 𝑚, 𝑛 ⩾ 𝑁 , 𝜌( 𝑓𝑛 , 𝑓𝑚 ) < 𝜀.
Fix 𝑥 ∈ 𝑋, noting that

| 𝑓𝑛 (𝑥) − 𝑓𝑚 (𝑥)| ⩽ sup | 𝑓𝑛 (𝑦) − 𝑓𝑚 (𝑦)| = 𝜌( 𝑓𝑛 , 𝑓𝑚 ) < 𝜀. ∗1

𝑦∈𝑋

Define, for this fixed 𝑥, a sequence in R { 𝑓𝑛 (𝑥)} 𝑛∈N . By ∗1 , we have that this sequence is Cauchy in R, but as R
complete, 𝑓𝑛 (𝑥) hence converges, to some limit we call 𝑓 (𝑥) := lim𝑛→∞ 𝑓𝑛 (𝑥). Note that 𝑥 is still fixed at this
point; these are but real numbers we are working with here.
Now, as 𝑥 was completely arbitrary, we can repeat this process for all of 𝑋, and define a function 𝑓 : 𝑋 → R
where 𝑓 (𝑥) := lim𝑛→∞ 𝑓𝑛 (𝑥).
For a fixed 𝑥, we have that 𝑓𝑚 (𝑥) → 𝑓 (𝑥) as 𝑚 → ∞. This implies:

0 ⩽ lim | 𝑓𝑛 (𝑥) − 𝑓𝑚 (𝑥)| ⩽ lim 𝜀 = 𝜀

𝑚→∞ 𝑚→∞

=⇒ | 𝑓𝑛 (𝑥) − 𝑓 (𝑥)| ⩽ 𝜀 ∀ 𝑛 ⩾ 𝑁
=⇒ 𝜌( 𝑓𝑛 , 𝑓 ) = sup | 𝑓𝑛 (𝑥) − 𝑓 (𝑥)| ⩽ 𝜀 =⇒ 𝑓𝑛 → 𝑓
𝑥∈𝑋

It remains to show that 𝑓 ∈ 𝐶(𝑋). Let 𝑐 ∈ 𝑋 and 𝜀 > 0, and the corresponding 𝑁 ∈ N : 𝜌( 𝑓𝑛 , 𝑓 ) < 3𝜀 ∀ 𝑛 ⩾ 𝑁.
By construction, 𝑓𝑁 ∈ 𝐶(𝑋), and is thus continuous at 𝑐. This gives that ∃𝛿 > 0 : | 𝑓𝑁 (𝑥) − 𝑓𝑁 (𝑐)| < 3𝜀 whenever
𝑑(𝑥, 𝑐) < 𝛿. 11
Hence, if 𝑑(𝑥, 𝑐) < 𝛿, we have

| 𝑓 (𝑥) − 𝑓 (𝑐)| ⩽ | 𝑓 (𝑥) − 𝑓𝑁 (𝑥)| + | 𝑓𝑁 (𝑥) − 𝑓𝑁 (𝑐)| + | 𝑓𝑁 (𝑐) − 𝑓 (𝑐)|

𝜀 𝜀
⩽ 𝜌( 𝑓 , 𝑓𝑁 ) + +
3 3
𝜀 𝜀 𝜀
< + + = 𝜀,
3 3 3

hence 𝑓 continuous at 𝑐, which was completely arbitrary, and thus 𝑓 ∈ 𝐶(𝑋). ■

10In this proof, the compactness is necessary for the norm to be well-defined.
10In this way, this becomes a Banach Space: a complete, normed vector space.
11Be careful here, there are three different metrics going on; 𝜌 from the vector space, 𝑑 from the underlying metric space, and |· · · |
from R.

7.1 Appendix: Notes from Tutorials 55

 ↩→ Theorem 7.2
Let (𝑋 , 𝑑)-complete. Let {𝐹𝑛 } be a decreasing family of non-empty closed sets with lim𝑛→∞ diam(𝐹𝑛 ) = 0.
Then, ∃𝑧 : 𝐹𝑛 = {𝑧}.
Ñ
𝑛∈N

↩→ Theorem 7.3
Let (𝑋 , 𝑑)-complete, and 𝑓 : 𝑋 → 𝑋 an “expanding map”, such that 𝑑(𝑥, 𝑦) ⩽ 𝑑( 𝑓 (𝑥), 𝑓 (𝑦)) ∀ 𝑥, 𝑦 ∈ 𝑋.
Then, 𝑓 is a surjective isometry, ie, 𝑓 (𝑋) = 𝑋 and 𝑑( 𝑓 (𝑥), 𝑓 (𝑦)) = 𝑑(𝑥, 𝑦) ∀ 𝑥, 𝑦 ∈ 𝑋.

↩→ Lemma 7.1
Differentiable =⇒ Continuous.

𝑓 (𝑥)− 𝑓 (𝑐)
Proof Let 𝑓 : 𝐼 → R, and 𝑐 ∈ 𝐼 arbitrary. Notice that ∀ 𝑥 ≠ 𝑐 ∈ 𝐼, 𝑓 (𝑥) − 𝑓 (𝑐) = (𝑥 − 𝑐)
Proof
Proof. 𝑥−𝑐 . Hence,

𝑓 (𝑥) − 𝑓 (𝑐)
lim ( 𝑓 (𝑥) − 𝑓 (𝑐)) = lim (𝑥 − 𝑐)
𝑥→𝑐 𝑥→𝑐 𝑥−𝑐
𝑓 (𝑥) − 𝑓 (𝑐)
= lim (𝑥 − 𝑐) · lim
𝑥→𝑐 𝑥→𝑐 𝑥−𝑐
′
= 0 · 𝑓 (𝑥) = 0
=⇒ lim 𝑓 (𝑥) = 𝑓 (𝑐),
𝑥→𝑐

hence 𝑓 continuous, noting that the splitting of the limits is valid as both are defined. ■

7.1 Appendix: Notes from Tutorials 56

 ⊛ Example 7.1

 𝑥2 𝑥∈Q



Let 𝑓 : R → R, 𝑓 (𝑥) :=
0
 𝑥∉Q

Claim: 𝑓 discontinuous at all 𝑥 ≠ 0.
Claim:

Proof Let 𝑥 ≠ 0 ∈ R. By density of Q ⊆ R, there exist sequences (𝑟𝑛 ) ∈ Q s.t. 𝑟𝑛 → 𝑥 and (𝑧 𝑛 ) ∈ R \ Q

Proof
Proof.
s.t. 𝑧 𝑛 → 𝑥. Then:

lim 𝑓 (𝑟𝑛 ) = lim 𝑟𝑛2 = 𝑥 2

𝑛→∞

lim 𝑓 (𝑧 𝑛 ) = lim 0,
𝑛→∞

hence 𝑓 discontinuous by the sequential criterion at 𝑥 ≠ 0. ■

Claim: 𝑓 ′(0) = 0.
Claim:

Proof Let 𝜀 > 0 and 𝛿 = 𝜀. Notice that 𝑓 (𝑥) ⩽ 𝑥 2 ∀ 𝑥. Then, we have that ∀ |𝑥| < 𝛿,
Proof
Proof.

𝑓 (𝑥) − 𝑓 (0) 𝑓 (𝑥)

−0 =
𝑥−0 𝑥
𝑥2
⩽ = |𝑥| < 𝛿 = 𝜀
𝑥

↩→ Definition 7.1
Let 𝑓 : 𝐼 → R. A point 𝑐 ∈ 𝐼 is a local max (resp min) if ∃𝛿 > 0 s.t. 𝑓 (𝑥) ⩽ 𝑓 (𝑐) (resp 𝑓 (𝑥) ⩾ 𝑓 (𝑐))
∀ 𝑥 ∈ (𝑐 − 𝛿, 𝑐 + 𝛿) ∩ 𝐼.

↩→ Lemma 7.2
Let 𝑓 : 𝐼 → R be differentiable at 𝑐 ∈ 𝐼 ◦ . If 𝑐 a local extrema of 𝑓 , then 𝑓 ′(𝑐) = 0.

Proof Assume wlog that 𝑐 a local max; if a local min, take 𝑓˜ := − 𝑓 and continue.
Proof
Proof.
Since 𝐼 ◦ open, ∃𝛿1 > 0 : (𝑐 − 𝛿 1 , 𝑐 + 𝛿 1 ) ⊆ 𝐼 ◦ ⊆ 𝐼. We also have that ∃𝛿 2 > 0 : 𝑓 (𝑥) ⩽ 𝑓 (𝑐) ∀ 𝑥 ∈
(𝑐 − 𝛿2 , 𝑐 + 𝛿2 ) ∩ 𝐼, by 𝑐 an extrema.
Let 𝛿 := min{𝛿1 , 𝛿2 }. Then, we have both (𝑐 − 𝛿, 𝑐 + 𝛿) ⊆ 𝐼 and 𝑓 (𝑥) ⩽ 𝑓 (𝑐) ∀ 𝑥 ∈ (𝑐 − 𝛿, 𝑐 + 𝛿).
𝑓 (𝑥)− 𝑓 (𝑐) 𝑓 (𝑥)− 𝑓 (𝑐)
Since 𝑓 ′(𝑐) exists, lim𝑥→𝑐 + 𝑥−𝑐 = lim𝑥→𝑐 − 𝑥−𝑐 . But we have from the property of being a maximum

7.1 Appendix: Notes from Tutorials 57

 that

𝑓 (𝑥) − 𝑓 (𝑐) 𝑓 (𝑥) − 𝑓 (𝑐)

lim+ ⩾ 0, lim− ⩽ 0,
𝑥→𝑐 𝑥−𝑐 𝑥→𝑐 𝑥−𝑐

hence, as these two limits must agree, they must equal 0 and thus 𝑓 ′(𝑐) = 0. ■

7.2 Miscellaneous

⊛ Example 7.2: Rudin, Chapter 7: Differentiability

1. Let 𝑓 be defined ∀ 𝑥 ∈ R, and suppose that | 𝑓 (𝑥) − 𝑓 (𝑦)| ⩽ (𝑥 − 𝑦)2 , ∀ 𝑥, 𝑦 ∈ R. Prove that 𝑓 is
constant.12

Proof Let 𝑥 > 𝑦 ∈ R. Then, as |𝑥 − 𝑦| = 𝑥 − 𝑦, we have

Proof
Proof.

𝑓 (𝑥) − 𝑓 (𝑦)
| 𝑓 (𝑥) − 𝑓 (𝑦)| ⩽ (𝑥 − 𝑦)2 =⇒ ⩽ 𝑥 − 𝑦 = |𝑥 − 𝑦| → 0 as 𝑦 → 𝑥
𝑥−𝑦
𝑓 (𝑥) − 𝑓 (𝑦)
=⇒ →0
𝑥−𝑦

This implies, then, that 𝑓 ′(𝑥) is defined ∀ 𝑥 ∈ R, and moreover, that 𝑓 ′(𝑥) = 0 ∀ 𝑥 ∈ R. We
conclude, then, that 𝑓 (𝑥) constant ∀ 𝑥 ∈ R. ■

2. Suppose 𝑓 ′(𝑥) > 0 in (𝑎, 𝑏). Prove that 𝑓 is strictly increasing in (𝑎, 𝑏), and let 𝑔 be its inverse
function. Prove that 𝑔 is differentiable, and that

1
𝑔 ′( 𝑓 (𝑥)) = (𝑎 < 𝑥 < 𝑏).
𝑓 ′(𝑥)

𝑓 (𝑥)− 𝑓 (𝑦)
Proof Fix 𝑥 > 𝑦 ∈ (𝑎, 𝑏). Then, by the mean value theorem, ∃𝑧 ∈ (𝑥, 𝑦) : 𝑓 ′(𝑧) =
Proof
Proof. 𝑥−𝑦 .
Since 𝑓 ′(𝑧) > 0, it follows that

𝑓 (𝑥) − 𝑓 (𝑦)
> 0 =⇒ 𝑓 (𝑥) − 𝑓 (𝑦) > 𝑥 − 𝑦 > 0 =⇒ 𝑓 (𝑥) > 𝑓 (𝑦),
𝑥−𝑦

hence, 𝑓 increasing, as 𝑥 > 𝑦 arbitrary.

Let now 𝑔 := 𝑓 −1 . ■

12Note that this means that 𝑓 Hölder continuous with constant 𝛼 = 2. Indeed, Hölder continuous functions with 𝛼 > 1 are always
constant by a similar proof. For 0 < 𝛼 ⩽ 1, we have the inclusion continuously differentiable =⇒ Lipschitz =⇒ 𝛼−Hölder =⇒
uniformly continuous =⇒ continuous.

7.2 Appendix: Miscellaneous 58

 7.3 Class Midterm Solutions

↩→ Question 7.1
Let 𝑋 be a topological space, and let 𝑓 , 𝑔 : 𝑋 → R be two continuous functions. Show that the set
{𝑥 ∈ 𝑋 : 𝑓 (𝑥) > 𝑔(𝑥)} is an open subset of 𝑋.

Proof Let 𝐴 ..= {𝑥 ∈ 𝑋 : 𝑓 (𝑥) > 𝑔(𝑥)}. Letting 𝜑(𝑥) ..= 𝑓 (𝑥) − 𝑔(𝑥) = ( 𝑓 − 𝑔)(𝑥), then remark that
Proof
Proof.
𝐴 ≡ {𝑥 ∈ 𝑋 : 𝜑(𝑥) > 0}, and since differences of continuous functions are continuous, 𝜑 continuous.
Letting 𝐵 ..= (0, ∞) ⊆ R, then, we have that 𝐴 = 𝜑−1 (𝐵). But 𝐵 an open set, and the inverse images of
open sets by continuous functions are open, hence 𝐴 open. ■

↩→ Question 7.2
(a) List three equivalent properties (definitions) of compact sets in metric spaces; you don’t need to
prove anything.

(b) Is the unit ball13 in the space ℓ 2 of infinite sequences compact? Prove or disprove. You may use any
of the properties from (a).

Proof
Proof.
Proof (a) Every open cover admits a finite subcover ⇐⇒ sequentially compact ⇐⇒ complete and
totally bounded.

(b) Denote the closed unit ball centered at (0, 0, . . . ) in ℓ 2 , 𝐵 ..= {𝑥 ∈ ℓ 2 : 𝑑22 (0, 𝑥) =
Í∞
𝑖=1 |𝑥 𝑖 | ⩽ 1}.
Consider the sequence of “unit sequences”

{𝑒 𝑛 } 𝑛∈N ∈ 𝐵, 𝑒 𝑖𝑛 ..= 𝛿 𝑖𝑛 .
√
Then, for any 𝑖 ≠ 𝑗, 𝑑2 (𝑒 𝑛 , 𝑒 𝑚 ) = 2 > 1. It follows that, although 𝑒 𝑛 ∈ 𝐵 for any 𝑛, there cannot exist
a subsequence of 𝑥 𝑛 that converges within 𝐵 (verify why this is!). Thus, 𝐵 cannot be sequentially
compact and thus not compact.

13Jakobson said in class this is supposed to be a closed ball.

7.3 Appendix: Class Midterm Solutions 59

 ↩→ Question 7.3
(a) Define a complete metric space.

(b) State (without proof) the contraction mapping theorem.

(c) Let 𝑓 : (0, 1) → (0, 1) be defined by 𝑓 (𝑥) = 𝑥/2. Is 𝑓 a contraction?

(d) Does 𝑓 have a fixed point in the open interval 𝐼 = (0, 1)? Does that contradict the contraction
mapping theorem?

Proof
Proof. (a) A complete metric space is a metric space in which every Cauchy sequence converges within
that space.

(b) Let (𝑋 , 𝑑) be a complete metric space, and let 𝑓 : 𝑋 → 𝑋 be a contraction mapping, ie for any
𝑥, 𝑦 ∈ 𝑋, 𝑑( 𝑓 (𝑥), 𝑓 (𝑦)) ⩽ 𝑐 · 𝑑(𝑥, 𝑦) for some 𝑐 ∈ (0, 1). Then, the contraction mapping states that 𝑓
has a unique fixed point 𝑧 ∈ 𝑋, ie 𝑓 (𝑧) = 𝑧 and lim𝑛→∞ 𝑓 (𝑛) (𝑥) = 𝑧 for any 𝑥.

(c) For any 𝑥, 𝑦 ∈ (0, 1), we have

𝑥−𝑦 1
𝑑( 𝑓 (𝑥), 𝑓 (𝑦)) = | 𝑓 (𝑥) − 𝑓 (𝑦)| = = |𝑥 − 𝑦| = 𝑐 · 𝑑(𝑥, 𝑦),
2 2

so 𝑓 indeed a contraction mapping with 𝑐 ..= 12 .

𝑥 𝑥 𝑥
(d) We have that for any 𝑥 ∈ 𝐼, 𝑓 (𝑛) (𝑥) = 2𝑛 so 𝑥 a fixed point iff 2𝑛 = 2𝑛−1
for some 𝑛, which is only
possible if 𝑥 = 0, but 0 ∉ 𝐼, so indeed 𝑓 has no fixed point in 𝐼. This is not a contradiction to the
1 1
contraction mapping theorem since 𝐼 ..= (0, 1) not complete (indeed, 𝑛 ∈ 𝐼 ∀ 𝑛 but 𝑛 → 0 ∉ 𝐼).

7.3 Appendix: Class Midterm Solutions 60

 ↩→ Question 7.4
Let 𝑥 = (𝑥1 , 𝑥2 , . . . ) and 𝑦 = (𝑦1 , 𝑦2 , . . . ) be infinite real sequences satisfying ||𝑥|| 2 ⩽ 2 and ||𝑦|| 2 ⩽ 3,
where || · · · || 2 the ℓ 2 norm.

(a) State Holder’s inequality and Minkowski inequality for sequences.

(b) Give an upper bound for 𝑥 · 𝑦 = 𝑥 𝑖 𝑦 𝑖 , and for ||𝑥 + 𝑦||.

Í
𝑖

Proof
Proof.
Proof (a) Holder’s inequality: for 𝑝, 𝑞 Holder conjugates and 𝑥 ∈ ℓ 𝑝 , 𝑦 ∈ ℓ 𝑞 we have

Õ
𝑥 𝑖 𝑦 𝑖 ⩽ ||𝑥|| 𝑝 ||𝑦|| 𝑞 .
𝑖=1

Minkowski inequality: for 𝑥, 𝑦 ∈ ℓ 𝑝 ,

||𝑥 + 𝑦|| 𝑝 ⩽ ||𝑥|| 𝑝 + ||𝑦|| 𝑝 .

(b) For 𝑥, 𝑦 as given; by Holders, 𝑥 · 𝑦 ⩽ ||𝑥|| 𝑝 ||𝑦|| 𝑞 = 2 · 3 = 6, and by Minkowski’s, ||𝑥 + 𝑦|| ⩽
||𝑥|| + ||𝑦|| = 2 + 3 = 5, so 6, 5 are upper bounds for 𝑥 · 𝑦, ||𝑥 + 𝑦|| respectively.

7.3 Appendix: Class Midterm Solutions 61

 ↩→ Question 7.5
(a) State (without proof) Taylor’s theorem.

(b) Let 𝑓 ∈ 𝐶 4 ([0, 2]), and let 𝑓 ′(1) = 𝑓 ′′(1) = 𝑓 ′′′(1) = 0 while 𝑓 (4) (1) = 2. Use (a) to show that 𝑓 (𝑥) has
a local extremum at 𝑥 = 1, and determine its type.

Proof
Proof.
Proof (a) Let 𝐼 ..= [𝑎, 𝑏] ⊆ R and let 𝑓 : 𝐼 → R such that 𝑓 ∈ 𝐶 𝑛 (𝐼), and 𝑓 (𝑛+1) (𝑥) exists on (𝑎, 𝑏). Then,
for 𝑥0 ∈ [𝑎, 𝑏], there exists some 𝑐 ∈ (min(𝑥, 𝑥0 ), max(𝑥, 𝑥0 )) such that

𝑓 (𝑛) (𝑥0 ) 𝑛 𝑓 (𝑛+1) (𝑐)

′
𝑓 (𝑥) = 𝑓 (𝑥0 ) + 𝑓 (𝑥0 )(𝑥 − 𝑥0 ) + · · · + (𝑥 − 𝑥0 ) + (𝑥 − 𝑥0 )𝑛+1 .
𝑛! (𝑛 + 1)!

(b) By Taylor’s, for any 𝑥 ∈ [0, 2], there exists some 𝑐 between 𝑥 and 1 such that

𝑓 (4) (𝑐)
𝑓 (𝑥) = 𝑓 (1) + 𝑓 ′(1)(· · · ) + 𝑓 ′′(1)(· · · ) + 𝑓 ′′′(1)(· · · ) + (𝑥 − 1)4
| {z } 4!
=0
𝑓 (4) (𝑐)
= 𝑓 (1) + (𝑥 − 1)4
4!
𝑓 (4) (𝑐)
=⇒ 𝑓 (𝑥) − 𝑓 (1) ⩾ (𝑥 − 1)4 ∀ 𝑥 ∈ [0, 2]
4!

By continuity of 𝑓 (4) , there exists some neighborhood 𝑉 of 𝑥0 = 1 such that 𝑓 (4) (𝑐) has the same
𝑓 (4) (𝑐) 𝑓 (4)(1) 2
sign of 𝑓 (4) (1). So, for any 𝑥 ∈ 𝑉, 4! ⩾ 0, since 4! = 4! ⩾ 0. Thus, since (𝑥 − 1)4 ⩾ 0, we have
that for such 𝑥 in 𝑉,
𝑓 (𝑥) − 𝑓 (1) ⩾ 0 =⇒ 𝑓 (𝑥) ⩾ 𝑓 (1).

Hence, we have a neighborhood of 1 such that for all 𝑥 in the neighborhood 𝑓 (𝑥) ⩾ 𝑓 (1). It follows
that 1 a local minimum of 𝑓 .

7.3 Appendix: Class Midterm Solutions 62

8 List of Theorems

↩→ Definition 1.1 (Metric Space) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

↩→ Definition 1.2 (Open Metric Space) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
↩→ Definition 1.3 (Normed Space) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
↩→Proposition 1.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
↩→Proposition 1.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
↩→ Definition 1.4 (Convex Set) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
↩→ Definition 1.5 (ℓ 𝑝 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
↩→Proposition 1.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
↩→ Definition 2.1 (Topological space) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
↩→ Definition 2.2 (Closed sets) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
↩→ Definition 2.3 (Equivalence of Metrics) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
↩→ Definition 2.4 (Interior, Boundary of a Topological Set) . . . . . . . . . . . . . . . . . . . . . . . . . 8
↩→ Definition 2.5 (Closure) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
↩→Proposition 2.1 (Properties of Int(𝐴)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
↩→Proposition 2.2 (Properties of 𝐴) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
↩→Proposition 2.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
↩→ Definition 2.6 (Basis for a Toplogy) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
↩→Proposition 2.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
↩→Proposition 2.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
↩→ Definition 2.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
↩→Proposition 2.6 (Consequences of Subspace Topologies) . . . . . . . . . . . . . . . . . . . . . . . . 10
↩→Proposition 2.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
↩→ Lemma 2.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
↩→ Lemma 2.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
↩→ Definition 2.8 (Continuous Function) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
↩→Proposition 2.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
↩→Proposition 2.9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
↩→ Theorem 2.1 (Continuity of Composition) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
↩→Proposition 2.10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
↩→ Definition 2.9 (Finite Product Spaces) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
↩→ Definition 2.10 (Cylinder Set) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
↩→ Definition 2.11 (Projection) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
↩→ Definition 2.12 (Coordinate Function) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
↩→Proposition 2.11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
↩→ Definition 2.13 (Product Topology/Cylinder Sets for ∞ Products) . . . . . . . . . . . . . . . . . . 14
↩→Proposition 2.12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
↩→Proposition 2.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
↩→ Question 2.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
↩→ Definition 2.14 (Compact) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
↩→Proposition 2.14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
↩→ Theorem 2.2 (On Compactness) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
↩→Proposition 2.15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
↩→Proposition 2.16 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
↩→Proposition 2.17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
↩→ Definition 2.15 (Connected) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
↩→ Theorem 2.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

63
↩→ Lemma 2.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
↩→ Theorem 2.4 (“Intermediate Value Theorem”) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
↩→ Theorem 2.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
↩→ Theorem 2.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
↩→ Definition 2.16 (Path Connected) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
↩→ Theorem 2.7 (Path connected =⇒ connected) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
↩→Proposition 2.18 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
↩→ Definition 2.17 (Connected Component) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
↩→ Definition 2.18 (Path Component) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
↩→Proposition 2.19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
↩→ Lemma 2.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
↩→ Lemma 2.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
↩→ Lemma 2.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
↩→Proposition 2.20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
↩→Proposition 2.21 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
↩→ Definition 2.19 (An Explicit Definition) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
↩→ Definition 2.20 (Complement Definition) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
↩→ Definition 3.1 (Hölder Conjugates) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
↩→Proposition 3.1 (Hölder’s Inequality) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
↩→ Lemma 3.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
↩→Proposition 3.2 (Minkowski Inequality) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
↩→ Theorem 3.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
↩→ Definition 3.2 (Complete Metric Space) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
↩→Proposition 3.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
↩→Proposition 3.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
↩→Proposition 3.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
↩→Proposition 3.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
↩→ Definition 3.3 (Contraction Mapping) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
↩→ Theorem 3.2 (Contraction Mapping Theorem) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
↩→ Theorem 3.3 (ℓ 𝑝 complete) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
↩→ Definition 3.4 (Totally Bounded) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
↩→ Theorem 3.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
↩→ Lemma 3.1 (Cantor Intersection Theorem) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
↩→ Definition 3.5 (Finite Intersection Property) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
↩→ Lemma 3.2 (Finite Interesection Formulation of Compactness) . . . . . . . . . . . . . . . . . . . . 33
↩→ Lemma 3.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
↩→ Lemma 3.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
↩→ Theorem 3.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
↩→ Definition 4.1 (Differentiable) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
↩→ Definition 4.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
↩→ Theorem 4.1 (Caratheodory’s Theorem) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
↩→ Theorem 4.2 (Chain Rule) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
↩→ Definition 4.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
↩→ Theorem 4.3 (Rolle’s) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
↩→ Theorem 4.4 (Gauss-Lucas) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
↩→ Definition 4.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
↩→ Theorem 4.5 (Mean Value) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
↩→ Theorem 4.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
↩→Proposition 4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
↩→ Lemma 4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

64
↩→ Theorem 4.7 (Darboux) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
↩→Proposition 4.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
↩→ Theorem 4.8 (Cauchy Mean Value) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
↩→Proposition 4.3 (More General L’Hopital) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
↩→Proposition 4.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
↩→ Theorem 4.9 (Taylor’s Theorem) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
↩→ Theorem 4.10 (Relative Extrema) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
↩→ Definition 4.5 (Convex Set) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
↩→ Definition 4.6 (Convex Function) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
↩→ Definition 5.1 (Partitions) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
↩→ Definition 5.2 (Riemann Sum) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
↩→ Definition 5.3 (Riemann Integrable) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
↩→Proposition 5.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
↩→Proposition 5.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
↩→Proposition 5.3 (Cauchy Criterion for Integrability) . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
↩→ Theorem 5.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
↩→ Lemma 5.1 (BS-7.2.4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
↩→ Theorem 5.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
↩→ Theorem 5.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
↩→ Theorem 5.4 (BS-7.2.7) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
↩→ Theorem 5.5 (Additivity; BS-7.2.8) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
↩→ Definition 5.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
↩→ Theorem 5.6 (Fundamental Theorem of Calculus) . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
↩→ Definition 5.5 (Upper/Lower Riemann Sums) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
↩→Proposition 5.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
↩→Proposition 5.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
↩→ Definition 5.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
↩→ Theorem 5.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
↩→ Theorem 5.8 (Another Fundamental Theorem of Calculus) . . . . . . . . . . . . . . . . . . . . . . 46
↩→ Corollary 5.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
↩→ Theorem 5.9 (Substitution/Change of Variables) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
↩→ Definition 5.7 (Lebesgue Measure 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
↩→ Theorem 5.10 (Lebesgue Integrability Criterion) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
↩→ Theorem 5.11 (Composition) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
↩→ Theorem 5.12 (Product Theorem) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
↩→ Theorem 5.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
↩→ Theorem 5.14 (Taylor’s Theorem, Remainder’s Version) . . . . . . . . . . . . . . . . . . . . . . . . 48
↩→ Definition 6.1 (Pointwise vs Uniform Convergence) . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
↩→ Theorem 6.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
↩→ Theorem 6.2 (Interchange of Limits) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
↩→ Theorem 6.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
↩→ Theorem 6.4 (Bounded Convergence Theorem) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
↩→ Theorem 6.5 (Dimi’s Theorem/Monotone Convergence) . . . . . . . . . . . . . . . . . . . . . . . . 49
↩→ Definition 6.2 (Absolute Convergence) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
↩→ Theorem 6.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
↩→ Definition 6.3 (Conditional Convergence) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
↩→ Theorem 6.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
↩→ Theorem 6.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
↩→Proposition 6.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
↩→Proposition 6.2 (Root Test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

65
↩→Proposition 6.3 (Ratio Test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
↩→Proposition 6.4 (Integral Test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
↩→Proposition 6.5 (Raube’s Test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
↩→ Corollary 6.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
↩→Proposition 6.6 (Alternating Series) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
↩→ Lemma 6.1 (Abel’s Lemma) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
↩→ Theorem 6.9 (Dirichlet’s Test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
↩→ Theorem 6.10 (Abel’s Test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
↩→ Definition 6.4 (Convergence) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
↩→Proposition 6.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
↩→Proposition 6.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
↩→ Theorem 6.11 (Cauchy Criterion) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
↩→Proposition 6.9 (Weierstrass M-Test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
↩→ Definition 6.5 (Power Series) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
↩→ Definition 6.6 (Radius of Convergence) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
↩→ Theorem 6.12 (Cauchy-Hadamard) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
↩→ Theorem 6.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
↩→ Theorem 7.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
↩→ Theorem 7.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
↩→ Theorem 7.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
↩→ Lemma 7.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
↩→ Definition 7.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
↩→ Lemma 7.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
↩→ Question 7.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
↩→ Question 7.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
↩→ Question 7.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
↩→ Question 7.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
↩→ Question 7.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

66