UMA101: Analysis and Linear Algebra

Shreeansh Hota
15 November 2023

§1 Basic Set Theory

Definition 1.1. A set is a well defined collection of objects. Remarks on the formal
definition.
ˆ The meaning of collection and objects are taken to be self evident.
ˆ A collection is said to be well defined if it is described by one rule in a specific
list called Axioms of Set Theory (ZF Axioms)
Some of the axioms of set theory are:
ˆ The collection with no elements is a set (Null set)
ˆ Two sets are said to be equal if they consist of the same list of elements
ˆ Axiom of Union: Given a non null set C of sets. The collection all of whose
elements belong to some set A ∈ C, and such that every S element of every A ∈ C
belongs to this collection, is a set. This set is denoted by A∈C A
Given sets A and B, the set difference A − B is the set of all elements in A that are not
in B. (The existence of such a set is guaranteed by the set builder axiom)

Theorem 1.2 (De Morgan’s Laws)

Let X be a set. Let C be a non null set whose elements are subsets of X. Then,
T S
1. X − A∈C A = A∈C (X − A)
S T
2. X − A∈C A = A∈C (X − A)

Question: Why do we need to codify the term well defined

Answer: It leads to contradictions. For example, consider the following,
A = {x ∈ S : x ∈
/ x}
where S is the set of all sets. The problem arises when we try to figure out whether
A ∈ A or not. This is called Russell’s Paradox

§2 Numbers
§2.1 Natural Numbers
Our approach will deviate from Apostol. The natural numbers (denoted as N) is a
collection of elements that satisfy a list of rules called the Peano Axioms

1
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Peano Axioms
1. There exists a special element in N called 0

2. For each n ∈ N, there is an unique element called the successor of n, denoted by

S(n)

3. 0 is not the successor of any n ∈ N

4. If m and n are two natural numbers and S(m) = S(n), then m = n

5. For each n ∈ N, let Σ(n) denote a statement involving n. If Σ(0) is true, and
Σ(S(n)) is true whenever Σ(n) is true, then Σ(n) is true for all n ∈ N

Remarks: It can be proven that N is a set. A natural number is simply any element of
N

Arithmetic
ˆ Peano Addition: n + 0 := n, ∀n ∈ N
n + S(m) = S(m + n), ∀m, n ∈ N

ˆ Peano Multiplication: n.0 := 0

n.S(m) = n.m + n

§2.2 Integers and Rational

Integers are denoted by Z. This set was created so as to realise the closure law with
respect to subtraction.
Rational are denoted by Q. This set was created to realise closure with respect to
division.

Field
A field (denoted by F) is a set equipped with two rules + and · defined for every pair of
elements x, y ∈ F, such that they obey,

ˆ Axiom 0: x + y ∈ F and x · y ∈ F

ˆ Axioms 1-6 given in Apostol

§2.3 Real Numbers

The real numbers (denoted by R) are collectively the ordered field containing Q as an
ordered sub-field and has the Lowest Upper Bound Property.

Definition 2.1. We can equip Q with a symbol > such that for any pair a, b ∈ Q, the
following properties hold :

ˆ a = b or a > b or b > a

ˆ a, b, c ∈ Q, and if a > b and b > c, then a > c

We write a ≥ b if either a > b or a = b

2
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Question: How can we tell xn> y, x, y ∈ Q+

a o
Answer: Step 1: Define Q+ := : a, b ∈ N − {0}
b
In addition to obeying Axioms 0-6, Q+ also obeys:
ˆ x, y ∈ Q+ =⇒ x + y ∈ Q+

ˆ 0∈
/ Q+

ˆ ∀x ∈ Q − {0}, exactly one of x and −x is in Q+

Step 2: We define for x, y ∈ Q, x > y ⇔ x − y ∈ Q+
Let (F, +, ·) be a field. We say F is an ordered field if ∃ a set F+ ⊂ F, such that (F, F+ )
satisfy the above axioms with Q replaced by F
Using the above rules, we assume that R is a ordered field. But we run into a few
problems.

ˆ We are demanding so many properties from R. Could these co-exist?

ˆ We will assume they co-exist in this course. This is a reasonable assumption because
Dedekind constructed the real numbers.

Let F be an ordered field and let > be given in F+ ⊂ F. Let S ̸= ϕ be a subset of F

1. u ∈ F is called an upper bound if u ≥ x for all x in S

2. We say S is bounded above if S has some upper bound

3. An element b ∈ F is called as a Least Upper Bound of S if,

ˆ b is an upper bound of S
ˆ If x ∈ F such that x < b, then x is not an upper bound of S

Problem (Self Thinking)

If F and S are as above, formulate definition for: Lower bound of S, S is bounded
below and Greatest lower bound of S

Definition 2.2. An ordered field has Least Upper Bounded property if any S ̸= ϕ,
S ⊂ F that is bounded above has a Least Upper Bound property.
Remark: Q does not have Least Upper Bound (consider the set S = {x ∈ a : 0 <
x2 < 2}, this doesn’t have a LUB in Q.

Problem 2.3
N as a subset of either Q or R is not bounded above.

Proof. First N ⊂ R. Assume N is bounded above, let u ∈ R be the LUB. Thus, u − 1 is

not an upper bound. Thus, there exists a natural number n such that,

n>u−1

But this means that n + 1 > u, and n + 1 ∈ N. Contradiction!

3
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 2.4 (Archimedean Property)

Let x ∈ R and x > 0. Given y ∈ R, ∃n ∈ P such that nx > y

Proof. We have two cases.

Case 1: y ≤ 0. In this case n = 1 works.
Case 2: y > 0. Assume that no such n exists. So, nx ≤ y, ∀n ∈ P. This implies that
n ≤ x−1 y. This contradicts the fact that N is not bounded above.

§3 Sequences
Definition 3.1. Let S ̸= ϕ be a set. A sequence is a function f : P → S.

Historically, we have denoted sequences as a1 , a2 , a3 , · · · So, we drop the functional

notation.
f (n) := an ∈ S, and denote the sequence as {an }∞
n=1 , or simply as {an }. an is known
as the nth term of the sequence.

Definition 3.2. Let {an } be a sequence in R. We say {an } converges to L (Alternatively,

has limit L), L ∈ R, if for any ϵ > 0, ∃N ∈ P (N depending on ϵ) such that |an − L|,
∀n ≥ N

ˆ If {an } has a limit, we say that it is convergent if {an } does not converge, we say
it is divergent.

ˆ If {an } has limit L, we denote this as limn→∞ an

Theorem 3.3
1
Let p ∈ P. Then the sequence np converges to 0.

Proof. Fix ϵ > 0. By the Archimedean property of R, ∃N ∈ P such that N ϵ > 1 =⇒

ϵ > 1/N . Thus,
1 1 1
=⇒ p
≤ ≤ ≤ ϵ, ∀n ≥ N
n n N
1
=⇒ 0 < p < ϵ∀n ≥ N
n
1
=⇒ −0 <ϵ
np
1
Since ϵ > 0 was arbritrary, by definition, limn→∞ np =0

Theorem 3.4
Let {an } be a convergent sequence in R. Then it has a unique limit.

4
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Proof. Suppose {an } has limit L and L′ . Fix ϵ > 0. By definition, ∃N ∈ P such that
|an − L| < ϵ∀n ≥ N . Also, ∃N ′ ∈ P such that |an − L′ | < ϵ∀n ≥ N ′ Now, we write,

|L − L′ | ≤ |an − L′ | + |an − L| < 2ϵ∀n ≥ max N, N ′

Since ϵ was arbitrary, L = L′

Theorem 3.5
Let {an } and {bn } be a sequence in R with limits L and T respectively.

ˆ {can } is convergent and limn→∞ (can ) = cL

ˆ {an + bn } is also convergent and limn→∞ (an + bn ) = L + T

ˆ {an bn } is also convergent and limn→∞ (an bn ) = LT

ˆ Suppose bn ̸= 0 for all n and that T ̸= 0. Then the sequence { abnn } is convergent
and limn→∞ { abnn } TL

Proof. We will only prove part third. As {an } and {bn } converge, they are bounded. So,
∃c1 , c2 > 0 such that |an | ≤ c1 and |bn | ≤ c2 ∀n

an bn − LM = (an bn − an M ) + (an M − LM )
|an bn − LM | ≤ |an ||bn − M | + |M ||an − L|
= c1 |bn − M | + M ′ |an − L|
(
|M | if M ̸= 0
Where M ′ := By definition,
1 if M = 0
ϵ
∃N1 , such that |an − L′ | < ∀n ≥ N1
2M ′
ϵ
∃N2 , such that |bn − M | < ∀n ≥ N2
2c1

From these equations, |an bn − LM | < ϵ for all n ≥ max N1 , N2

Problem (Example)
n2 n2 +1
Consider an = n+1 − n , does {an } converge?

Proof.
n2 n2 + 1
{an } = −
n+1 n
n − (n + 1)(n2 + 1)
3 −n2 − n − 1 −1 − 1/n − 1/n2
= = =
n2 + n 2
n +n 1 + 1/n
By theorem of algebraic combination of convergence of sequences, and since limit of
denominator ̸= 0, {an } converges to the limit 1

5
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Definition 3.6. Let {an } be a sequence in R. We say that {an } is monotone if

ˆ a1 ≤ a2 ≤ · · · (monotonic increasing)

ˆ a1 ≥ a2 ≥ · · · (monotonic decreasing)

Recall that if S ⊂ R such that S ̸= ϕ and bounded above, its LUB is unique. This is
called the supremum of S, denoted by sup S. Similarly, if S is bounded below, it has a
unique GLB and this is called the infimum of S, denoted by inf S

Theorem 3.7 (Monotone Convergence Theorem)

A monotone bounded sequence is convergent.

Proof. We have two cases. We will only show when the sequence is monotone decreasing.
The other part is similar and left as an exercise.
Case 1: {an } is monotone decreasing.
The set S = {an : n ∈ P} is bounded below. Let I be the inf S. Fix ϵ > 0. By definiton,
I + ϵ is not a lower bound. So, ∃N such that aN < I + ϵ

=⇒ I − ϵ < an < I + ϵ∀n ≥ N (Because sequence is monotone)

=⇒ |an − I| < ϵ∀n ≥ N

Since ϵ was arbitrary, we are done!

§4 Infinite Series
P∞
Definition 4.1. Let F be a field. An infinite series is an expression of the form n=1 an ,
where a1 , a2 , · · · ∈ F

PAssociated with a series of the above form, we have a sequence {Sn }, where Sn :=
n
k=1 ak . This is called the nth partial sum.

Definition 4.2. Let ∞

P
n=1 an be a series, an ∈ R∀n. We say that this series converges if
the associated sequence Sn converges. The limit L := limn→∞ Sn is called the sum of
the series.

Example: 1 − 1 + 1 − 1 + · · ·
(
1, if n is odd
Sn =
0, if n is even

In this example, we cannot define the sum because Sn does not converge.

Definition 4.3. We say that a real series ∞

P
n=1 an is a telescopic series if ∃ a real
sequence {bn } such that an = bn − bn+1 ∀n ∈ P

ˆ S1 = a1 = b1 − b2

ˆ S2 = a1 + a2 = b1 − b3

ˆ Sn = b1 − bn+1

6
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 4.4
Let ∞
P
n=1 an be a telescoping series. Let {bn } such that an = bn − bn+1 ∀n. The
series converges iff {bn } converges. In this case, sum∞
n=1 an = b1 − limn→∞ bn

Proof. We will first prove a lemma.

Lemma: Fix k ∈ P. Let {bn } be a sequence that converges to L. Let αn := bn+k ∀n,
then {αn } converges to L.
Sketch of Proof: Given ϵ > 0, pick N to be the number for {bn }. Then the same N
works for {αn } because n + k > n ≥ N , which will deliver the proof.

∞
X
an ⇔ {Sn }
n=1
⇔ {b1 − bn+1 }
⇔ {−(b1 − bn+1 + b1 )}
⇔ {bn+1 }
⇔ {bn }

Where the above denotes that RHS converges ⇔ LHS converges. Hence, its sum is
(b1 − limn→∞ bn )

§4.1 Tests for convergence

Theorem 4.5 (Geometric Series Test)

Series of the form ∞ n−1 , where, a, r ∈ R. Read section 10.8 from Apostol.
P
n=1 ar

Theorem 4.6 (Divergence Test)

Let ∞
P P∞
n=1 an be a real series. If n=1 an exists then the sequence of nth terms is
convergent and limn→∞ an = 0.If

ˆ {an } does not converge, or

ˆ {an } converges but not to 0

We immediately deduce that ∞

P
n=1 an does not converge. Hence the name Diver-
gence Test

Theorem 4.7 (p series test)

The series ∞ 1
P
n=1 np , where p ∈ Q converges ↔ p > 1

7
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 4.8
Let ∞
P P∞
n=1 an and n=1 bn be real converging series. Then,

ˆ Let α ∈ R. Then the series ∞

P P∞
n=1 αan converges and the sum is α n=1 an
P∞ P∞ P∞
ˆ n=1 an + bn converges and the sum is n=1 an + n=1 bn

(a) (b)
Proof. Let Sn and Sn be partial sums of the two series.
Pn (a) (a)
ˆ j=1 αSn . By an earlier theorem, this converges to α limn→∞ Sn
Pn (a) (b) (a) (b)
ˆ j=1 aj + bj = Sn + Sn . By an earlier theorem, {Sn + Sn } converges to
(a) (b)
limn→∞ Sn + limn→∞ Sn

Tests for non-negative series

Theorem 4.9 (Comparision Test)

P∞ P∞
Let n=1 an and n=1 bn be non negative series. Suppose ∃N ∈ P such that
an ≤ bn ∀n ≥ N

ˆ If ∞
P P∞
n=1 bn converges, then so does n=1 an
P∞ P∞
ˆ if n=1 an diverges, then so does n=1 bn

1
Does ∞
P
Example: Suppose an = √
3n2 +7n−1 n=1 an converge?
Hint: Develop an √ 1
intuition that 3n2 +7n−1 is like √13n
P∞
ˆ By p-series and linearity, looks like n=1 an diverges

ˆ Need to find smaller divergent series

Using the above information, note that

3n2 + 7n − 1 ≤ 3n2 + 7n ≤ 4n2

1
for all n ≥ 7. Thus, an ≥ 2n . Since this diverges, we are done by comparision test.
Proof of comparision test second part.

Proof. We will P
first prove a lemma.
Lemma: LetP ∞
P∞
n=1 an be a real series. Let N ∈ N {0, 1}. Then the series n=1 an
converges iff ∞n=N n a converges.

By ourPlemma, can assume WLOG that an ≤ bn ∀n (if not, then consider the
we P
series ∞ a
n=1 n
′ and ∞ ′ ′ ′
n=1 bn where an = aN +n−1 and bn = bN +n−1 )

(a) (b) (a)

Note that {Sn } and {Sn } are monotonic increasing sequences. By assumption, {Sn }
(b)
is not bounded above. Now assume that {Sn } is bounded. Then ∃c > 0 such that
(b) (a) (b)
0 ≤ Sn ≤ c. But then, Sn ≤ Sn ≤ c because an ≤ bn . This is a contradiction.

8
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 4.10 (Ratio Test)

Let ∞ an+1
P
n=1 an be a series with only positive terms. Assume { an } converges. Let L
denote the limit.

ˆ If L < 1, the series converges

ˆ If L > 1, the series diverges

ˆ If L = 1, we cannot draw any conclusions

P∞ n! an+1
Ratio test is effective for a series like n=1 2n where an is hard to analyse, but an is
easy to.

Theorem 4.11
nq
Let a > 1 be a real number. Let q ∈ Q such that q ≥ 0, then limn→∞ an =0

nq nq
Sketch. Consider ∞ that L = 1/a < 1. Thus, ∞
P P
n=1 an and show
q
n=1 an converges, and
by an above theorem, limn→∞ ann = 0

ˆ The sketch can be made more rigorous if q ∈ N

ˆ The general Slogan is that “an grows faster than nq ”

Proof of Ratio Test. We will proceed case wise.

Case 1: L < 1
Fix ϵ > 0 such that L + ϵ < 1 ∃N ∈ P such that,
an+1
L−ϵ< < L + ϵ∀n ≥ N
an

By a simple induction, we can prove that aN +k < aP k

N (L + ϵ) . Now comparing to the
geometric series, and using the fact that L + ϵ < 1, ∞ n=1 an converges.
Case 2: L > 1 (Left as an exercise)
Case 3: L = P1
Notice that ∞ 1 P∞ 1
n=1 n diverges and n=1 n2 converges

Theorem 4.12 (Root Test)

1/n
Let ∞
P
n=1 an be a non negative series. Assume that limn→∞ an exists. Call this
limit R

ˆ If R < 1, the series converges

ˆ If R > 1, the series diverges

ˆ If R = 1, then no conclusions

9
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

§4.2 Order for trying tests

P∞
If the non negative series n=1 an is not a telescopic or a geometric series, useful order
for trying out tests is:

ˆ Divergence Test

ˆ p-series Test

ˆ Ratio/Root Test

ˆ Comparision Test
P∞ n(2+(−1)n )n
Example: n=1 5n
Hint: Note that the Divergence Test, p-series Test and Ratio/Root test does not work.
Hence, use the Comparision Test.

Absolute Convergence
Definition 4.13. GivenPa real series ∞
P
n=1 an , we say this series converges absolutely
if the associated series ∞
n=1 |an | converges

Remark: Every test for non-negative series is also a test for absolute convergence!
Conditional Convergence: Series which converges but not absolutely.

§5 Limits of Function
ˆ What do we mean by limx→0 sin x
x =1

ˆ One way to understand is to take a sequence {xn } such that limn→∞ xn = 0 and
observe that limn→∞ sinxnxn = 1, where we must only consider sequences {xn } such
that xn ̸= 0∀n

ˆ Also, we must get limn→∞ sin xn

xn to exist, and be independent of any {xn } as above.

Definition 5.1. Let I be an open interval and p ∈ I. Let f be a real valued funciton
such that f (x) is defined ∀x ∈ I, except perhaps at x = p. Let A ∈ R, we say that the
limit of f (x) as x approaches p is A, denoted by limx→p f (x) = A, if for any sequence
{xn } ⊂ I − {p} such that limn→∞ xn = p, we have that limn→∞ f (xn ) = A

This is an excellent definition in the sense that theorems whose burden can be passed
onto sequences and series are easily proven. For an example, consider these theo-
rems.

Theorem 5.2
Let I, p, f be as in the above definition. Let A1 and A2 be such that limx→p f (x) = A1
and limx→p f (x) = A2 , then A1 = A2 .

10
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 5.3
Let I be an open interval and p ∈ I. Let f and g be real valued functions such
that f and g are defined ∀x ∈ I, except perhaps at x = p. Let c ∈ R. Suppose that
limx→p f (x) = A and limx→p g(x) = B . Then,

ˆ cf (x) has a limit as x approaches p, and limx→p cf (x) = cA

ˆ f + g has a limit as x approaches p, and limx→p f (x) + g(x) = A + B

ˆ f g has a limit as x approaches p, and limx→p f (x)g(x) = AB

These two theorem are direct corollaries of the corresponding theorems from sequences
and series.
The sequencial definition has the problem that given a f and p, we would have to analyse
every {f (xn )} for every {xn }. This seems HARD!
If we fix ϵ > 0 and want to know for which x, |f (x) − A| < ϵ, instead of looking at xn ∀n
sufficiently large, we consider all x sufficiently close to p, and not equal to p. This allows
us to get rid of sequences {xn } ⊂ I − {p}

Definition 5.4 (ϵ-δ definition). Let I, p, f be as above. Let A ∈ R. We say that the
limit of f (x) as x approaches p is A, if given ϵ > 0, ∃δ > 0 (which depends on ϵ) such
that,
|f (x) − A| < ϵ whenever x ∈ I and 0 < |x − p| < δ

Two standard limits

Theorem 5.5
sin x
limx→p x exists and equals 1

Sketch. Won’t prove; notice that the proof done in high school can be recasted as
demonstration of the sequence definition. The Missing Ingredient is the rigorous proof
of Squeeze Theorem

Theorem 5.6
Let k ∈ P and p ∈ R. Then limx→p xk exists and equals pk

Proof. We proceed by cases.

Case 1: k = 1. Nothing to prove Case 2: k ̸= 1. Fix a sequence {xn } ⊂ R − {p} such
that limn→∞ xn = p. Thus, there exists N1 ∈ P such that |xn − p| < 1. Using the triangle
inequality, we get that
|xn | < 1 + |p|∀n ≥ N1
Fix a sequence {xn } ⊂ R − {p} such that limn→∞ xn = p. Fix ϵ > 0. Thus, ∃N2 ∈ P
such that,
ϵ
|xn − p| < ∀n ≥ N2
k(|p| + 1)k−1

11
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Thus, we get that,

k−1
X
k k
|x − p | ≤ |xn − p| |xn |k−j−1 |p|j
j=0
k−1
ϵ X
≤ (|p| + 1)k−j−1 |p|j
k(|p| + 1)k−1
j=0
k−1
≤ k(|p| + 1) ∀n ≥ max N1 , N2

Because ϵ > 0 was arbitrary and {xn } ∈ R − {p} was also arbitrary, we are done.

Theorem 5.7 (Equivalence of the two definitions)

Let I, p, f be as in our definitions. Suppose that for any {xn } ⊂ I − {p} such that
limn→∞ xn = p, we have limn→∞ f (xn ) = A, then given ϵ > 0, ∃δ > 0 (depends of
ϵ) such that

|f (x) − A| < ϵ whenever x ∈ I and 0 < |x − p| < δ

and vice versa.

Proof. We will prove the vice versa implication. Fix ϵ > 0 and let δ be as given as
above. Let {xn } ⊂ I − {p} such that limn→∞ xn = p. By definition, ∃N1 ∈ P such that,

0 < |x − p| < δ =⇒ |f (x) − A| < ϵ

Because ϵ > 0 was arbitrary, limn→∞ f (xn ) = A. As {xn } was arbitrary, vice versa
follows

§6 04/10/23
When we defined limits, we had,
ˆ an open interval I and p ∈ I

ˆ an R valued function f such that f (x) is defined ∀x ∈ I except perhaps at x = p

Reason for the choice of an open interval I are:
1. It is the closest to the definition in Apostol

2. The limit concept is key to differentiation , I ⊆open R admits generalization to 2

or more variable.
Let us recall the definition of limit. We say that limx→p f (x) = A, iff, given ϵ > 0, ∃δ > 0
(depending on ϵ) such that, |f (x) − A| < ϵ whenever x ∈ I AND 0 < |x − p| < δ
Notice that the boxed words ensure that we can define a limit even if I is not an open
interval.
Now, we will begin our discussion on continuity. Let I be an interval and f : I → R be
a function.
Definition 6.1. Let f and I be as above. Let p ∈ I. We say that f is continuous at p
if, limx→p f (x) = f (p). In general, we say that f is continuous if f is continuous ∀p ∈ I

12
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Definition 6.2 (“ϵ- δ” definition of continuity). Let f and I be as above. Let p ∈ I.

We say that f is continuous at p, if given any ϵ > 0, ∃δ > 0 such that |f (x) − f (p)| < ϵ
whenever x ∈ I AND |x − p| < δ
Note that,
1. Sequential definition of limit (used in proving the theorem on sum and products of
limits) is equivalent to the ϵ- δ definition.
2. The above discussion on open interval extends the notion of limits to all intervals.

Theorem 6.3
Let I be an interval, and p ∈ I and let f, g : I → R. Let c ∈ R. Assume f and g are
continuous at p. Then,

ˆ f + g is continuous at p

ˆ f g is continuous at p

§7 06/10/23
Last time
1. Introduced the concept of continuity
2. Proved the algebra of continuity theorem.

Today
We will take the theorem on algebra of continuity a bit further. Namely, we will add the
following propositions.

Theorem 7.1 (Continuation of the algebra of continuity)

Consider the same hypothesis as in the earlier theorem, we have that,

ˆ cf is continuous at p

ˆ Suppose that g(p) ̸= 0. Then, ∃r > 0, such that, g(x) ̸= 0∀x ∈ (−r+p, r+p)∩I,
and f /g is continuous at p

§7.1 Examples
First note that the previous theorem allows us to VASTLY expand our stock of continuous
functions.
ˆ Any constant function is continuous on R

ˆ The function f (x) = x (which is sometimes denoted by idR ) is continuous on R

(Note that ϵ = δ works)
ˆ We can repeatedly apply the previous theorem on the identity function to get that
Any polynomial is continuous on R

13
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

ˆ (Rational Functions) Let f and g be polynomials. Let I ⊆ R be an interval such

that g(x) ̸= 0 ∀x ∈ I. Then f /g (defined on I) is continuous on I.

ˆ Fix p ∈ R. We will try to show that sin x is continous at p. For that, we will use
the following lemma.

Lemma 7.2
Suppose that f is a function (defined on I) such that |f (a) − f (b)| ≤ c|x − p|
∀a, b ∈ I, then this function is continous.

Now we compute,

| sin(θ + p) − sin p| = | sin θ cos p + cos θ sin p − sin p|

≤ | sin θ|| cos p| + | sin p|| cos θ| ≤ | sin θ| + 2 sin2 θ/2

Now suppose that θ ̸= 0, then by the above lemma,

 
sin θ 2 sin θ/2
sin(θ + p) − sin p ≤ |θ| +
θ |θ|

Now use the ϵ- δ definition with ϵ = 1 to get a known c. Finish it as a homework.

§8 09/10/23
Last time
Vastly expanded our stock of continuous functions

Today
Let S1 , S2 ⊂ R be non empty sets and f : S1 → R, g : S2 → R. We will denote as f (S1 )
the set
f (S1 ) := {f (x) : x ∈ S1 }
Suppose f (S1 ) is a subset of S2 , then we can define a function which is denoted as g ◦ f ,
and it is defined as
g ◦ f := g(f (x))

Theorem 8.1 (Continuity of Composition)

Let I1 , I2 ⊂ R be intervals, let f : I1 → R and g : I2 → R. Suppose that f (I1 ) ⊆ I2
Let p ∈ I1 . Suppose that f is continuous at p and g is continuous at f (p). Then the
function g ◦ f is continuous at p

Proof. Fix an ϵ > 0. Write q := f (p). Then ∃σ > 0 such that |g(y) − g(q)| < ϵ whenever
y ∈ I2 AND |y − q| < σ.
By the continuity of f at p, ∃δ > 0, such that |f (x) − f (p)| < σ whenever x ∈ I1 and
|x − p| < δ.
By our assumption, whenever x ∈ I1 =⇒ f (x) ∈ I2 . From the above statements,
taking y = f (x), we get that |g(f (x)) − g(f (p))| < ϵ whenever x ∈ I1 and |x − p| < δ

14
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 8.2 (Bolzano Theorem)

Let a < b ∈ R. Let f : a, b → R be continuous. If f (a) and f (b) have opposite signs,
then ∃c ∈ (a, b) such that f (c) = 0

Proof. We will begin by first proving a lemma.

Lemma 8.3 (Sign Preserving Property)

Let I be an interval, g : I → R and let c ∈ I. Suppose that g is continuous at
c. Suppose that g(c) ̸= 0. Then ∃r > 0 such that g(x) has the same sign as g(c),
∀x ∈ (c − r, c + r) ∩ I.

Proof of Lemma. We will only prove for the case g(c) < 0. The other case is left as an
exercise. By continuity, ∃r > 0 such that |g(x) − g(c)| < |g(c)|/2 =⇒ g(x) − g(c) <
−g(c)/2 (because g(c) < 0). Rearranging, we get that g(x) < g(c)/2 < 0.

Having done this, define,

S := {x ∈ [ a, b] : f (x) ≤ 0}
We immediately notice the following:
ˆ S ̸= ϕ (because a ∈ S)
ˆ S is bounded
Hence, by the LUB property, c := sup S exists.

§9 10/10/2023
Last time
1. Continuity of composition of functions
2. Began proving Bolzano’s Theorem

Today
Continuation of Yesterday’s proof. Assume that f (c) ̸= 0. Then f (c) < 0 or f (c) > 0.
We will get a contradiction in either case.
f (c) > 0 =⇒ ∃r > 0 such that,
f (x) > 0 ∀x ∈ (c − r, c + r) ∩ [a, b]
Note that a ∈/ (c − r, c + r) =⇒ a < c − r < c, i.e, c − r is not even an upper bound of
S. Thus, ∃x′ : c − r < x′ ≤ c such that f (x′ ) ≤ 0. Contradiction!
Do the f (c) < 0 case as a homework.

Theorem 9.1
Let I be an interval, and let f : I → R be continuous. Let x1 < x2 ∈ I. Suppose that
f (x1 ) ̸= f (x2 ). Then, for any k ∈ R lying between f (x1 ) and f (x2 ), ∃c ∈ (x1 , x2 )
such that f (c) = k

15
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Proof. Let g(x) = f (x)|[x1 ,x2 ] − k : [x1 , x2 ] → R. Clearly, this is continuous. Also
notice that max [f (x1 ), f (x2 )] > k and min [f (x1 ), f (x2 )] < k. So, g(x1 ) and g(x2 ) have
opposite signs. Thus, by using Bolzano’s theorem, ∃c ∈ (x1 , x2 ) such that g(c) = 0. But,
this gives that f (c) = g(c) + k = k

§9.1 Interlude
A and B are non empty sets. We define the Cartesian Product as,

A × B := {(a, b) : a ∈ A and b ∈ B}

Notice that the Cartesian Product is a collection, but is this a set?

Answer 1
Take as an axiom that A × B is a set.

Answer 2
A × B can be constructed as a set thanks to something called the Pair Set Axiom

Definition 9.2. Let A and B be two non empty sets. A relation between two sets, say
→, is a subset S→ ⊆ A × B. We say that a → b ⇔ (a, b) ∈ S→

§10 11/10/23
Last time
ˆ Bolzano Theorem

ˆ Cartesian Product

Today

Example 10.1
> is a relation between R and itself. y > x ⇔ y − x ∈ R+ . More formally,

S> = {(y, x) ∈ R × R : y − x ∈ R+ }

Working inductively using the above argument, given non empty sets, A1 , · · · , An , we
can define

A1 × A2 × · · · × An := {(a1 , · · · , an ) : aj ∈ Aj , j = 1, · · · , n}

Definition 10.2. Let A and B be non empty sets. A function f : A → B is a relation

Γf ⊆ A × B such that,

1. For each a ∈ A, ∃(a, b) ∈ Γf (A is the domain of f )

2. For each a ∈ A, ∃ a unique b ∈ B such that (a, b) ∈ Γf (b := f (a))

16
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 10.3
Let f : [−1, 1] → [−1, 1]. Suppose f is continuous. Then, ∃c ∈ [−1, 1] such that,
f (c) = c

Proof. We proceed by cases.

Case 1: f (−1) = −1 or f (1) = 1. Trivial. Nothing to prove.
Case 2: f (−1) ̸= −1 and f (1) ̸= 1. Define g(x) = f (x) − x∀x ∈ [−1, 1]. Note that g
is continuous. Further,

g(−1) = f (−1) − (−1) >0

g(1) = f (1) − 1 <0

Now, by Bolzano’s theorem, ∃c ∈ (−1, 1) such that g(c) = f (c) − c = 0 =⇒ f (c) = c

A Short note on the relevance of the above theorem

We proved the n = 1 case of the Brower’s Fixed Point Theorem. To understand the
theorem, consider the set,
v
u n 2
uX
n t
Bn := {(x1 , · · · , xn ) ∈ R : xj ≤ 1}
j=1

Theorem 10.4
Any continuous function from Bn to itself has a fixed point i.e ∃ a point (x1 , · · · , xn )
such that, f (x1 , · · · , xn ) = (x1 , · · · , xn )

There are many practical applications of this theorem, one of them being PAGE RANK
of a search engine.

§11 13/10/20
Last time
ˆ More about relations and functions

ˆ Brouwer’s Fixed Point Theorem (n = 1 in syllabus)

Today
Let S ̸= ϕ be a set. Let f be a function f : S → R such that

{f (x) ∈ R : x ∈ S} (range of f )

is bounded above and below.

ˆ Such an f is called a bounded function

ˆ inf(range) and sup(range) exists.

ˆ Abbreviate terms as inf f and sup f

17
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 11.1
Let a < b ∈ R. Let f : [a, b] → R be continuous. Then f is bounded

Proof. Read Theorem 3.11 from Apostol

Corollary 11.2
Let [a, b] and f be as above. Then,

1. inf f and sup f exists

2. ∃c, d ∈ [a, b] such that, f (c) = inf f and f (d) = sup f . Thus, f has a point of
absolute maximum and a point of absolute minimum.

Proof. For the first part, by the previous theorem, range(f ) is bounded and non empty.
the first claim follows from the GLB property and 2nd from the LUB property.
For the second part, it suffices to show that ∃d such that, f (d) = sup f . Suppose not,
and let M := sup f . Then M − f (x) > 0∀x ∈ [a, b]. By our theorem on continuity of
algebraic combinations,
1
g(x) := (Continuous on [a, b])
M − f (x)

By the last theorem, ∃c > 0 such that,

1
0< ≤ C ∀x ∈ [a, b]
M − f (x)
=⇒ M − f (x) ≥ 1/c
=⇒ f (x) ≤ M − 1/c ∀x ∈ [a, b]

But (M − 1/c) is not an upper bound of range(f ), which contradixts the last line.

§12 16/10/23
Last time
1. If f : [a, b] →continuous R, then f is bounded

2. Corollary: Existance of points of absolute maximum and minimum for above class
of functions

Today
Definition 12.1. Let S be a non empty set and f : S → R. A point p ∈ S is called a
point of absolute maximum (respectively, absolute minimum) if, f (p) ≥ f (x)∀x ∈ S
(respectively, f (p) ≤ f (x)∀x ∈ S)

18
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

§13 Differentiability
Definition 13.1. Let I ⊆ R be an open interval and x ∈ I. Let f : I → R. We say that
f is differentiable at x if the following limit exists.

f (x + h) − f (x)
lim := f ′ (x)
x→0 h
ˆ If f is differentiable at each x ∈ I, we say that f is differentiable

ˆ In the above case, f induces the function f ′ : I → R

Theorem 13.2
Let I be an open interval, a ∈ I and f : I → R. If f is differentiable at a, then it is
continuous at a

Proof. Write
f (x) − f (a)
f (x) − f (a) = (x − a) ∀x ∈ I − {a}
x−a
f (x) − f (a)
Since lim exists, by our theorem on limit of products, this implies,
x→a x−a

lim (f (x) − f (a)) = 0

x→a
=⇒ lim f (x) = f (a)
x→a

Example 13.3
Show that f (x) = cos x is differentiable

Solution.
cos(x + h) − cos x cos x cos h − sin x sin h − cos x
=
h h  
− cos x(1 − cos h) sin h
= − sin x
h h
 2   
h sin (h/2) sin h
= −2 cos x − sin x
4 (h/2)2 h

Finish this argument as an homework.

§14 18/10/23
Last time
Introduction to differentiation.

19
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Today

Theorem 14.1
Let I ⊆ R be an open interval. Let a be a point in I and let f, g : I → R. Let
c1 , c2 ∈ R. Assume that f and g are differentiable at a. Then

ˆ c1 f + c2 g is differentiable at a and (c1 f + c2 g)′ (a) = c1 f ′ (a) + c2 g ′ (a)

ˆ f g is differentiable at a and (f g)′ (a) = f ′ (a)g(a) + f (a)g ′ (a)

ˆ Assuming that g(a) ̸= 0 Then ∃r > 0 such that g(x) ̸= 0 ∀ (a − r, a + r).

Moreover, f /g is differentiable at a and
 ′
f f ′ (a)g(a) − f (a)g ′ (a)
=
g g(a)2

Proof of part 2. We compute,

f g(a + h) − f g(a) f (a + h)(g(a + h) − g(a)) + g(a)(f (a + h) − f (a))

=
h h
g(a + h) − g(a) f (a + h) − f (a)
= f (a + h) + g(a)
h h
Also, we know that, limh→0 f (a+h) = f (a). Now, by using limit of algebraic combinations,

f (a + h)g(a + h) − f (a)g(a)
lim = f (a)g ′ (a) + f (a)g ′ (a)
h→0 h

Theorem 14.2 (Chain Rule)

Let I1 , I2 ⊆ R be open intervals. Let f : I1 → R and g : I2 → R. Let f (I1 ) ⊆ I2
and let F := g ◦ f . Let a ∈ I1 and suppose f is differentiable at a and let g is
differentiable at f (a). Then, F is differentiable at a and

F ′ (a) = g ′ (f (a))f ′ (a)

How ”Not” to prove the chain rule

Here’s a faulty argument.

1. Write:
F (x) − F (a) g(f (x)) − g(f (a)) f (x) − f (a)
= ·
x−a f (x) − f (a) x−a

2. We could justify writing y := f (x) and stating

lim (y − f (a)) = lim (f (x) − f (a)) = 0

y→f (a) x→a

The first limit is self evident and the second one follows by continuity of f

20
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

3. But EQUATING:
g(y) − g(f (a)) g(f (x)) − g(f (a))
=
y − f (a) f (x) − f (a)
is a problen because there could, for each r > 0, ∃xr ∈ (a − r, a + r) ∩ I1 , xr ̸= a
such that f (xr ) = f (a).

Self-tudy
Work on all ”applied” problems in 4.12

§15 20/10/23
Last time
1. Differentiation of algebraic combinations

2. Chain rule

Today
Definition 15.1. Let S be a non empty set and let f : S → R. A point c ∈ S is called

i) a point of local or relative maximum if ∃r > 0 such that f (c) ≥ f (x)∀x ∈

S ∩ (c − r, c + r)

ii) a point of local or relative maximum if ∃r > 0 such that f (c) ≤ f (x)∀x ∈
S ∩ (c − r, c + r)

Theorem 15.2
Let I ⊆ R be an open interval and let f : I → R. Let f be differentiable. If c ∈ I is
a point of local maximum or local minimum, then f ′ (c) = 0

Proof. Let c is a point of local extremum. then, ∃R > 0 such that ,

f (c) ≥ 0 OR
f (c) ≤ 0

Suppose f ′ (c) ̸= 0. Then, it suffices to assume f ′ (c) > 0 (WHY?). Define Df : I − {c} →
R, such that,
f (x) − f (c)
Df (x) :=
x−c
By assumption, limx→c Df (x) exists (= f ′ (c)). Now, by the sign preserving property,
∃r ∈ (0, R) such that,

(c − r, c + r) ⊆ I AND Df (x) > 0∀x : 0 < |x − c| < r

21
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

§16 23/10/2023
§16.1 Last time
Point of local maximum or minimum algorithm for determining absolute max and min
values

§16.2 Today

Theorem 16.1 (Rolle’s Theorem)

Let a < b ∈ R and f : [ a, b] → R be continuous. Assume f is differentiable on (a, b)
and f (a) = f (b). Then ∃c ∈ (a, b) such that f ′ (c) = 0

Proof. Assume f ′ (x) ̸= 0∀x ∈ (a, b). By assumption, f has a point of absolute maximum
and also a point of absolute minimum. By the previous theorem, none of the latter are
in (a, b). Thus,
min{f (a), f (b)} ≤ f (x) ≤ max{f (a), f (b)}∀x ∈ [ a, b]
This implies that f is a constant, thus, f ′ (x) = 0∀x ∈ (a, b), a contradiction.

Theorem 16.2 (Lagrange’s Mean Value Theorem)

Let a < b ∈ R and f : [ a, b] → R and f is continuous. Further assume that f is
differentiable on (a, b). Then ∃c ∈ c(a, b) such that

f (b) − f (a)
f ′ (c) =
b−a

Sketch of Proof. Apply Rolle’s Theorem to h : [ a, b] → R where

h(x) := (b − a)f (x) − x(f (b) − f (a))

Corollary 16.3
Let a < b ∈ R and f : [ a, b] → R is continuous. Suppose f is differentiable on (a, b),
then,

ˆ If f ′ (x) > 0∀x ∈ (a, b), then f is strictly increasing

ˆ If f ′ (x) < 0∀x ∈ (a, b), then f is strictly decreasing

ˆ If f ′ (x) = 0∀x ∈ (a, b), then f is constant

Proof. We will only do part (i). Let x < y ∈ [ a, b] . Apply LMVT to f (first check all
conditions to apply LMVT). Thus, ∃c ∈ (x, y) such that,
f (y) − f (x) = f ′ (c)(y − x) > 0
because f ′ (x) > 0 by hypothesis. Thus, f (y) > f (x). Since, x < y are arbritrary, (i)
follows.

22
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Definition 16.4. Let S ⊆ R be a non empty set. If f : S → R is one one, then

∃f −1 : f (S) → S such that , f (f −1 (y)) = y, y ∈ f (S)

GENERAL EXAMPLES
ˆ With S as above, if f is either strictly increasing or strictly decreasing, f is one-one

Theorem 16.5
Let I ⊆ R be an interval and let f be constant. If f is one one, then f is either
strictly increasing or strictly decreasing.

§17 25/10/2023
Last time
ˆ 2 mean value theorems (Rolle’s and LMVT)
ˆ Invertible functions

Today
We begin with an interesting question. If f is invertible, is ontinuity of f inherited
by f −1 . The answer is that it depends on the domain of f . Here is the appropriate
theorem.

Theorem 17.1
Let I ⊆ R be an interval and f : I → R be an invertible funciton. If f is continuous,
then so is f −1

The proof of the above theorem was not done in class, but it’s a nice exercise to try
out (Hint: Try proving one one first) Now, we are ready to talk about derivatives of
inverses.

Theorem 17.2
Let I ⊆ R be an open interval and f : I → R be an invertible function. Let a ∈ I.
Assume f is continuous and f is differentiable at a. If f ′ (a) ̸= 0, then,

ˆ f (I) is an open interval

ˆ f −1 is differentiable at f (a) and,

1
(f −1 )′ [ f (a)]
f ′ (a)

A short sketch/hint. Consider a sequence {yn } ⊂ J − {f (a)}. We need to analyse if,

f −1 (yn ) − f −1 (f (a))
has a limit
yn − f (a)
Now set xn := f −1 (yn ), and try it out as an homework.

23
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Example 17.3
Find the derivate of arcsin x (Note that in general arcsin x is defined only if the sin
function is restricted to [2n − 1/2π, 2n + 1/2π])

solution. Note that if we define,

g = f −1 where f (θ) = sin θ∀θ ∈ [−π/2, π/2]

Let x ∈ (−1, 1). Any such x is of the form x = sin θ for a unique θ ∈ (−π/2, π/2). Thus,
we can infer
1
(f −1 (sin θ) = )
cos θ
Now complete the argument and prove that
1
(f −1 (x) = √ )
1 − x2

§18 27/11/23
Last time
ˆ Continuity and Differentiability of inverse in context of function defined on intervals

ˆ Found the derivative of the arcsin x function

Today
Definition 18.1. Let I ⊆ R be an open interval and f : I → R. If f is differentiable
on I, write f (1) := f ′ and f (0) := f . We say that f hasderivatives of order n, n ∈ P, if
′
f (n−1) exists and is differentiable where f (n) := f (n−1) for 1 ≤ k ≤ n

Integration
Rb R
(Discussion) We want to make precise the meaning of a f (x)dx (from which f (x)dx
follows) For f : [a, b] → R such that f ≥ 0, want,
Z b
f (x)dx = Area {(x, y) ∈ R2 : 0 ≤ y ≤ f (x), x ∈ [a, b]}
 
a

Definition 18.2. Let a < b ∈ R

ˆ A partition of [a, b] is a finite set of real number x0 , · · · , xn such that a = x0 <

x1 < x2 < · · · < xn = b

ˆ A function on [a, b] is called a step function if ∃ a partition a = x0 < x1 < · · · <

xn = b and ∃ c1 , c2 , · · · , cn ∈ R such that s(x) = cj if x ∈ (xj−1 , xj )

24
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 18.3
Let a < b ∈ R and let s, t : [a, b] be step functions.

ˆ cS is a step function for any c ∈ R

ˆ s + t is a step function for any c ∈ R

§19 30/10/23
Last time
Rb
ˆ Motivation for the integral a f (x)dx (Specifically, if f ≥ 0, integral is the area of
the function)
ˆ Defined step functions

Today
Definition 19.1. Let a < b and s : [a.b] → R be a step function. Let a = x0 < x1 <
· · · < xn = b and c1 , · · · , cn ∈ R be the data defining S. Then,
Z b n
X
s(x)dx = cj (xj − xj−1 )
a j=1

Remarks: A partition determining S is not unique!

Definition 19.2. Let a < b ∈ R. Let f : [a, b] → R be a bounded function. Define
(Z )
b
Sf+ := s(x)dx : s is a step function and s ≥ f
a
(Z )
b
Sf− := s(x)dx : s is a step function and s ≤ f
a

ˆ Sf+ ̸= ϕ (As f is bounded, ∃ M > 0 such that −M ≤ f (x) ≤ M ∀x, s+ := M ≥ f ,

s− := −M ≤ f )
ˆ Sf+ is bounded below (CHECK: −M (b − a) is a lower bound) and similarly Sf− is
bounded above
ˆ I¯ := inf Sf+ (Reimann upper integral)

ˆ I := sup Sf− (Reimann lower integral)

Definition 19.3. Let a < b ∈ R and f : [a, b] → R be a bounded function. We say that
¯ ) = I(f ). The common value is
f is Reimann Integrable denoted f ∈ R [a, b] iff I(f
Z b
called the integral of f , denoted f (x)dx
a

§20 03/11/23
Last time
Defined the Reimann Integral

25
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Today

Theorem 20.1
Z b
Let a < b ∈ R and s : [a, b] ∈ R be a step function. Then s(x)dx is independent
a
of partitions determining S

Sketch. Step 1: Special case where P ⊊ P ∗ such that P ∗ −P = {c}; P ∗ and P determine
partition.
P : a = x0 < x1 < · · · < xn = b
and ∃ unique j ∗ such that 1 ≤ j ∗ ≤ n such that, c ∈ (xj ∗ −1 , xj ∗ ). If

∗
xj , 0 ≤ j ≤ j − 1

∗
P : a = y0 < y1 < · · · < yn+1 = b =⇒ yj = c, j = j ∗

xj−1 , j ≥ j ∗ − 1


Now compare,
Z b Z b
(P) s(x)dx and (P ∗ ) s(x)dx
a a
Step 2: General case. Given determining partitional, P1 ̸= P2 , define P ∗ = P1 ∪ P2 .
Thus, P ∗ ⊋ P1 , P2 . Now, use induction + Step 1 to show that,
Z b Z b Z b
∗
(P1 ) s(x)dx = (P ) s(x)dx = (P2 ) s(x)dx
a a a

Theorem 20.2
Let a < b ∈ R, s, t : [a, b] → R be step functions. Let α ∈ R and c ∈ (a, b)
Z b Z b
ˆ (Homogenity) αs(x)dx = α s(x)dx
a a
Z b Z b Z b
ˆ (Additivity) (s + t)(x)dx = s(x)dx + t(x)dx
a a a
Z b Z b
ˆ (Comparision Theorem) If s ≥ t, then s(x)dx ≥ t(x)dx
a a
Z b Z c Z c
ˆ (Additivity wrt interval) s(x)dx = s(x)dx + s(x)dx
a a b
Z b Z b+α
ˆ (Translation invariance) s(x)dx = s(x − α)dx
a a+α

Proof. We will only proof third part. Notice that,

Z b Z b Z b
s(x)dx − t(x)dx = (s − t)(x)dx ≥ 0
a a a

The inequalities follow by the part one and two of the theorem.

26
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

§21 06/11/23
Last time
Properties of integral of step functions.

Today
First, notice that, the value of the integral of the step function given by the general formula
is the same as the reimann integral. This fact is a homework assignment.

Theorem 21.1
Let a < b ∈ R and let f, g : [a, b] → R such that f, g ∈ R [a, b]. Let c ∈ (a, b) and
α ∈ R. Then,

ˆ The first three properties of the theorem on 3/11 hold. (with f + g, αf ∈

R [a, b])

ˆ (Additivity wrt interval) f |[a,c] ∈ R [a, c], f |[c,b] ∈ R [c, b] and,

Z b Z c Z b
f (x)dx = f (x)dx + f (x)dx
a a c

ˆ (Translational invariance) f (x − α) ∈ R [a + α, b + α] and

Z b Z b+α
f (x)dx = f (x − α)dx
a a+α

Proofs are lengthy and omitted. Now, proceeding further, we want to ask a ques-
tion.

Example 21.2 (Question)

Other than step functions, is there some other class of functions that are integrable?

Answer. This question needs some steps.

Step 1:
Definition 21.3. Let I ⊆ R be an interval and f : I → R. We say that f is
uniformly continuous if given ϵ > 0, ∃ δ > 0 depending ONLY on ϵ such that,

|f (x) − f (y)| < ϵ∀x, y ∈ I AND such that |x − y| < δ

(In general, if f is continuous at p ∈ I, δ depends on both ϵ and p)

Step 2: Examples

ˆ The sine function is uniformly continuous. Recall that,

| sin(θ + x) − sin x| ≤ | sin θ| + 2 sin2 θ/2

Note that the LHS doesn’t have any term involving x, which is enough for uniform
continuity (complete the proof)

27
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

ˆ f (x) = xn is not uniformly continuous on R if n ∈ N − {0, 1}

ˆ f (x) = x is uniformly continuous (given ϵ > 0, δ = ϵ works)

§22 7/11/23
Last time
Extended properties of integral of step functions to general Reimann Integrable functions.
Intro to uniform continuity.

Today
The general question asked was to find a class of Reimann Integrable functions.

Theorem 22.1
Let a < b ∈ R. If f : [a, b] → R is continuous, then it is uniformly continuous.

This is by itself a deep result. We will not prove this, but it is useful to know that
the proof re-uses ideas for ”boundedness theorem”. The following theorem answers our
question.

Theorem 22.2
Let a < b ∈ R. If f : [a, b] → R is continous, then f ∈ R [a, b]

Homework
ˆ READ statement of “small span theorem” (Theorem 3.13 in Apostol)

ˆ Study proof of the above theorem (Theorem 3.14 in Apostol)

Now, we will introduce some conventions. Let a < b ∈ R and let f : [a, b] → R be
bounded.
Ra Rb
ˆ If f ∈ R [a, b], set b f (x)dx := − a f (x)dx
Rc
ˆ If c ∈ [a, b], set c f (x)dx = 0 (This can be justified)

§23 Fundamental Theorem of Calculus

Theorem 23.1 (1st Fundamental Theorem of Calculus)

Let a < b ∈ R and let f : [a, b] → R. Let c ∈ [a, b]. Define
Z x
F (x) := f (t)dt∀x ∈ [a, b]
c

Then F is differentiable at each x ∈ (a, b) at which f is continuous and F ′ (x) = f (x)

28
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Proof. Before proving the theorem, check that if f be defined as above and c1 , c2 , c3 ∈ [a, b]
(in no particular order). Then,
Z c3 Z c2 Z c3
f (x)dx = f (x)dx + f (x)dx
c1 c1 c2

Now, let x ∈ (a, b) at which f is continous. Let |h| be so small that x + h ∈ [a, b]. Then,
Z x+h Z c 
F (x + h) − F (x) 1
− f (x) = f (t)dt − f (t)dt − f (x)
h h c c
1 x+h
Z
= f (t)dt − f (x)
h x
1 x+h
Z
= (f (t) − f (x))dt
h x
Z max(x,x+h)
1
≤ |f (t) − f (x)|dt (Comparision Theorem)
|h| min(x,x+h)
Fix ϵ > 0. By continuity at x, ∃δ > 0 such that, |f (y) − f (x)| < δ whenever y ∈ [a, b] and
|x − y| < δ. If t ∈ [min(x, x + h), max(x, x + h)] and |h| < δ, then we have |t − x| < δ
and so, |f (x) − f (y)| < ϵ. Now by the above chain of steps,
Z max(x,x+h)
F (x + h) − F (x) 1
− f (x) ≤ ϵdt
h |h| min(x,x+h)
∀h : 0 < |h| < δ
Since ϵ > 0 was arbritrary,
F (x + h) − F (x)
lim = f (x)
h→0 h

Definition 23.2. Let I ⊆ R be an open interval and let f : I → R. A function P : I → R

is called a primitive (or anti derivative) if P is differentiable, and P ′ (x) = f (x)∀x ∈ I

Theorem 23.3 (2nd Fundamental Theorem of Calculus)

Let I ⊆ R be an open interval and f : I → R is continuous. Then f admits a
primitive. Let P be a primitive and c ∈ I. Then,
Z x
P (x) = P (c) + f (t)dt∀x ∈ I
c

In particular, if a < b ∈ I,
Z b
f (x)dx = P (b) − P (a)
a

Remarks
ˆ Let f and I be as in the definition of primitve and let f admit a primitive. If P1
and P2 are two primitives, then ∃ a constant c such that, P1 = P2 + c (follows from
the corollary to the LMVT)

29
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

ˆ Let I be as above and f : I → R is continuous. Let a < b ∈ I. If we can guess a

primitive for f then we can easily calculate the integral by the above theorem

Corollary 23.4 (Integration by parts)

Let I ⊆ R be open. Let f, g : I → R be differentiable functions such that f ′ and g ′
are continuous. Let c ∈ I. Then,
Z x Z x
′
f (x)g(x) = f (c)g(c) + f (t)g(t)dt + f (t)g ′ (t)dt
c c

Let G : I → R be continuous. Let PG be a primitive of G. Let a < b ∈ I

Z b b Z b
f (x)G(x)dx = f PG − f ′ (x)PG (x)dx
a x=a a

Proof. By hypothesees, f ′ g +f g ′ is continuous on I. By the product rule, f g is a primitive

of (f ′ g + f g ′ ). The assertion follows from the second FTC.

§24 10/11/23
Last time
Finished proof of the 1st FTC. Stated 2nd FTC. Integration by parts

Today
Leibnizian notation: Let I ⊆ R be a non empty interval and f : I → R is continuous.
The 2nd FTC gives existence of primitive of f , which is non unique.
Z
1. We denote by f (x)dx := a function P , UNDETERMINED up to a constant,
such that, P ′ = f on I

2. If we can guess a specific P as in (a), we write,

Z
f (x)dx = P (x) + c

Logarithms
What is a logarithm?
Answer: Any function f : (0, ∞) → (−∞, ∞) such that,

f (xy) = f (x) + f (y)∀x, y ∈ (0, ∞)

Definition 24.1. Let x ∈ (0, ∞). The natural logarithm (denoted by log) of x is defined
as, Z x
1
log(x) := dt
1 x

30
Shreeansh Hota — 15 November 2023 UMA101: Analysis and Linear Algebra

Theorem 24.2 (Properties of logarithm)

Let log be defined as above. Then,

ˆ log(xy) = log x + log y ∀x, y > 0

ˆ log is strictly increasing

ˆ Range of log = R (tricky to prove rigorously)

Proof. We will only prove first one. Fix x, y > 0. Then,

Z xy
1
log(xy) = dt
1 t
Z x Z xy
1 1
= dt + dt
1 t x t
Z y
1
= log(x) + x dt
1 xt
= log(x) + log(y)

31