Week 1: Real numbers and their bounded subsets

August 30, 2022

Lecture 1
Sets
A set is a collection of objects called elements or members of that set.
The empty set (denoted ∅) is the set with no elements.

Notations:

ˆ a ∈ S means that ”a is an element in S.”

ˆ a∈
/ S means that ”a is not an element in S.”

ˆ ∀ means ”for all.”

ˆ := means ”define.”

ˆ ∃ means ”there exists.”

ˆ ∃! means ”there exists a unique.”

ˆ =⇒ means ”implies.”

ˆ ⇐⇒ means ”if and only if.”

Definition 0.1 (Set Relations). We want to relate different sets, and thus we get the following
notation/definitions:

1. A set A is a subset of B, A ⊂ B, if every element of A is in B. Given A ⊂ B, if

a ∈ A =⇒ a ∈ B.

2. Two sets A and B are equal, A = B, if A ⊂ B and B ⊂ A.

3. A set A is a proper subset of B, A ⊊ B if A ⊂ B and A ̸= B.

1
 One way we can describe a set is using ”set building notation”. We write
{x ∈ A | P (x)} or {x | P (x)}
to mean ”all x ∈ A that satisfies property P (x)”. One example of this would be {x | x is an
even number}.
There are a few key sets that we will use throughout this class:
1. The set of natural numbers: N = {1, 2, 3, 4, . . . }.
2. The set of integers: Z = {0, 1, −1, 2, −2, 3, −3, . . . }.
3. The set of rational numbers: Q = { m
n | m, n ∈ Z and n ̸= 0}.

4. The set of real numbers: R.

It follows that
N ⊂ Z ⊂ Q ⊂ R.

1. The union of A and B is the set A ∪ B = {x | x ∈ A or x ∈ B}.

2. The intersection of A and B is the set A ∩ B = {x | x ∈ A and x ∈ B}.
3. The set difference of A and B is the set A \ B = {x ∈ A | x ∈
/ B}.
4. The complement of A is the set Ac = {x | x ∈
/ A}.
5. A and B are disjoint if A ∩ B = ∅.

Theorem 0.1 (De Morgan’s Laws). If A, B, C are sets then

1. (B ∪ C)c = B c ∩ C c ,
2. (B ∩ C)c = B c ∪ C c ,
3. A \ (B ∪ C) = (A \ B) ∩ (A \ C),
4. and A \ (B ∩ C) = (A \ B) ∪ (A \ C).
We will prove the first statement to give an example of how such a proof would go, but the
rest will be left to you.
Proof : Let B, C be sets. We must prove that
(B ∪ C)c ⊂ B c ∩ C c and B c ∩ C c ⊂ (B ∪ C)c .
If x ∈ (B ∪ C)c =⇒ x ∈ / B ∪ C =⇒ x ∈ / C. Hence, x ∈ B c and x ∈ C c =⇒
/ B and x ∈
c c c
x ∈ B ∩ C . Thus, (B ∪ C) ⊂ B ∩ C .c c

If x ∈ B c ∩ C c then x ∈ B c and x ∈ C c =⇒ x ∈/ B and x ∈

/ C. Hence, x ∈
/ B ∪ C =⇒ x ∈
c c c
(B ∪ C) . Thus, B ∩ C ⊂ (B ∪ C) . c

2
Lecture 2
Mathematical Induction
This concept of induction was invented by Pascal in 1665.

Theorem 0.2 (Induction). Let P (n) be a statement depending on n ∈ N. Assume that

1. (Base case) P (1) is true and

2. (Inductive step) if P (m) is true then P (m + 1) is true.

Then, P (n) is true for all n ∈ N.

Let’s see an example of induction in action.

Exercise 0.3. For all c ̸= 1 in the real numbers, and for all n ∈ N,

1 − cn+1
1 + c + c2 + · · · + cn = .
1−c
Proof : We will prove this by induction. First, we prove the base case (n = 1). The left
2 (1−c)(1+c)
hand side of the equation is 1 + c for n = 1. The right hand side is 1−c
1−c = 1−c = 1 + c.
Hence, the base case has been shown.
Assume that the equation is true for k ∈ N, in other words

1 − ck+1
1 + c + c2 + · · · + ck = .
1−c

Thus,

=⇒ 1 + c + c2 + · · · + ck + ck+1 = (1 + c + c2 + · · · + ck ) + ck+1
1 − ck+1
= + ck+1
1−c
1 − ck+1 + ck+1 (1 − c)
=
(1 − c)
1 − c(k+1)+1
= .
1−c
Therefore, our proof is complete.
Let’s do another example:

Exercise 0.4. For all c ≥ −1, (1 + c)n ≥ 1 + nc for all n ∈ N.

3
 Proof : We prove this through induction. In the base case, we have: (1 + c)1 = 1 + 1 · c. For
the inductive step, suppose that
(1 + c)m ≥ 1 + mc.
Then,

(1 + c)m+1 = (1 + c)m · (1 + c).

By assumption,

≥ (1 + mc) · (1 + c)
= 1 + (m + 1)c + mc2
≥ 1 + (m + 1)c.

By induction, our proof is complete.

Absolute Value
Definition 0.2. If x ∈ R we define
(
x, x≥0
|x| := .
−x, x≤0

Theorem 0.5. We can prove a bunch of theorems about the absolute value function that we
usually take for granted:

1. |x| ≥ 0 and |x| = 0 ⇐⇒ x = 0.

2. ∀x ∈ R, | − x| = |x|.

3. ∀x, y ∈ R, |xy| = |x||y|.

4. |x2 | = x2 = |x|2 .

5. If x, y ∈ R, then |x| ≤ y ⇐⇒ −y ≤ x ≤ y.

6. ∀x ∈ R, x ≤ |x|.

7. If x, y ∈ R, then |x + y| ≤ |x| + |y|

Proof :

1. If x ≥ 0 then |x| = x ≥ 0. If x ≤ 0, then −x ≥ 0 =⇒ |x| = −x ≥ 0. Thus, |x| ≥ 0.

Now suppose x = 0. Then, |x| = x = 0. For the other direction, suppose |x| = 0. Then, if
x ≥ 0 =⇒ x = |x| = 0. If x ≤ 0, then −x = |x| = 0. Therefore, x = 0 ⇐⇒ |x| = 0.

4
 2. If x ≥ 0 then −x ≤ 0. Thus, |x| = x = −(−x) = | − x|. If x ≤ 0 then −x ≥ 0 and thus
| − x| = | − (−x)| = |x|.

3. If x ≥ 0 and y ≥ 0, then xy ≥ 0 and |xy| = xy = |x||y|. If x ≤ 0 and y ≤ 0, then

xy ≤ 0 =⇒ |xy| = −xy = (−x)y = |x||y|.

4. Take x = y in 3). Then, |x2 | = |x|2 . Since x2 ≥ 0, it follows that |x2 | = x2 .

5. Suppose |x| ≤ y. If x ≥ 0, then −y ≤ 0 ≤ x = |x| ≤ y. Therefore, −y ≤ x ≤ y. If x ≤ 0,

then −x ≥ 0 and | − x| ≤ y. Hence, −y ≤ −x ≤ y =⇒ −y ≤ x ≤ y.

6. Take y = |x| in 5).

Lecture 3

Completeness Axioms
Definition 0.3 (Bounded Above/Below). Let S be a subset of R. Then,

1. If there exists b ∈ R such that x ≤ b for all x ∈ S, then S is bounded above and b is an
upper bound of S.

2. If ∃c ∈ R such that x ≥ c for all x ∈ S, then S is bounded below and c is a lower bound of
S.

Definition 0.4. We say that b0 is the least upper bound, or the supremum of E if

1. b0 is an upper bound for E and

2. if b is an upper bound for E then b0 ≤ b.

We denote this as b0 = sup E.

Similarly, we say that c0 is the greatest lower bound, or the infinimum of E if

1. c0 is a lower bound for E and

2. if c is a lower bound for E then c < c0 .

We denote this as c0 = inf E.

Example 0.1. Here are a few examples of infimums and supremums:

ˆ S = {−2, −1, 0, 1, 2}. Then, inf S = −2 and sup S = 2.

5
 ˆ But, note that the supremum nor the infimum need to be in S. Consider the sets

S = {q ∈ R | 0 < q < 1} = (0, 1).

Then, inf S = 0 ∈
/ E and sup S = 1 ∈
/ S.

ˆ Furthermore, neither the supremum nor the infimum need exist. Consider the sets S = Z .
Then, inf S = 1, but sup S does not exist as there is not an integer greater than all natural
numbers.
Now, we are ready to state the completeness Axiom.
ˆ If S is a non-empty subset of R which is bounded above, then it has a least upper bounded
in R

ˆ If S is a non-empty subset of R which is bounded below, then it has a greatest lower

bounded in R
Definition 0.5. Let S ⊂ R and m, M ∈ R, then
ˆ M is the maximum element of S if M ∈ S and a ≤ M for all a ∈ S.

ˆ m is the minimum element of S if m ∈ S and m ≤ a for all a ∈ S.

Theorem 0.6. If x ∈ R and

x = sup{q ∈ Q | q > 0, q 2 < 2}

√
then x > 0 and x2 = 2 (x = 2).
Proof : Let E equal the set on the right hand side, and suppose x ∈ R such that x = sup E.
Then, since 1 ∈ E and x is an upper bound for E, 1 ≤ x =⇒ x > 0.n o
2−x2
We now prove that x2 ≥ 2. Suppose that x2 < 2. Define h = min 21 , 2(2x+1) < 1. Then, if
x2 < 2 then h > 0. We now prove that x + h ∈ E. Indeed,

(x + h)2 = x2 + 2xh + h2
< x2 + h(2x + 1)

as h < 1. Hence
2x + 1
(x + h)2 ≤ x2 + (2 − x2 ) ·
2(2x + 1)
2 − x2
= x2 +
2
2−2
<2+
2
= 2.

6
Therefore, x + h ∈ E and x + h > x =⇒ x is not an upper bound for E. Therefore, x ̸= sup E
which is a contradiction. Hence, x2 ≥ 2.
2
We now prove that x2 ≤ 2. Suppose x2 > 2. Let h = x 2x−2 . Hence, if x2 > 2 then h > 0 and
x − h > 0. We will show that x − h is an upper bound for E. We have

(x − h)2 = x2 − 2xh + h2
= x2 − (x2 − 2) + h2
= 2 + h2
> 2.

Let q ∈ E. Then, q 2 < 2 < (x − h)2 =⇒ (x − h)2 − q 2 > 0. Hence,

((x − h) + q)((x − h) + q) > 0 =⇒ (x − h) − q > 0.

Thus, for all q ∈ E, q < x − h < x =⇒ x ̸= sup E. This is a contradiction. Therefore,

x2 = 2.

Theorem
√ 0.7. The set E = {q ∈ Q | q > 0 and q 2 < 2} does not have a supremum in Q
( 2∈
/ Q).

Proof : Suppose there exists an x ∈ Q such that x = sup E. Then, by our previous theorem,
x2 = 2. In particular, note that x > 1 as otherwise x ≤ 1 =⇒ 2 = x2 < 12 . Thus, ∃m, n ∈ N
such that m > n and x = m n . Therefore, ∃n ∈ N such that nx ∈ N. Let

S = {k ∈ N | kx ∈ N}.

Then, S ̸= ∅ since n ∈ S. By the well-ordering property of N, S has a least element k0 ∈ S. Let

k1 = k0 x − k0 ∈ Z. Then, k1 = k0 (x − 1) > 0 since k0 ∈ N and x > 1. Therefore, k1 ∈ N. Now
x2 = 2 =⇒ x < 2, as otherwise x2 > 4 > 2. Thus, k1 = k0 (x − 1) < k0 (2 − 1) = k0 . So, k1 ∈ N
and k1 < k0 =⇒ k1 ∈ / S as k0 is the least element of S. But,

xk1 = k0 x2 − xk0 = 2k0 − xk0 = k0 − k1 ∈ N =⇒ k1 ∈ S.

This is a contradiction. Thus, ̸ ∃x ∈ Q such that x = sup E.

The Archimedian Property, Density of the Rationals

x+y
For all x, y ∈ R and x < y, there exists an r ∈ R such that x < r < y (take r = 2 ).
Can we find r ∈ Q such that x < r < y? (The answer is yes!)

Theorem 0.8. 1. (Archimedian Property) If x, y ∈ R and x > 0, then ∃n ∈ N such that

nx > y.

2. (Density of Q) If x, y ∈ R and x < y then ∃r ∈ Q such that x < r < y.

7
 Proof :
1. Suppose that x, y ∈ R and x > 0. Then we wish to show that ∃n ∈ N such that n > xy .
Suppose this is not the case. Then, ∀n ∈ N, n ≤ xy . In other words, N is bounded above
by xy . However, this is a contradiction. Therefore, ∃n ∈ N such that n ≥ xy .

2. Suppose x, y ∈ R and x < y. Then, there are three cases:

ˆ 0 ≤ x < y,
ˆ x < 0 < y, and
ˆ x < y ≤ 0.

For the second case, take r = 0 ∈ Q. So, assume that 0 ≤ x < y. Then, by the Archimedian
Property, ∃n ∈ N such that n(y − x) > 1. Again by the Archimedean property, ∃l ∈ N
such that l > nx. Thus, consider the set

S = {k ∈ N | k > nx}.

By the well-ordering property of N, S has a least element, m ∈ S =⇒ nx < m =⇒ x <

m
n ∈ Q.
m
Since m − 1 ∈
/ S, m − 1 ≤ nx =⇒ m ≤ nx + 1 < ny. Hence, n < y. Therefore,
m
x< < y.
n
If instead we have x < y ≤ 0, then 0 ≤ −y < −x =⇒ ∃r̃ ∈ Q such that

−y < r̃ < x =⇒ x < −r̃ < y

by the previous case.

Corollary 0.9. For every x ≥ 0 there is an n ∈ N such that

n − 1 ≤ x < n.
1

Exercise 0.10. 1 = sup 1 − n |n∈N .
If n ∈ N, then 1 − n1 < 1 =⇒ 1 is an upper bound of this set. Suppose that x is an upper bound
for the set {1−1/n | n ∈ N}. We now prove that x ≥ 1. For the sake of contradiction, assume that
x < 1. By the Archimedean property, there exists an n ∈ N such that 1 < n(1 − x). Therefore,
∃n ∈ N such that x < 1 − 1/n. Hence, x is not an upper bound for the set {1 − 1/n | n ∈ N} if
x < 1. Thus, if x is an upper bound, x ≥ 1. Therefore,
 
1
sup 1 − | n ∈ N = 1.
n
We now begin proving some theorems about supremums and infinimums which will make
them easier to use.

8
Theorem 0.11. Suppose that S ⊂ R is nonempty and bounded above. Then, x = sup S if and
only if

1. x is an upper bound for S.

2. for all ϵ > 0, ∃y ∈ S such that x − ϵ < y ≤ x.

Proof : This is left as an exercise.

For x ∈ R and A ⊂ R, define

x + A := {x + a | a ∈ A} & xA := {xa | a ∈ A}. (0.12)

Theorem 0.13. Using this new notation, we have the following theorems:

1. If x ∈ R and A is bounded above, then x + A is bounded above and

sup(x + A) = x + sup A.

2. If x > 0 and A is bounded above then xA is bounded above and

sup(xA) = x sup A.

Proof :

1. Suppose that x ∈ R and A is bounded above. Therefore, sup A ∈ R by the least upper
bound property of R. Then, ∀a ∈ A, a ≤ sup A. Hence,

∀a ∈ A, x + a ≤ x + sup A.

Hence, x + sup A is an upper bound for x + A. Let ϵ > 0. Then, ∃y ∈ A such that

sup A − ϵ < y ≤ sup A =⇒ (x + sup A) − ϵ < y + x ≤ x + sup A.

Therefore, by our previous theorem, x + sup A = sup(x + A).

2. Suppose that x > 0 and A is bounded above. Thus, sup A ∈ R. Then, ∀a ∈ A, a ≤ sup A
and thus xa ≤ x sup A. Hence, x sup A is an upper bound of xA. Let ϵ > 0. Then ∃y ∈ A
such that
ϵ
sup A − < y ≤ sup A =⇒ x sup A − ϵ < xy ≤ x sup A.
x
Therefore, by the previous theorem, sup(xA) = x sup A.

Theorem 0.14. Let A, B ⊂ R such that ∀x ∈ A, ∀y ∈ B, x ≤ y. Then, sup A ≤ inf B.

Proof : The proof of this is left to the reader.

9
 Week 2,3: Sequences

1 Lecture 4:
Convergent Sequences of the Real Numbers
Definition 1.1. A sequence {xn } is said to converges to x ∈ R, if for every ε > 0 there exists
N ∈ N such that |xn − x| < ε for every n ≥ N. We write

lim xn = x.
n→∞

A sequence that converges is said to be convergent. Otherwise, we say the sequence diverges or
that it is divergent.
1
Example 1.1. limn→∞ n2 +2n+100
= 0.
−1
Proof. Let ϵ > 0 and choose M ∈ N such that M > ϵ 2 . Then, ∀n ≥ M ,
 
 1  1 1 1
 n2 + 2m + 100 − 0 ≤ n2 + 2m + 100 ≤ 2n ≤ 2M < ϵ.
 

Example 1.2. Consider the sequence xn = (−1)n . This sequence is divergent.

Proof. Let x ∈ R. We claim limn→∞ (−1)n ̸= x. To prove this, we simply need find an epsilon
that stops the sequence from converging. For instance, consider ϵ0 = 21 . Then, for M ∈ N,

1 = |(−1)M − (−1)M +1 | ≤ |(−1)M − x| + |(−1)M +1 − x|.

1
Thus, either |(−1)M − x| ≥ 2 or |(−1)M +1 − x| ≥ 12 . In either case, this shows that the limit
cannot converge to x.

A sequence {xn } is bounded if there exists a B ∈ R such that

|xn | ≤ B, ∨n ∈ N.

Theorem 1.1. If {xn } is convergent, then {xn } is bounded.

Proof : Suppose that limn→∞ xn = x. Thus, there exists an M ∈ N such that |xn − x| < 1
for all n ≥ M . Let
B = max{|x1 |, |x2 |, . . . , |xM −1 |, |x| + 1}.
If n < M , then |xn | ≤ B by construction. If n ≥ M , then

|xn | ≤ |xn − x| + |x| < 1 + |x| ≤ B.

10
Definition 1.2 (Monotone). A sequence {xn } is monotone increasing if ∀n ∈ N, xn ≤ xn+1 .
A sequence {xn } is monotone decreasing if ∀n ∈ N, xn ≥ xn+1 . If {xn } is either monotone
increasing or monotone decreasing, we say {xn } is monotone.
1
Example 1.3. For example, xn = n is monotone, yn = − n1 is monotone increasing, and (−1)n
is neither.

Theorem 1.2. Let {xn } be a monotone increasing sequence. Then, {xn } is convergent if and
only if {xn } is bounded. Moreover, limn→∞ xn = sup{xn | n ∈ N}.

Proof : Firstly, we know that if {xn } is convergent then it is bounded by the previous
theorem, Now assume that {xn } is bounded. Then, x := sup{xn | n ∈ N} exists in R by the
lowest upper bound property of R. We now prove that

lim xn = x.
n→∞

Let ϵ > 0. Then, ∃M0 ∈ N such that

x − ϵ < x M0 < x

since x is the supremum of the set. Let M = M0 . Then, ∀n ≥ M , we have

x − ϵ < xM0 = xM ≤ xn ≤ x < x + ϵ.

Therefore, |xn − x| < ϵ. Therefore, xn → x.

Theorem 1.3. Let {xn } be a monotone decreasing function. Then, {xn } is convergent if and
only if {xn } is bounded. Moreover,

lim xn = inf{xn | n ∈ N}
n→∞

The proof of this is similar to the previous theorem and is thus omitted.

Definition 1.3 (Subsequence). Let {nk } be a strictly increasing sequence of natural numbers.
Then the sequence
{xnk }∞
k=1

is called a subsequence of {xn }.

Consider the sequence {xn } = n – in other words, the sequence 1, 2, 3, 4, . . . . Then, the
following are subsequences of xn :

1, 3, 5, 7, 9, 11, . . .
2, 4, 6, 8, 10, . . .
2, 3, 5, 7, 11, 13, . . . .

11
The first two are described by xnk = x2k and xnk = x2k−1 respectively.
How would we describe the third?
Continuing to let {xn = n}n , the following are not subsequences:

1, 1, 1, 1, 1, 1, . . .
1, 1, 3, 3, 5, 5, . . . .

Now consider the sequence {(−1)n }. Then we have the subsequences

xnk = x2k−1 → −1, −1, −1, . . .

ynk = x2k → 1, 1, 1, . . . .

Theorem 1.4. If {xn } converges to x, then any subsequence of xn will converge to x.

Proof : Suppose limn→∞ xn = x. Let ϵ > 0. Then, ∃M0 ∈ N such that ∀n ≥ M0 ,

|xn − x| < ϵ.

Choose M = M0 . If k ≥ M , then nk ≥ k ≥ M = M0 . Hence, for all ϵ > 0 there exists an M

such that for all nk > M ,
|xn−k − x| < ϵ.

Remark 1.5. Notice that this also implies that the sequence {(−1)n }n is divergent.

Lecture 6
The Squeeze Theorem
Theorem 1.6 (Squeeze Theorem). Let {an }, {bn }, and {xn } be sequences such that ∀n ∈ N,

an ≤ xn ≤ bk .

Suppose that {an } and {bn } converge and

lim an = x = lim bn .
n→∞ n→∞

Therefore, {x} converges and limn→∞ xn = x.

Remark 1.7. We sometimes abbreviate the Squeeze Theorem to ST.

Proof : Let ϵ > 0. Since limn→∞ an = x, there exists an M0 ∈ N such that for all n ≥ M0 ,

|an − x| < ϵ =⇒ x − ϵ < an .

12
Since limn→∞ bn = x, ∃M1 ∈ N such that ∀n ≥ M1 ,

|bn − x| < ϵ =⇒ bn < x + ϵ.

Choose M = max{M0 , M1 }. Then, if n ≥ M , then

x − ϵ < an ≤ xn ≤ bn < x − ϵ =⇒ |xn − x| < ϵ.

Therefore, {xn } is convergent and limn→∞ xn = x.

Theorem 1.8. Another way to check that a sequence xn → x, is stated below:

lim xn = x ⇐⇒ lim |xn − x| = 0.

n→∞ n→∞

Hence, we can consider a sequence like the following:

Example 1.4. Show that

n2
lim = 1.
n→∞ n2 + n + 1

Proof. We have

n2
   
   −n − 1  n+1 n+1 1
 n2 + n + 1 − 1 =  n2 + n + 1  = n2 + n + 1 ≤ n2 + n = n .
   

Thus,
n2 n2
   
  1  
0≤ 2
 − 1 ≤ → 0 =⇒ lim  2
  − 1 = 0
n +n+1 n n→∞ n + n + 1

by the Squeeze Theorem.

How do limits interact with ordering?

Theorem 1.9. Let {xn } and {yn } be sequences of real numbers. Then,

1. if {xn } and {yn } are convergent sequences and ∀n ∈ N xn ≤ yn , then

limn→∞ xn ≤ limn→∞ yn .

2. if {xn } is a convergent sequence and ∀n ∈ N x ≤ xn ≤ bn then a ≤ limn→∞ xn ≤ b.

Proof :

1. Let x = limn→∞ xn and y = limn→∞ yn . Suppose for the sake of contradiction that y < x.
Then, ∃M0 ∈ N such that ∀n ≥ M0
x−y
|yn − y| <
2

13
 And ∃M1 ∈ N such that for all n ≥ M1 ,
x−y
|xn − x| < .
2
Then, if M = M0 + M1 ≥ max{M0 , M1 },
x−y x+y x−y
yM < +y = =x− + x < xM .
2 2 2
However, this would imply that yM < xM which contradicts ∀n ∈ Nxn ≤ yn .

2. Apply part 1 to proof part 2, by considering yn = a ≤ xn ≤ b = zn for all n ∈ N.

Operations Involving Convergent Sequences

Theorem 1.10. Suppose limn→∞ xn = x and limn→∞ yn = y. Then,

1. {xy + yn }n is convergent and limn→∞ (xn + yn ) = x + y.

2. ∀c ∈ R, {cxn }n is convergent and limn→∞ cxn = cx.

3. {xn · yn } is convergent and limn→∞ xn yn = xy.

4. If ∀n ∈ N, yn ̸= 0 and y ̸= 0, then {xn /yn }n is convergent and

xn x
lim = .
n→∞ yn y

Proof :

1. Let ϵ > 0. Then, since xn → x, ∃M0 ∈ N such that ∀n ≥ M0 , |xn − x| < 2ϵ . Since yn → y,
∃M1 ∈ N such that ∀n ≥ M1 , |yn − y| < 2ϵ . Hence, letting M = max{M0 , M1 }, we get for
all n ≥ M ,
ϵ ϵ
|xn + yn − (x + y)| ≤ |xn − x| + |yn − y| < + = ϵ.
2 2
ϵ
2. Let ϵ > 0. Since xn → x, ∃M0 ∈ N such that ∀n ≥ M0 , |xn − x| < |c|+1 . Let M = M0 .
Then, ∀n ≥ M ,
|c|
|cxn − cx| = |c||xn − x| ≤ ·ϵ<ϵ
|c| + 1
|c|
since |c|+1 < 1.

14
 3. Since yn → y, {yn } is bounded. In other words, ∃B ≥ 0 such that ∀n ∈ N, |yn | ≤ B.
Then,

|xn yn − xy| = |(xn − x)yn + (yn − y)x|

≤ |xn − x||yn + |x||yn − y|
≤ B|xn − x| + |x||yn − y|.

Therefore, 0 ≤ |xn yn − xy| ≤ B|xn − x| + |x||yn − y|. Since B|xn − x| + |x||yn − y| → 0,

by the Squeeze Theorem limn→∞ |xn yn − xy| = 0.

4. We prove y1n → y1 . We first prove ∃b > 0 such that ∀n ∈ N, |yn | ≥ b. Since yn → y and
y ̸= 0, ∃M0 ∈ N such that ∀n ≥ M0 ,

|y|
|yn − y| < .
2
By the Triangle Inequality, ∀n ≥ M0 ,

|y| |y|
|y| ≤ |yn − y| + |yn | ≤ + |yn | =⇒ |yn | ≥ .
2 2
n o
Let b = min |y1 |, . . . , |yM0 −1 |, |y|
2 . Then, ∀n ∈ N, |yn | ≥ b. Therefore,
 
1 1  |yn − y| 1
0≤ − =
 ≤ |yn − y|.
yn y |yn ||y| b|y|
 
By the Squeeze Theorem, limn→∞  y1n − y1  = 0. Therefore, limn→∞ 1
= y1 . Furthermore,
 
yn
 
by the proof before this (3.), it follows that limn→∞ xn · y1n = xy .

Remark 1.11. By induction, one can prove that

lim (xn )k = xk .
n→∞
√
Theorem 1.12. If {xn } is a convergent sequence such that ∀n ∈ N, xn ≥ 0, then { xn } is
convergent and
√ q
lim xn = lim xn .
n→∞ n→∞

Proof : Let x = limn→∞ xn .

Case 1: x = 0. Let ϵ > 0. Then, since xn → 0, there exists an M0 ∈ N such that ∀n ≥ M0 ,
xn = |xn − 0| < ϵ2 . Choose M = M0 . Then, ∀n ≥ M ,
√ √ √ √
| xn − 0| = xn < ϵ2 = ϵ.

15
 Case 2: x > 0. We have ∀n ∈ N,
√ √
√ √ 

√  xn − x √
| xn − x| =  √
 √ · ( xn + x)
xn + x
1
=√ √ |xn − x|
xn + x
1
≤ √ |xn − x|.
x
Hence,
√ √ 1
0 ≤ | xn − x| ≤ √ |xn − x|
x
∀n ∈ N. Hence, by the Squeeze Theorem,
√ √
lim | xn − x| = 0.
n→∞

Theorem 1.13. If {xn } is convergent and limn→∞ xn = x, then {|xn |} is convergent and
limn→∞ |xn | = |x|.
√
Proof : Firstly, note that ∀x ∈ R, x2 = |x|. Then,
p √
lim |xn | = lim x2n = x2 = |x|
n→∞ n→∞
by the previous theorem.
Theorem 1.14. If c ∈ (0, 1), then limn→∞ cn = 0. If c > 1, then {cn } is unbounded.
Proof : If 0 < c < 1, we claim that ∀n ∈ N, 0 < cn+1 < cn < 1. We can prove this
through induction. Firstly, notice that 0 < c2 < c < 1 since c > 0 and c < 1. Now assume that
0 < cm+1 < cm . Then, multiply by c > 0 to obtain
0 < cm+1 · c = c(m+1)+1 < cm · c = c(m+1) .
By induction, our claim holds. Thus, {cn } is a monotone decreasing sequence and is bounded
below. Thus, {cn } is convergent. Let L = limn→∞ cn . We will prove that L = 0. Let ϵ > 0.
Then, ∃M ∈ N such that ∀n ≥ M , |cn − L| < (1 − c) 2ϵ . Therefore,
(1 − c)|L| = |L − cL| = |L − cM +1 + cM +1 − cL|
≤ |L − cM +1 | + c|cM − L|
ϵ ϵ
< (1 − c) + c(1 − c) < (1 − c)ϵ.
2 2
Therefore, ∀ϵ > 0, |L| < ϵ =⇒ L = 0.
Now let c > 1. We have to show that ∀B ≥ 0, ∃n ∈ N such that cn > B. Let B ≥ 0. Choose
B
n ∈ N such that n > c−1 . Then,
cn = (1 + (1 − c))n ≥ 1 + n(c − 1) ≥ n(c − 1) > B.
To see why this center inequality is true, see the last theorem shown in Lecture 1.

16
Lecture 7:
Bolzano-Weierstrass theorem
Theorem 1.15. Some Special Sequences

1. If p > 0, then limn→∞ n−p = 0.

1
2. If p > 0 then p n = 1.
1
3. limn→∞ n n = 1.

Proof :

1. Let ϵ > 0. Then, choose M > (1/ϵ)1/p . Hence, if n ≥ M ,

 
1
 − 0 = 1 ≤ 1 < ϵ.

 np  |np | Mp

2. Suppose p > 1. Then, p1/n − 1 > 0 which may be proven by induction. Furthermore, we
have

p = (1 + (p1/n − 1))n
≥ 1 + n(p1/n − 1).

Therefore, 0 < p1/n − 1 ≤ p−1 n . Hence, we may apply the Squeeze Theorem, obtaining
limn→∞ |p 1/n − 1| = 0. If p < 1, then
1 1
lim p1/n = lim = = 1.
n→∞ n→∞ (1/p)1/n 1

Furthermore, if p = 1 then it is clear that limn→∞ p1/n = 1. Hence, in all cases, the limit
is 1.

3. Let xn = n1/n − 1 ≥ 0. We want to show that limn→∞ xn = 0, as this will imply the end
result. Notice that
n    
X n j n 2 n! n(n − 1) 2
n = (1 + xn )n = xn ≥ x = · x2 = · xn .
j 2 n 2(n − 2)! n 2
j=0

Thus, for n > 1, r

2
0 ≤ xn ≤ =⇒ xn → 0.
n−1

17
Lim sup Lim inf
Does a bounded sequence have a convergent subsequence?

Definition 1.4 (Limsup/Liminf). Let {xn } be a bounded sequence. We define, if the limits
exist,

lim sup xn := lim (sup{xk | k ≥ n})

n→∞ n→∞

lim inf xn := lim (inf{xk | k ≥ n}).

n→∞ n→∞

These are called the limit superior and limit inferior respectively.

We will now show that these limits always exist.

Theorem 1.16. Let {xn } be a bounded sequence, and let

an = sup{xk | k ≥ n}
bn = inf{xk | k ≥ n}.

Then,

1. {an } is monotone decreasing and bounded, and {bn } is monotone increasing and bounded.

2. lim inf n→∞ xn ≤ lim supn→∞ xn .

Proof

1. Since, ∀ ∈ N,
{xk | k ≥ n + 1} ⊆ {xk | k ≥ n},
we have that an+1 = sup{xk | k ≥ n + 1} ≤ sup{xk | k ≥ n} = an .
Similarly, ∀n ∈ N, bn+1 ≥ bn . Given {xn } is a bounded sequence, ∃B ≥ 0 such that
∀n ∈ N,
−B ≤ xn ≤ B.
Therefore, ∀n ∈ N,
−B ≤ bn ≤ an ≤ B
which implies both sequences are bounded.

2. By the above equation, ∀n ∈ N, bn ≤ an =⇒ lim inf n→∞ xn = limn→∞ bn ≤ limn→∞ an =

lim supn→∞ xn .

Let’s consider a few examples.

Example 1.5. Let xn = (−1)n . Calculate the lim inf and lim sup of this sequence.

18
Proof. Notice that {(−1)k | l ≥ n} = {−1, 1}. Thus, the supremum of these sets is always 1 and
the infimum is always −1. Therefore,

lim sup xn = 1 and lim inf xn = −1.

n→∞ n→∞

Example 1.6. Let xn = n1 . Calculate the lim inf and lim sup of this sequence.
Proof. We may do this directly:
1
sup{1/k | k ≥ n} = → 0 =⇒ lim sup xn = 0.
n n→∞
inf{1/k | k ≥ n} = 0 → 0 =⇒ lim inf xn = 0.
n→∞

The limit inferior and the limit superior allow us to answer the question posed at the begin-
ning of this section.
Theorem 1.17. Let {xn } be a bounded sequence. Then, there exists subsequences {xnk } and
{xmk } such that

lim xnk = lim sup xn

k→∞ n→∞
lim xmk = lim sup xn .
k→∞ n→∞

Proof : Let an = sup{xk | k ≥ n}. Then, ∃n1 ∈ N such that a1 − 1 < xn1 ≤ a1 . Now,
∃n2 > n1 such that
1
an1 +1 − < xn2 ≤ an1 +1
2
since
an+1 = sup{xk | k ≥ n1 + 1}.
Similarly, ∃n3 > n2 such that
1
an2 +1 −< xn3 ≤ an2 +1 .
3
Continuing in this way, we obtain a sequence of integers n1 < n2 < n3 < . . . such that
1
ank +1 − < xnk ≤ ank +1 .
k+1
Given limk→∞ ank +1 = lim supn→∞ xn , by the Squeeze Theorem,

lim xnk = lim sup xn .

k→∞ n→∞

The direction for the lim inf works out the same way so that portion of the proof is left to the
reader.

19
Theorem 1.18 (Bolzano-Weierstrass). Every bounded sequence has a convergent subsequence.

Remark 1.19. We may abbreviate the Bolzano-Weierstrass theorem to B-W.

Proof : This follows immediately from the previous theorem, but is so important that it itself
is a theorem.
When it is clear, we may have the following notational shorthand: lim inf n→∞ xn := lim inf xn ,
and
lim supn→∞ xn := lim sup xn .

Theorem 1.20. Let {xn } be a bounded sequence. Then, {xn } converges if and only if lim inf xn =
lim sup xn .

Proof ( ⇐= ) Suppose lim inf xn = lim sup xn . Then, ∀n ∈ N,

inf{xk | k ≥ n} ≤ xn ≤ sup{xk | k ≥ n}.

By the Squeeze Theorem, since limk→∞ inf{xk | k ≥ n} = limk→∞ sup{xk | k ≥ n} by assump-

tion, we have
lim xn = lim inf xn = lim sup xn .
n→∞

Therefore, xn converges.
( =⇒ ) Let x = limn→∞ xn . Therefore, every subsequence of {xn } converges to x, so
lim inf xn = x and lim sup xn = x by a theorem we proved in Lecture 7. Hence, lim inf xn =
lim sup xn .

Lecture 8:
Cauchy Sequences
Definition 1.5. A sequence {xn } is Cauchy if ∀ϵ > 0 ∃M ∈ N such that for all n, k ≥ M ,

|xn − xk | < ϵ.
1
Example 1.7. Show the sequence xn = n is Cauchy.
1
Proof. Let ϵ > 0 and choose M ∈ N such that M < 2ϵ . Then, if n, k ≥ M , then
 
1
 − 1  ≤ 1 + 1 ≤ 2 < ϵ.

n k n k M

By the negation of the definition, a sequence {xn } is not Cauchy if ∃ϵ0 > 0 such that for all
M ∈ N, ∃n, k ≥ M such that |xn − xk | ≥ ϵ0 .

20
Example 1.8. Show the sequence xn = (−1)n is not Cauchy.
Proof. Choose ϵ = 1 and let M ∈ N. Choose n = M and k = M + 1. Then,
|(−1)n − (−1)k | = 2 ≥ 1.

Theorem 1.21. If {xn } is Cauchy, then {xn } is bounded.

Proof : If {xn } is Cauchy then ∃M ∈ N such that for all n, k ≥ M ,
|xn − xk | < 1.
Then, for all n ≥ M , |xn − xM | < 1. Hence,
|xn | ≤ |xn − xM | + |xM | < |xM | + 1.
Let B = |x1 | + · · · + |xM | + 1. Then, for all n ∈ N, |xn | ≤ B.
Theorem 1.22. If {xn } is Cauchy and a subsequence {xnk } converges, then {xn } converges.
Proof : Suppose that {xnk } is a subsequence of {xn } such that limk→∞ xnk = x. We claim
that xn → x. Let ϵ > 0. Since xnk → x, there exists M0 ∈ N such that ∀k ≥ M0 ,
ϵ
|xnk − x| < .
2
Since {xn } is Cauchy, there exists an M1 ∈ N such that for all n ≥ M1 and m ≥ M1 ,
ϵ
|xn − xm | < .
2
Choose M = M0 + M1 . If n ≥ M , then nM ≥ M ≥ M0 and n≥ M1 . Therefore,
ϵ ϵ
|xn − x| ≤ |xn − xnM | + |xnM − x| < + = ϵ.
2 2

Theorem 1.23. A sequence of real numbers {xn } is Cauchy if and only if {xn } is convergent.
Proof : ( =⇒ ) If {xn } is Cauchy, then {xn } is bounded. Therefore, {xn } has a convergent
subsequence by Bolzano-Weierstrass. By the previous theorem, we thus have that {xn } is
convergent.
( ⇐= ) Suppose that {xn } is convergent and x = limn→∞ xn . Let ϵ > 0. Since xn → x,
∃M0 ∈ N such that ∀n ≥ M0 ,
ϵ
|xn − x| < .
2
Choose M = M0 . Then, if n, k ≥ M ,
ϵ ϵ
|xn − xk | ≤ |xn − x| + |xk − x| < + = ϵ.
2 2
Therefore, {xn } is Cauchy.

21
Week 4: Series
P∞ P
Definition
P 1.6. Given {xn }, the symbol n=1 xn or xn is the series associated to {xn }. We
say xn converges if the sequence
( m
)∞
X
sm = xn
n=1 m=1
P
converges.PWe call the terms of {sm } the partial sums. If limm→∞ sm = s, we write s = xn
and treat xn as a number.

Remark 1.24. A series need not start at n = 1.

P∞ 1
Example 1.9. n=1 n(n+1) converges.

Proof. We may do show this directly by consider the partial sums:

m m
X 1 X1 1
sm = = −
n(n + 1) n n+1
n=1 n=1
   
1 1 1 1 1
= 1 + + ··· + − + ... +
2 m 2 m m+1
1
=1− .
m+1
1
Thus, sm = 1 − → 1. Hence, the partial sums converge and thus the series converges.
m+1

Theorem 1.25. If |r| < 1 then ∞ n

P
n=0 r converges and

∞
X 1
rn = .
1−r
n=0

Proof : We have ∀m ∈ N,
m
X 1 − rm+1
sm = rn =
1−r
n=0

by induction. Since |r| < 1, limm→∞ |r|m+1 = 0. Therefore,

1−0 1
lim sm = = .
m→∞ 1−r 1−r

P∞
Theorem 1.26. Let {xn } be a sequence and let M ∈ N. Then, n=1 xn converges if and only
if ∞
P
n=M x n converges.

22
 Proof : The partial sums satisfy, for all m ∈ N,
m
X m
X M
X
xn = xn + xn .
n=1 n=M n=1

P
Definition 1.7. xn is Cauchy if the sequence of partial sums is Cauchy.
P P
Theorem 1.27. xn is Cauchy ⇐⇒ xn is convergent.
Proof : This follows by the analogous theorem for regular sequences of real numbers proven
earlier.
P
Theorem 1.28. xn is Cauchy if and only if ∀ϵ > 0, ∃M ∈ N such that for all m ≥ M and
ℓ > m,  ℓ 
 X 
xn  < ϵ.
 

 
n=m+1
P
Proof : ( =⇒ ) Suppose xn is Cauchy. Let ϵ > 0. Then, ∃M0 ∈ N such that ∀m, ℓ ≥ M0 ,
|sm − sℓ | < ϵ.
Choose M = M0 . Then, if m ≥ M and ℓ > m, then
 ℓ 
 X 
xn  = |sℓ − sm | < ϵ.
 

 
n=m+1

The other direction is left as an exercise.

P
Theorem 1.29. If xn converges then limn→∞ xn = 0.
P P P
Proof : Suppose xn converges. Then, xn is Cauchy. Let ϵ > 0. Since xn is Cauchy,
∃M0 ∈ N such taht for all ℓ > m ≥ M0 ,
 
ℓ
 X 
xn  < ϵ.
 

 
n=m+1

Choose M = M0 + 1. Then, if m ≥ M =⇒ m − 1 ≥ M0 . Therefore,

 
Xm 
|xm | =  xn  < ϵ
 
 
n=m

by taking ℓ = m.
P∞ n
Theorem 1.30. If |r| ≥ 1, then n=0 r diverges.
Proof : If |r| ≥ 1, then limm→∞ rm ̸= 0. Therefore, ∞ n
P
n=0 r diverges, as if this wasn’t the
case then
limm→∞ rm = 0 by the previous theorem which is a contradiction.

23
Convergence tests
P
Recall that
P if xn converges then limn→∞ xn = 0. Is the converse true? Does limn→∞ xn =
0 =⇒ xn converges?

Example 1.10. The series ∞ 1

P
n=1 n does not converge.

Proof : We will show that there exists a subsequence of sm = m 1

P
n=1 n which is unbounded,
which will imply the series diverges. Consider, for ℓ ∈ N,

2 ℓ
X 1
s2ℓ = .
n
n=1

Then,
       
1 1 1 1 1 1 1
s2ℓ = 1 + + + + + ··· + + ... + · · · +
2 3 4 5 8 2ℓ−1 + 1 2ℓ
ℓ 2λ
X X 1
=1+
n
λ=1 n=2λ−1 +1
ℓ 2λ
X X 1
≥1+
2λ
λ=1 n=2λ−1 +1
ℓ
X 1 λ
=1+ (2 − (2λ−1 + 1) + 1)
2λ
λ=1
ℓ
X 2λ−1
=1+
2λ
λ=1
ℓ
=1+ .
2
Thus, {s2ℓ }∞
ℓ=1Pis unbounded which implies {s2ℓ } does not converge.
1
The series n is called the harmonic series.
P P P
Theorem 1.31. Let α ∈ R and xn and yn be convergent series. Then the series (αxn +
yn ) converges and X X X
(αxn + yn ) = α xn + yn .

Proof : The partial sums satisfy

m
X m
X m
X
(αxn + yn ) = α xn + yn .
n=1 n=1 n=1

24
By linear properties of limits, it follows that
m
X X X
lim (αxn + yn ) = α xn + yn .
m→∞
n=1

Series with non-negative terms are easier to work with than general series as then {sn } is a
monotone sequence.
P
Theorem 1.32. If ∀n ∈ N xn ≥ 0, then xn converges if and only if {sm } is bounded.
Proof : If xn ≥ 0 for all n ∈ N then
m+1
X m
X
sm+1 = xn = xn + xm+1 = sm + xm+1 ≥ sm
n=1 n=1

Thus, {sm } is a monotone increasing sequence. Therefore, {sm } converges if and only if {sm }
is bounded.
P P
Definition 1.8. xn converges absolutely if |xn | converges.
P P
Theorem 1.33. If xn converges absolutely then xn converges.
P P
Proof : Suppose P|xn | converges. We will then show that xn is Cauchy.
m Pm
Claim: ∀m ≥ 2, | n=1 xn | ≤ n=1 |xn |. We prove this claim by induction. For m = 2,
this
P states that |x1 + x2 | ≤ |x1 | + |x2 |, which follows by the Triangle Inequality. Suppose for all
 ℓ  Pℓ
 n=1 xn  ≤ n=1 |xn |. Then,
   
Xℓ+1  X ℓ  Xℓ ℓ+1
X
xn  ≤  xn  + |xℓ+1 | ≤ |xn | + |xℓ+1 | = |xn |.
   

   
n=1 n=1 n=1 n=1
P P P
We now prove that xn is Cauchy. Let ϵ > 0. Since |xn | converges, |xn | is Cauchy.
Therefore, there exists an M0 ∈ N such that for all ℓ > m ≥ M0 ,
ℓ
X
|xn | < ϵ.
n=m+1

Choose M = M0 . Then, for all ℓ > m ≥ M ,

 
 Xℓ  ℓ
X
x ≤ |xn | < ϵ.
 
 n
 
n=m+1 n=m+1
P
Hence, xn is Cauchy, and thus converges.
(−1)n
We will see that ∞
P
n=1 n is convergent but not absolutely convergent. Notice that it is
P  (−1)n  P 1
immediately clear that this series is not absolutely convergent as  n  = n (the harmonic
series), which doesn’t converge.

25
Theorem 1.34 (Comparison Test). Suppose for all n ∈ N 0 ≤ xn ≤ yn . Then,
P P
1. if yn converges, then xn converges.
P P
2. if xn diverges, then yn diverges.

Proof :
∞
yn converges, then { m
P P
1. If n=1 yn }m=1 is bounded. In other words, there exists a B ≥ 0
such that for all m ∈ N,
Xm
yn ≤ B.
n=1

Thus, for all m ∈ N, m

P Pm
n=1
P x n ≤ n=1 yn ≤ B. Therefore, the partial sums of {xn } are
bounded, which implies xn converges.

xn diverges, then { n=1 xn }∞

P Pm
2. If m=1 is unbounded. We now prove that
( m
)∞
X
yn
n=1 m=1

is also unbounded. Let B ≥ 0. Then, ∃m ∈ N such that

m
X
xn ≥ B.
n=1
Pm Pm Pm ∞
Therefore,
P n=1 y n ≥ n=1 x n ≥ B. Thus, { n=1 yn }m=1 is unbounded, which implies
yn diverges.

P∞ 1
Theorem 1.35. For p ∈ R, the series n=1 np converges if and only if p > 1.

Proof : ( =⇒ ) We prove thisPdirection through contradiction. Suppose ∞ 1

P
converges
n=1 np P
1 1 1 1
and p ≤ 1. Then, P np ≥ n , and n diverges. Therefore, by the Comparison Test, np also
1
diverges. Hence, if np converges, then p > 1.
( ⇐= ) Suppose p > 1. We first prove that a subsequence of the partial series is bounded.

26
 1
Claim 1: ∀k ∈ N, s2k ≤ 1 + 1−2−(p−1)
. Proof:

k 2 ℓ
X X 1
s2k = 1 +
np
ℓ=1 n=2ℓ−1 +1
k 2 ℓ
X X 1
≤1+
(2ℓ−1 + 1)p
ℓ=1 n=2ℓ−1 +1
k
X
≤1+ 2−p(ℓ−1) (2ℓ − (2ℓ−1 + 1) + 1)
ℓ=1
Xk
=1+ 2−(p−1)(ℓ−1)
ℓ=1
k−1
X
=1+ 2−(p−1)ℓ
ℓ=0
∞
X
≤1+ 2−(p−1)ℓ
ℓ=0
1
=1+
1 − 2−(p−1)
− 1 > 0, and using properties of geometric series. Thus, Claim 1 is proven.
using the fact that pP
Claim 2: {sm = m 1 m
n=1 np } is bounded. Proof: Let m ∈ N. Since 2 > m, we have that

m 2 m
X 1 X 1
sm = p
≤ n−p ≤ 1 + .
n=1
n
n=1
1 − 2−(p−1)

Hence, the partial sums are bounded, which implies {sm } converges.

Theorem 1.36 (Ratio test). Suppose xn ̸= 0 for all n and

|xn+1 |
L = lim
n→∞ |xn |

exists. Then,
P
1. if L < 1 then xn converges absolutely.
P
2. if L > 1 then xn diverges.

Proof : We will first prove the second part of this theorem.

27
 2) Suppose L > 1 and α ∈ (1, L). Then, there exists M0 ∈ N such that for all N ≥ M0 ,
|xn+1 |
|xn | ≥ α ≥ 1. Thus, for all n ≥ M0 ,

|xn+1 | ≥ |xn | =⇒ lim |xn | =

̸ 0.
n→∞
P
Therefore, xn diverges.

1) Now suppose that L < 1. Let α ∈ (L, 1). Then, there exists M0 ∈ N such that ∀n ≥ M0 ,
|xn+1 |
|xn | < α. Therefore,∀n ≥ M0 , |xn+1 | ≤ α|xn |. In other words, for all n ≥ M0 ,

|xn | ≤ α|xn−1 | ≤ α2 |xn−2 | ≤ · · · ≤ αn−M0 |xM0 |.

Let m ∈ N. Then,
m
X M
X 0 −1 m
X
|xn | = |xn | + |xn |
n=1 n=1 n=M0
M
X 0 −1 m
X
≤ |xn | + |xM0 | αn−M0
n=1 n=M0
M
X 0 −1 ∞
X
≤ |xn | + |xM0 | αℓ
n=1 ℓ=0
M 0 −1
X |xM0 |
= |xn | + .
1−α
n=1
∞
Therefore, { m
P P
n=1 |xn |}m=1 is bounded, and thus |xn | converges. Hence, xn is absolutely
convergent.

Let’s consider two examples where we can use the Ratio test.
(−1)n
Example 1.11. Show the series ∞
P
n=1 n2 +1 converges absolutely.

Proof. Notice
 (−1)n 
 
1 1
 n2 + 1  ≤ n2 + 1 < n2 ,
 

and hence  
 (−1)n+1 
 (n+1)2 +1  n2
lim  (−1)n  < lim = 1.
n→∞  n→∞ (n + 1)2
2
n +1 

P∞ xn
Example 1.12. Show that ∀x ∈ R, n=0 n! converges absolutely.

28
Proof. This immediately follows from the Ratio test, noting that

|x|n+1 n! |x|
lim · n = lim = 0.
n→∞ (n + 1)! |x| n→∞ n + 1

Remark 1.37. As seen above, the Ratio test can be really helpful to use when we have a (−1)n
or a factorial in the argument. Also note that if L = 1 then we the test doesn’t apply.
P
Theorem 1.38 (Root test). Let xn be a series and suppose that

L = lim |xn |1/n

n→∞

exists. Then,
P
1. if L < 1 then xn converges absolutely.
P
2. if L > 1 then xn diverges.

Proof :

1. Suppose L < 1. Let L < r < 1. Then, since |xn |1/n → L, ∃M ∈ N such that ∀n ≥ M ,
|xn |1/n < r. Therefore, for all n ≥ M , |xn | ≤ rn . Thus, for all m ∈ N,
m
X M
X −1 m
X
|xn | = |xn | + |xn |
n=1 n=1 n=M
M
X −1 Xm
≤ |xn | + rn
n=1 n=M
M
X −1 X∞
≤ |xn | + rn
n=1 n=M
M −1
X rM
= |xn | + .
1−r
n=1

∞
Thus, { m
P P
n=1 |xn |}m=1 is bounded, and thus |xn | converges.

2. Suppose L > 1. Then, since |xn |1/n → L > 1, there exists an M ∈ N such that for all
n ≥ M , |xP 1/n > 1. In other words, for all n ≥ M , |x | > 1. Therefore, lim
n| n n→∞ xn ̸= 0,
and thus xn diverges.

Remark 1.39. Again, note that if L = 1 then the test doesn’t apply.

29
Theorem 1.40P (Alternating Series test). Let {xn } be a monotone decreasing sequence such that
xn → 0. Then, (−1)n xn converges.

Proof : Let sm = m n
P
n=1 (−1) xn . Then,

2k
X
s2k = (−1)n xn
n=1
= (x2 − x1 ) + (x4 − x3 ) + · · · + (x2k − x2k−1 )
≥ (x2 − x1 ) + · · · + (x2k − x2k−1 ) + (x2k+2 − x2k+1 )
= s2(k+1)

as {xn } is a monotone decreasing sequence. Thus, {s2k }∞ k=1 is monotone decreasing. Further-
more,
s2k = −x1 + (x2 − x3 ) + (x4 − x5 ) + · · · + (x2k−2 − x2k−1 ) + x2k ≥ −x1 .
In other words, {s2k } is a bounded below monotone decreasing sequence. Thus, {s2k }∞
k=1 con-
∞
verges. Let s = limk→∞ s2k . We now prove {sm }m=1 converges to s.
Let ϵ > 0. Since s2k → s, ∃M0 ∈ N such that for all k ≥ M0 ,
ϵ
|s2k − s| < .
2
Since xn → 0, ∃M1 ∈ N such that ∀n ≥ M1 ,
ϵ
|xn | < .
2
m
Choose M = max{2M+ 0 + 1, M1 }. Suppose m ≥ M . If m is even, then 2 ≥ M0 + 1/2 ≥ M0 .
Therefore,
ϵ
|sm − s| = |s2· m2 − s| < < ϵ.
2
m−1
If m is odd, let k = 2 so m = 2k + 1. Then, m ≥ M =⇒ k ≥ M0 and m ≥ M1 . Then,

|sm − s| = |sm−1 + xm − s|
≤ |s2k − s + xm |
ϵ ϵ
≤ |s2k − s| + |xm | < + = ϵ.
2 2
(−1)n xn converges.
P
Thus, sm → s, and thus
P (−1)n
Corollary 1.41. We already showed that n does not absolutely converge. However,
P (−1)n
n converges.

Proof : This follows immediately from the Alternating Series test.

30
 P P
Theorem 1.42. Suppose P xn converges absolutely and xnP= x. Let σ : N → N be a
bijective function. Then, xσ(n) is absolutely convergent and xσ(n) = x. In other words,
absolute convergence implies if we rearrange the sequence the new series will still converge to
the same value of the original series.
P
Pm Proof : We first show |xσ(n)
P | converges, which is equivalent to showing the partial sums
n=1 |xσ(n) | is bounded. Since xn converges, ∃B ≥ 0 such that for all ℓ ∈ N,

ℓ
X
|xn | ≤ B.
n=1

Let m ∈ N. Then, σ({1, . . . , m}) is a finite subset of N. Thus, there exists an ℓ ∈ N such that

σ({1, . . . , m}) ⊂ {1, . . . , ℓ}.

Thus,
m
X X ℓ
X
|xσ(n) | = |xn | ≤ |xn | ≤ B.
n=1 n∈σ({1,...,m}) n=1
P P∞
Therefore, |xσ(n) | converges. Let x = n=1 xn , and let ϵ > 0. Then, ∃M0 ∈ N such that
∀m ≥ M0 , m 
X  ϵ
xn − x < .
 
 2


n=1
P
Since |xn | converges, ∃M1 ∈ N such that for all ℓ > m ≥ M1 ,
ℓ
X ϵ
|xn | < .
2
n=m+1

Let M2 = max{M0 , M1 }. Then, ∀ℓ > m ≥ M2 ,

m  ℓ
X  ϵ X ϵ
xn − x < and |xn | < .
 
 2 2


n=1 n=m+1

Since σ −1 ({1, . . . , M2 }) is a finite set, ∃M3 ∈ N such that

{1, . . . , M2 } ⊂ σ({1, . . . , M3 }).

31
Choose M = M3 . Thus, if m′ ≥ M ,
 m′   

X   X 
xσ(n′ ) − x =  xn − x
  

 ′  
n =1 n∈σ({1,...,m′ }) 
 
M
X X 

= xn − x + xn 
n=1 n∈σ({1,...,m′ })\{1,...,M } 
  max σ({1,...,m′ })
XM  X
≤ xn − x + |xn |
 
 
n=1 n=M +1
 
XM  X ℓ
≤ xn − x + |xn |
 
 
n=1 n=M +1
ϵ ϵ
< + = ϵ.
2 2

Week 5: Continuous Functions

Limits of Functions
Definition 1.9. Let S ⊂ R. x ∈ R is a neighborhood point of S if ∀δ > 0, (x−δ, x+δ)∩S \{x} =
̸
∅.
Examples.
1. S = {1/n | n ∈ N}. Here, 0 is a neighborhood point of S.

2. S = (0, 1). The set of neighborhood points of S is [0, 1].

3. S = Q. The set of neighborhood points of S is R.

4. S = {0}. There are no neighborhood points of S.

5. S = Z. There are no neighborhood points of S.

Theorem 1.43. Let S ⊂ R. Then, x is a cluster point of S if and only if there exists a sequence
{xn } of elements in S \ {x} such that xn → x.
Definition 1.10. Let S ⊂ R, let c be a neighborhood point of S, and f : S → R. We say that
f (x) converges to L ∈ R at c if ∀ϵ > 0 ∃δ > 0 such that if x ∈ S and 0 < |x − c| < δ, then
|f (x) − L| < ϵ.
We write limx→c f (x) = L.

32
Theorem 1.44. Let c be a neighborhood point of S ⊂ R, and let f : S → R. If f (x) → L1 and
f (x) → L2 as x → c, then L1 = L2 .

Proof : We will show ∀ϵ > 0, |L1 − L2 | < ϵ. Let ϵ > 0. Then, since f (x) → L1 and
f (x) → L2 , ∃δ1 such that if x ∈ S and 0 < |x − c| < δ1 then

|f (x) − L1 | < ϵ/2

and ∃δ2 > 0 such that if x ∈ S and 0 < |x − c| < δ2 , then

|f (x) − L2 | < ϵ/2.

Let δ = min{δ1 , δ2 } > 0. Then, since c is a cluster point of S, ∃x0 ∈ S such that

0 < |x0 − c| < δ =⇒ |L1 − L2 | = |L1 − f (x0 ) + f (x0 ) + L2 | ≤ |L1 − f (x0 )| + |f (x0 ) − L2 | < ϵ.

Example 1.13. Let f (x) = ax + b. Then, for all c ∈ R, limx→c f (x) = ac + b.

ϵ
Proof. Let ϵ > 0. Choose δ = 1+|a| . Then, if x ∈ R and 0 < |x − c| < δ, then

|f (x) − (ac + b)| = |a(x − c)|

= |a||x − c|
< |a|δ
|a|
= ϵ < ϵ.
1 + |a|

√ √
Example 1.14. Let f (x) = x. Then, ∀c > 0, limx→c f (x) = c.
√
Proof. Let ϵ > 0. Choose δ = ϵ c. Then, if x > 0 and 0 < |x − c| < δ, then
√ √ √
|f (x) − c| = | x − c|
 √ √ √ √ 
 ( x − c)( x + c) 
=  √ √ 
x+ c 
|x − c|
=√ √
x+ c
|x − c|
≤ √
c
δ
< √ = ϵ.
c

33
 (
1 x ̸= 0
Example 1.15. Let f (x) = . Then, limx→0 f (x) = 1. Notably, limx→0 f (x) ̸= f (0)!
2 x=0
Proof. Let ϵ > 0 and choose δ = 1. Then, if 0 < |x − 0| < 1 then x ̸= 0 =⇒

|f (x) − 1| = |1 − 1| = 0 < ϵ.

Theorem 1.45. Let S ⊂ R, c a cluster point of S, and let f : S → R. Then, the following are
equivalent:
1. limx→c f (x) = L and

2. for every sequence {xn } in S \ {c} such that xn → c, we have f (xn ) = L.

Proof : (1. =⇒ 2.): Suppose limx→c f (x) = L. Let {xn } be a sequence in S \ {c} such
that xn → c. We want to show that f (xn ) → L. Let ϵ > 0. Given limx→c f (xn ) = L, ∃δ > 0
such that if x ∈ S and 0 < |x − c| < δ then |f (x) − L| < ϵ. Since xn → c, ∃M0 ∈ N such that
∀n ≥ M0 , 0 < |xn − c| < δ.
Choose M = M0 . Then, ∀n ≥ M , if 0 < |xn − c| < δ then |f (xn ) − L| < ϵ. Thus, f (xn ) → L.
(2. ⇐= 1.): Suppose 2. holds, and assume for the sake of contradiction that 1) is false.
Then, ∃ϵ0 > 0 such that ∀δ > 0, ∃x ∈ S such that

0 < |x − c| < δ and |f (x) − L| ≥ ϵ0 .

1
Then, ∀n ∈ N, ∃xn ∈ S such that 0 < |xn − c| < n and |f (xn ) − L| ≥ ϵ0 . By the Squeeze
Theorem applied to
1
0 < |xn − c| < ,
n
xn → c. Then, by 2.,
0 = lim |f (xn ) − L| ≥ ϵ0
n→∞
which is a contradiction.
Definition 1.11 (Bounded Functions). A function f : S → R is bounded if ∃B ≥ 0 such that
for all x ∈ S,
|f (x)| ≤ B.
Theorem 1.46. If f : [a, b] → R is continuous then f is bounded.
Proof : Suppose for the sake of contradiction that f : [a, b] → R is continuous and f is
unbounded. Then, ∀n ∈ N, ∃xn ∈ [a, b] such that |f (xn )| ≥ n. By the Bolzano-Weierstrass
theorem, ∃ a subsequence {xnk }k of {xn }n and an x ∈ R such that xnk → x. Since a ≤ xnk ≤ b
for all k, a ≤ x ≤ b. Given f is continuous at x by assumption,

f (x) = lim f (xnk ) =⇒ |f (x)| = lim |f (xnk )|.

k→∞ k→∞

34
Therefore, {|f (xnk )|} is bounded, and thus {nk } is bounded since nk ≤ |f (xnk )|. But by the
definition of a subsequence, we must have k ≤ nk for all k, contradicting the boundedness of
{nk }.
Definition 1.12. Let f : S → R. Then, f achieves an absolute minimum at c if ∀x ∈ S,
f (x) ≥ f (c). Similarly, f achieves an absolute maximum at d if ∀x ∈ S, f (x) ≤ f (d).
Theorem 1.47 (Min-Max Theorem). Let f : [a, b] → R. If f is continuous, then f achieves an
absolute maximum and absolute minimum.
Proof : We will prove this for the absolute maximum. If f is continuous, then f is bounded
by the previous theorem. Thus, the set

E = {f (x) | x ∈ [a, b]}

is bounded above. Let L = sup E. Then,

1. L is an upper bound for E, i.e.

∀x ∈ [a, b], f (x) ≤ L.

2. There exists a sequence {f (xn )}n with xn ∈ [a, b] such that f (xn ) → L.
By the Bolzano-Weierstrass theorem, there exists a subsequence {xnk }k of {xn } and d ∈ [a, b]
such that xnk → d as k → ∞. Hence,

f (d) = lim f (xnk ) = lim f (xn ) = L

k→∞ n→∞

by the continuity of f . Thus, f achieves an absolute maximum at d.

We leave the absolute minimum proof to the reader.
Consider (
1
x = 0, 1
f (x) = 2 .
x x ∈ (0, 1)
Here, f neither achieves an absolute maximum nor an absolute minimum on [0, 1].
If f : [a, b] → R is continuous, then f ([a, b]) ⊂ [f (c), f (d)] where f achieves an absolute
minimum at c and an absolute maximum at d.
Theorem 1.48. Let f : [a, b] → R. If f (a) < 0 and f (b) > 0, then ∃c ∈ (a, b) such that
f (c) = 0.
Proof : We prove this using a bisection method. Let a1 = a and b1 = b, and define a2 , b2
as follows: If f ((a1 + b1 )/2) ≥ 0, define a2 = a1 , b2 = a1 +b
2 . If f ((a1 + b1 )/2) < 0, define
1

a1 +b1
a2 = 2 and b2 = b1 . In general, if we know an , bn , we choose an+1 and bn+1 as follows:
If f ((an + bn )/2) ≥ 0, define an+1 = an , b2 n + 1 = an +b
2 . If f ((an + bn )/2) < 0, define
n

an +bn
an+1 = 2 and bn+1 = bn . Thus, we have:

35
 1. ∀n ∈ N, a ≤ an ≤ an+1 ≤ bn+1 ≤ bn ≤ b.
bn −an
2. ∀n ∈ N, bn+1 − an+1 = 2 .

3. ∀n ∈ N, f (an ) < 0 and f (bn ) ≥ 0.

By 1., {an } and {bn } are monotone increasing and monotone decreasing respectively, both of
which are bounded. Thus, ∃c, d ∈ [a, b] such that an → c and bn → d. By 2.,
bn−1 − an−1 1 1
bn − an = = (bn−2 − an−2 ) = · · · = n−1 (b − a).
2 4 2
Thus,
1
d − c = lim (bn − an ) = lim (b − a) = 0 =⇒ d = c.
n→∞ n→∞ 2n−1
Therefore, limn→∞ an = limn→∞ bn = c. By 3., f (c) = limn→∞ f (an ) ≤ 0 and f (c) =
limn→∞ f (bn ) ≥ 0. Therefore, f (c) = 0.
Theorem 1.49 (Bolzano IVT). Let f : [a, b] → R be continuous. If f (a) < f (b), and y ∈
(f (a), f (b)), ∃c ∈ (a, b) such that f (c) = y. If f (b) < f (a) and y ∈ (f (b), f (a)), ∃c ∈ (a, b) such
that f (c) = y.
Remark 1.50. This is known as the Intermediate Value Theorem or IVT for short.
Proof : Suppose f (a) < f (b). Let y ∈ (f (a), f (b)). Define g(x) = f (x)−y. Then, g : [a, b] →
R is continuous, g(a) = f (a)−y < 0 and g(b) = f (b)−y > 0. Therefore, by the previous theorem,
∃c ∈ (a, b) such that g(c) = 0. Therefore, ∃c ∈ (a, b) such that g(c) = f (c)−y = 0 =⇒ f (c) = y.
The other direction is analogous.
Theorem 1.51. Let f : [a, b] → R be continuous. let c ∈ [a, b] be where f achieves an absolute
minimum and d ∈ [a, b] be where f achieves an absolute maximum. Then,
f ([a, b]) = [f (c), f (d)].
In other words, every value between the absolute minimum value and the absolute maximum
value is achieved.
Proof : We know that f ([a, b]) ⊆ [f (c), f (d)]. Hence, we prove the other direction. By the
IVT applied to f : [c, d] → R,
[f (c), f (d)] ⊆ f ([c, d]) ⊆ f ([a, b]).
Therefore, f ([a, b]) = [f (c), f (d)].
Of course, Bolzano IVT is false if we assume f is not continuous (as can be seen by the
following diagram):
Theorem 1.52. The polynomial f (x) = x2021 + x2020 + 9.03x + 1 has at least one real root.
Proof : Notice that f (0) = 1 > 0 and f (−1) = −1 + 1 − 9.03 + 1 = −8.03 < 0. Thus, by
IVT, ∃c ∈ (−1, 0) such that f (c) = 0.

36
Uniform Continuity
Example 1.16. Consider the function f (x) = x1 . f is continuous on (0, 1).
n 2 o
Proof. Let ϵ > 0. Choose δ = min 2c , c2 ϵ . Suppose |x − c| < δ. Then, |x − c| < c
2 =⇒ |x| >
c − |x − c| > 2c . Thus, 1
|x| < 2c . Therefore,
 
1
 − 1  |x − c|
=
x c |xc|
δ
<
|x||c|
2
< 2δ
c
2 c2 ϵ
≤ 2 = ϵ.
c 2

As shown in the previous example. δ depended on both ϵ and c.

Definition 1.13 (Uniformly Continuous). Let f : S → R. Then, f is uniformly continuous on
S if ∀ϵ > 0, ∃δ = δ(ϵ) > 0 such that ∀x, c ∈ S,
|x − c| < δ =⇒ |f (x) − f (c)| < ϵ.
Remark 1.53. Thus, in the definition of uniform continuity, δ only depends on ϵ!
Example 1.17. The function f (x) = x2 is uniformly continuous on [0, 1].
Proof. Let ϵ > 0. Choose δ = 2ϵ . Then, if x, c ∈ [0, 1] then |x − c| < δ implies that
|x2 − c2 | = |x + c||x − c| ≤ 2|x − c| < 2δ = ϵ.

However, there are of course continuous functions that are not uniformly continuous. For
example, we will show that f (x) = x1 is not uniformly continuous on (0,1), but first we consider
the negation of the definition.
Let f : S → R. Then, f is not uniformly continuous on S if ∃ϵ0 > 0, ∀δ > 0 such that
∃x, c ∈ S with
|x − c| < δ and |f (x) − f (c)| ≥ ϵ0 .
Proof : Choose ϵ0 = 2 (in fact, any ϵ0 > 0 will show that x1 is not uniformly continuous on
(0, 1)). Then, let δ > 0. Choose c = min δ, 12 and x = 2c . Then, |x − c| = 2c ≤ 2δ < δ and

   
 1 1 2 1 1
 −  =  −  = ≥ 1 = 2.
x c   c c c 1
2

37
Theorem 1.54. Let f : [a, b] → R. Then, f is continuous if and only if f is uniformly
continuous.

Proof : ( ⇐= ) This direction is left as an exercise to the reader.

( =⇒ ) Suppose f is continuous and assume for the sake of contradiction that f is not
uniformly continuous. Then, ∃ϵ0 > 0 such that for all n ∈ N, ∃xn , cn ∈ [a, b] such that
1
|xn − cn | < and |f (xn ) − f (cn )| > ϵ0 .
n
By Bolzano-Weierstrass, ∃ a subsequence {xnk } of {xn } and x ∈ [a, b] such that limk→∞ xnk =
x. Similarly, by Bolzano-Weierstrass, ∃ a subsequence {cnk } of {cn } and c ∈ [a, b] such that
limk→∞ cnk = c. Note that subsequence {xnkj } of {xnk } satisfies limj→∞ xnkj = x.
Then,
1
|x − c| = lim |xnkj − cnkj | ≤ lim − 0.
j→∞ j→∞ nkj

Thus, x = c. But, since f is continuous at c,

0 = |f (c) − f (c)| = lim |f (xnkj ) − f (cnkj )| ≥ ϵ0 .

j→∞

This is a contradiction.

2 Derivative
Definition 2.1. Let I be an interval, let f : I → R, and let c ∈ I. We say that f is differentiable
at c if the limit
f (x) − f (c)
lim
x→c x−c
exists.

If f is differentiable at c, we write

f (x) − f (c)
f ′ (c) := lim .
x→c x−c
df
Furthermore, if f is differentiable at every c ∈ I, we write f ′ or dx for the function f ′ (x).

Example 2.1. Consider the function f (x) = ax + b. Then, for all c ∈ R, f ′ (c) = a.

Proof. This follows as

f (x) − f (c) ax + b − (ac + b) x−c

lim = lim = a lim = lim a = a.
x→c x−c x→c x−c x→c x − c x→c

38
Example 2.2. For all n ∈ N, if f (x) = αxn , then for all c ∈ R,

f ′ (c) = αncn−1 .

Proof. We note that for all n ∈ N,

n−1
X n−1
X n−1
X
(x − c) xn−1−j cj = xn−j cj − xn−1−j cj+1 .
j=0 j=0 j=0

Letting ℓ = j + 1, we obtain
n−1
X n−1
X n
X
(x − c) xn−1−j cj = xn−j cj − xn−ℓ cℓ
j=0 j=0 ℓ=1
n−0 0 n−n n
=x c −x c
n n
=x −c .

Therefore,
n−1 n−1
αxn − αcn X X
lim = α lim xn−1−j cj = α cn−1−j cj = αncn−1 .
x→c x−c x→c
j=0 j=0

Theorem 2.1. Let f : I → R, g : I → R be differentiable at c ∈ I. Then,

1. (Linearity) ∀α ∈ R, (αf + g)′ (c) = αf ′ (c) + g ′ (c) .

2. (Product rule) (f g)′ (c) = f ′ (c)g(c) + f (c)g ′ (c).

3. (Quotient rule) If g(x) ̸= 0 for all x ∈ I, then

 ′
f f ′ (c)g(c) − f (c)g ′ (c)
(c) = .
g (g(c))2

Proof :
1. We can compute this directly:
(αf + g)(x) − (αf + g)(c) f (x) − f (c) g(x) − g(c)
lim = lim α + = αf ′ (x) + g ′ (x).
x→c x−c x→c x−c x−c

2. We first write
f (x)g(x) − f (c)g(c) f (x) − f (c) g(x) − g(c)
= · g(x) + f (c) ·
x−c x−c x−c
and use the fact that limx→c g(x) = f (c).

39
 3. The quotient rule is left as an exercise to the reader.

Theorem 2.2 (Chain Rule). Let I1 , I2 be two intervals, g : I1 → I2 be differentiable at c ∈ I1 ,

and f : I2 → R differentiable at g(c). Then, f ◦ g : I1 → R is differentiable at c and

(f ◦ g)′ (c) = f ′ (g(c))g ′ (c).

Proof: Let h(x) = f (g(x)) and d = g(c). We want to prove that h′ (c) = f ′ (d)g ′ (c). Define
the following ( (
f (y)−f (d) g(x)−g(c)
y−x y ̸
= d x−c x= ̸ c
u(y) = ′
and v(y) = ′
.
f (d) y=d g (d) x=c
Then,
f (y) − f (d)
lim u(y) = lim = f ′ (d) = u(d).
y→d y→d y−d
Similarly,
g(x) − g(c)
lim v(x) = lim = g ′ (c) = v(c).
x→c x→c x−c
In other words, u is continuous at d and v is continuous at c. Now,

f (y) − f (d) = u(y)(y − d)

g(x) − g(c) = v(x)(x − c)
=⇒ h(x) − h(c) = f (g(x)) − f (d)
= u(g(x))(g(x) − g(c))
= u(g(x))v(x)(x − c).

Therefore,

h(x) − h(c)
lim = lim u(g(x))v(x)
x→c x−c x→c
= u(g(c))v(c)
= f ′ (g(c))g ′ (c).

3 Mean Value Theorem

Definition 3.1 (Relative Maximum/Minimum). Let S ⊂ R and f : S → R. Then, f has a
relative maximum at c ∈ S if ∃δ > 0 such that for all x ∈ S, |x − c| < δ =⇒ f (x) ≤ f (c). The
definition for relative maximum is analogous.

40
Theorem 3.1. If f : [a, b] → R has a relative max or min at c ∈ (a, b) and f is differentiable
at c, then
f ′ (c) = 0.
Proof : If f has a relative maximum at c ∈ (a, b) then ∃δ > 0 such that (c − δ, c + δ) ⊂ (a, b)
and ∀x ∈ (c − δ, c + δ), f (x) ≤ f (c). Let
δ
xn = c − ∈ (c − δ, c).
2n
Then, xn → c so
f (xn ) − f (c)
f ′ (c) = lim ≥ 0.
n→∞ xn − c
Now let
δ
yn = c + ∈ (c, c + δ).
2n
Then, yn → c so
f (yn ) − f (c)
f ′ (c) = lim ≤ 0.
n→∞ yn − c
Therefore, f ′ (c) = 0. The proof for relative minimum is similar and thus left to the reader.
Theorem 3.2 (Rolle). Let f : [a, b] → be continuous and differentiable on (a, b). If f (a) = f (b),
then ∃c ∈ (a, b) such that f ′ (c) = 0.
Proof : Let K = f (a) = f (b). Since f is continuous on [a, b], ∃c1 , c2 ∈ [a, b] such that f
achieves an absolute maximum at c1 and absolute minimum at c2 . If f (c1 ) > K =⇒ c1 ∈ (a, b).
Therefore, f ′ (c1 ) = 0 by the previous theorem. Similarly, if f (c2 ) < K, then c2 ∈ (a, b) =⇒
f ′ (c2 ) = 0. If
f (c1 ) ≤ K ≤ f (c2 ) =⇒ f (x) = K ∀x ∈ [a, b] =⇒ f ′ (c) − 0 for any c ∈ (a, b).

Theorem 3.3 (Mean Value Theorem). Let f : [a, b] → R be continuous, and let f be differen-
tiable on (a, b). Then, ∃c ∈ (a, b) such that
f (b) − f (a) = f ′ (c)(b − a).
Remark 3.4. The Mean Value Theorem is sometimes denoted MVT.
Proof : Define g : [a, b] → R with
f (b) − f (a)
g(x) = f (x) − f (b) + (b − x).
b−a
Then, g(a) = g(b) = 0. Thus, by Rolle’s theorem, ∃c ∈ (a, b) with g ′ (c) = 0, and hence
f (b) − f (a)
0 = f ′ (c) − .
b−a

We now look at some useful applications of the MVT.

41
Theorem 3.5. If f : I → R is differentiable and f ′ (x) = 0 for all x ∈ I, then f is constant.

Proof : Let a, b ∈ I with a < b. Then, f is continuous on [a, b] and differentiable on (a, b).
Therefore, ∃c ∈ (a, b) such that f (b) − f (a) = (b − a)f ′ (c) = 0. Hence, f (b) = f (a) for all a, b ∈ I
such that a < b.

Theorem 3.6. Let f : I → R be differentiable. Then,

1. f is increasing if and only if f ′ (x) ≥ 0 for all x ∈ I and

2. f is decreasing if and only if f ′ (x) ≤ 0 for all x ∈ I.

Proof :

1. ( ⇐= ) Suppose f ′ (x) ≥ 0 for all x ∈ I. Let a, b ∈ I with a < b. Then, by MVT, ∃c ∈ (a, b)
such that
f (b) − f (a) = (b − a)f ′ (c) ≥ 0 =⇒ f (a) ≤ f (b).

( =⇒ ) Suppose f is increasing. Let c ∈ I and let {xn } be a sequence in I such that xn → c

such that ∀n, xn < c. Then, for all n, f (xn ) − f (c) ≤ 0 =⇒ f (xxnn)−f
−c
(c)
≥ 0. Therefore,

f (xn ) − f (c)
f ′ (c) = lim ≥ 0.
n→∞ xn − c

Now let {xn } be a sequence in I such that xn → c such that ∀n, xn > c. Then, for all n,
f (xn ) − f (c) ≥ 0 =⇒ f (xxnn)−f
−c
(c)
≥ 0. Therefore,

f (xn ) − f (c)
f ′ (c) = lim ≥ 0.
n→∞ xn − c

Hence, in either case, f ′ (c) ≥ 0.

2. Notice that f is decreasing if and only if −f is increasing, and apply 1. to −f .

4 Taylor’s Theorem
Remark 4.1. Taylor’s theorem is essentially the Mean Value Theorem for higher order deriva-
tives.

Definition 4.1 (n-times Differentiable). We say f : I → R is n-times differentiable on J ⊂ I

if f ′ , f ′′ , . . . , f (n) exist at every point in J.

We denote the n-th derivative of f as f (n) (as used above).

42
Theorem 4.2 (Taylor). Suppose f : [a, b] → R is continuous and has n continuous derivatives
on [a, b] such that f (n+1) exists on (a, b). Given x0 , x ∈ [a, b], there exists a c ∈ (x0 , x) such that
n
X 1 (k) f (n+1) (c)
f (x) = f (x0 )(x − x0 )k + (x − x0 )n+1 .
k! (n + 1)!
k=0

Denote the large sum as Pn (x) and the last term with Rn (x).
Definition 4.2. Pn (x) is the n-th order Taylor polynomial for f at x0 . Rn (x) is the n-th order
remainder term.
We will essentially apply the Mean Value Theorem n + 1 times to prove Taylor’s theorem.
Proof : Let x, x0 ∈ [a, b]. If x = x0 then any c will satisfy the theorem. So, suppose x ̸= x0 .
Let Mx,x0 = f(x−x
(x)−Pn (x)
0)
n+1 . Hence,

f (x) = Pn (x) + Mx,x0 (x − x0 )n+1 .

Now, for 0 ≤ k ≤ n,
f k (x0 ) = Pn(k) (x0 ).
Let g(s) = f (s) − Pn (s) − Mx,x0 (s − x0 )n+1 (notably, n + 1-times differentiable. Then,
g(x0 ) = f (x0 ) − Pn (x0 ) − Mx,x0 (x0 − x0 )n+1 = 0
g ′ (x0 ) = f ′ (x0 ) − Pn′ (x0 ) − Mx,x0 (n + 1)(x0 − x0 )n = 0
..
.
g (n) (x0 ) = f (n) (x0 ) − Pn(n) (x0 ) − Mx,x0 (n + 1)!(x0 − x0 ) = 0.
Now, notice that g(x) = 0 and g(x0 ) = 0. By the MVT, there exists an x1 ∈ (x0 , x) such that
g ′ (x1 ) = 0. Thus, g ′ (x0 ) = 0 and g ′ (x1 ) = 0. Therefore, ∃x2 ∈ (x0 , x1 ) such that g ′′ (x2 ) = 0.
Continuing, we analogously find xn between x0 and xn−1 such that g (n) (xn ) = 0. Then, finally,
g (n) (x0 ) = 0 and g (n) (xn ) = 0 implies ∃c ∈ (x0 , xn ) (and thus between x0 and x) such that
g (n+1) (c) = 0.
We may compute
dn+1
Mx,x0 (s − x0 )n+1 = Mx,x0 (n + 1)!.
ds(n+1)
(n+1)
Furthermore, Pn (c) = 0 since Pn is a polynomial of degree n. Thereforee,
0 = g (n+1) (c) = f (n+1) (c) − Mx,x0 (n + 1)!,
f (n+1)! (c)
which implies Mx,x0 = (n+1)! and thus

f (n+1) (c)
f (x) = Pn (x) + (x − x0 )n+1 .
(n + 1)!

43
Theorem 4.3 (Second Derivative Test). Suppose f : (a, b) → R has two continuous derivatives.
If x0 ∈ (a, b) such that f ′ (x0 ) = 0 and f ′′ (x0 ) > 0, then f has a strict relative minimum at x0 .

Proof : Since f ′′ is continuous at x0 and

lim f ′′ (c) = f ′′ (x0 ) > 0,

c→x0

we have that ∃δ > 0 such that for all c ∈ (x0 − δ, x0 + δ), f ′′ (c) > 0. Let x ∈ (x0 − δ, x0 + δ)
(as you will show in your homework). Then, by Taylor’s theorem, ∃c between x and x0 (hence
c ∈ (x0 − δ, x0 + δ)) such that

f ′′ (c)
f (x) = f (x0 ) + (x − x0 )2 ≥ f (x0 ),
2
with f (x) > f (x0 ) if x ̸= x0 .

The Riemann Integral

Definition 4.3. We define the set

C([a, b]) := {f : [a, b] → R | f is continuous}.

Definition 4.4 (Partition). A partition x of [a, b] is a finite set

x = {a = x0 < x1 < · · · < xn = b}.

The norm of x, denoted ∥x∥, is the number

∥x∥ := max{x1 − x0 , x2 − x1 , . . . , xn − xn−1 }.

Definition 4.5 (Tag). If x is a partition, a tag of x is a finite set ξ = {ξ1 , . . . , ξn } such that

a = x0 ≤ ξ1 ≤ x1 ≤ ξ2 ≤ x2 ≤ · · · ≤ xn−1 ≤ ξn ≤ xn = b.

The pair (x, ξ) is referred to as a tagged partition.

Example 4.1. Consider the tagged partition (x, ξ) = ({1, 3/2, 2, 3}, {5/4, 7/4, 5/2}. Then,

∥x∥ = max{3/2 − 1, 2 − 3/2, 3 − 2} = 1.

Definition 4.6 (Riemann sum). The Riemann sum of f corresponding to (x, ξ) is the number
n
X
Sf (x, ξ) := f (ξk )(xk − xk−1 ).
k=1

44
Theorem 4.4 (Riemann Integral). Let f ∈ C([a, b]). Then, there exists a unique number de-
Rb
noted a f (x) dx ∈ R with the following property: for all sequences of tagged partitions {(xr , ξ r )}
such that ∥xr ∥ → 0, we have
Z b
r r
lim Sf (x , ξ ) = f (x) dx.
r→∞ a

Remark 4.5. Uniqueness follows immediately R b from uniqueness of limits of sequences of real
numbers. All we need to prove is existence of a f (x) dx.

Before giving the proof of the theorem, we first prove some useful facts. Note that we number
the next few theorems to use them in our proof of the above theorem.

Definition 4.7 (Modulus of Continuity). For f ∈ C([a, b]), η > 0, we define the modulus of
continuous
wf (η) = sup{|f (x) − f (y)| | |x − y| ≤ η}.

Remark 4.6. Note that we number the following theorems to reference later.

Theorem 4.7 (Theorem I). For all f ∈ C([a, b]), limη→0 wf (η) = 0. In other words, for all
ϵ > 0, there exists a δ > 0 such that ∀η < δ, wf (η) < ϵ.

Proof : Let ϵ > 0. Since f ∈ C([a, b]), f is uniformly continuous on [a, b] (as continuity on
a bounded interval is equivalent to uniform continuity on the same interval). Thus, ∃δ0 > 0
such that if |x − y| < δ9 , then |f (x) − f (y)| < ϵ/2. Choose δ = δ0 and let η < δ. Then, if
|x − y| ≤ η < δ = δ0 , then
ϵ
|f (x) − f (y)| < .
2
Therefore, ϵ/2 is an upper bound for {|f (x) − f (y)| | |x − y| ≤ η}. Hence,

wf (η) ≤ ϵ/2 < ϵ.

Theorem 4.8 (Theorem II). If (x, ξ) and (x′ , ξ ′ ) are tagged partitions of [a, b] such that x ⊂ x′ ,
then if f ∈ C([a, b]) then

|Sf (x, ξ) − Sf (x′ , ξ ′ )| ≤ wf (∥x∥)(b − a).

Definition 4.8 (Refinement). If (x, ξ) and (x′ , ξ ′ ) are tagged partitions of [a, b] such that x ⊂ x′ ,
we say x′ refinement of x.

Remark 4.9. Refinements of x are obtained by adding more partition points.

45
 Proof : For k = 1, . . . , n, let
y(k) = {xk−1 = x′ℓ , x′ℓ+1 , . . . , x′m = xk }
′ ′ ′
η(k) = {ξℓ+1 , ξℓ+2 , . . . , ξm }.
Then,
 

Xm 
′ ′ ′
 
|f (ξk )(xk − xk−1 ) − Sf (y(k), η(k))| = f (ξk ) − f (ξj )(xj − xj−1 )
 j=ℓ+1 
 
 m 
 X ′ ′ ′

=  (f (ξk ) − f (ξj ))(xj − xj−1 )
j=ℓ+1 
Pm ′
since j=1 xj − x′j−1 = xm − x′ℓ = xk − xk−1 . Hence,
m
X
|f (ξk )(xk − xk−1 ) − Sf (y(k), η(k))| ≤ |f (ξk ) − f (ξj′ )|(x′j − x′j−1 )
j=ℓ+1
Xm
≤ wf (|xk − xk−1 |)(x′j − x′j−1 )
j=ℓ+1

≤ wf (∥x∥(xk − xk−1 ).
Thus,
 
Xm 
|Sf (x, ξ) − Sf (x′ , ξ ′ )| =  (f (ξk )(xk − xk−1 ) − Sf (y(k), η(k)))
 
 
k=1
m
X
≤ |f (ξk )(xk − xk−1 ) − Sf (y(k), η(k))|
k=1
n
X
≤ wf (∥x∥) xk − x1 = wf (∥x∥)(b − a).
k=1

Theorem 4.10 (Theorem III). If (x, ξ) and (x′ , ξ ′ ) are any two tagged partitions of [a, b] and
f ∈ C([a, b]), then
|Sf (x, ξ) − Sf (x′ , ξ ′ )| ≤ (wf (∥x) + wf (∥x′ ∥))(b − a).
Proof : Let x′′ = x∪x′ (i.e. a common refinement), and ξ ′′ be a tag of x′′ . Then, by Theorem
II,
|Sf (x, ξ) − Sf (x′ , ξ ′ )| ≤ |Sf (x, ξ) − Sf (x′′ , ξ ′′ )| − |Sf (x′ , ξ ′ ) − Sf (x′′ , ξ ′′ )|
≤ wf (∥x∥)(b − a) + wf (∥x′ ∥)(b − a).

46
 We now have the theorems necessary to prove the big theorem at the beginning of this
section.
Proof : Let {y(r), ζ(r)}r be a sequence of tagged partitions with ∥y(r)∥ → 0 as r → ∞.
Claim 1: {Sf (y(r), ζ(r)}r is a Cauchy sequence. Proof: Let ϵ > 0. By Theorem I, ∃δ > 0
such that ∀η < δ,
ϵ
wf (η) < .
2(b − a)
Since ∥y(r)∥ → 0, ∃M0 ∈ N such that ∀r ≥ M0 ,

∥y(r)∥ < δ.

Choose M = M0 . Then, if r, s ≥ M = M0 ,

|Sf (y(r), ζ(r)) − Sf (y(s), ζ(r))| ≤ (w(∥y(r)∥ + w(∥y(s)∥))(b − a)

by Theorem III. Hence, by the above inequalities, it follows that

 
ϵ ϵ
|Sf (y(r), ζ(r)) − Sf (y(s), ζ(r))| < + (b − a) = ϵ.
2(b − a) 2(b − a)
This proves Claim 1. Let L = limr→∞ Sf (y(r), ζ(r)), which exists as Cauchy sequences of real
numbers are convergent sequences.
Claim 2: Let {(x(r), ξ(r))}r be any sequence of partitions with ∥x(r)∥ → 0. Then,

lim Sf (x(r), ξ(r)) = L.

r→∞

With (y(r), ζ(r)) as before, we have by the Triangle Inequality and Theorem III that

|Sf (x(r), ξ(r)) − L| ≤ |Sf (x(r), ξ(r)) − Sf (y(r), ζ(r))| + |Sf (y(r), ζ(r)) − L|
≤ (wf (∥x(r)∥) + wf (∥y(r)∥))(b − a) + |Sf (y(r), ζ(r)) − L| → 0

as r → ∞ by Theorem I and by the definition of L. Thus, by the Squeeze Theorem,

|Sf (x(r), ξ(r)) − L| → 0 as r → 0.

5 Properties of the Riemann Integral

Rb Rb
We will often abbreviate a f (x) dx to a f.
Theorem 5.1 (Linearity). If f, g ∈ C([a, b]) and α ∈ R, then
Z b Z b Z b
(αf + g) = α f+ g.
a a a

47
 Proof : Let {(x(r), ξ(r))}r be a sequence of tagged partitions such that ∥x(r)∥ → 0. Then,
Sαf +g (x(r), ξ(r)) = αSf (x(r), ξ(r)) + Sg (x(r), ξ(r)).
Therefore,
Z b
(αf + g) = lim Sαf +g (x(r), ξ(r))
a r→∞

= lim (αSf (x(r), ξ(r)) + Sg (x(r), ξ(r))

r→∞
Z b Z b
=α f+ g.
a a

Theorem 5.2 (Additivity). If f ∈ C([a, b]) and a < c < b, then

Z b Z c Z b
f= f+ f.
a a c

Proof : Let {(y(r), ζ(r))}r and {(z(r), η(r))}r be tagged partitions of [a, c] and [c, b] respec-
tively such that ∥y(r)∥ → 0 and ∥z(r)∥ → 0. Define

x(r) = y(r) ∪ z(r)

ξ(r) = ζ(r) ∪ η(r),
a sequence of tagged partitions of [a, b]. Then,
∥x(r)∥ ≤ ∥y(r)∥ + ∥z(r)∥ → 0.
Thus,
Z b
f = lim Sf (x(r), ξ(r))
a t→∞

= lim (Sf (y(r), ζ(r)) + Sf (z(r), η(r)))

r→∞
Z c Z b
= f+ f.
a c

Theorem 5.3. Let f ∈ C([a, b]), and

mf = inf{f (x) | x ∈ [a, b]} ∈ R
Mf = sup{f (x) | x ∈ [a, b]} ∈ R.
Then,
Z b
mf (b − a) ≤ f ≤ Mf (b − a).
a

48
 Proof : Let {(x(r), ξ(r))}r be a sequence of tagged partitions with ∥x(r)∥ → 0. Then,
n
X n
X
Sf (x(r), ξ(r)) = f (ξk (r))(xk (r) − xk−1 (r)) ≥ mf (xk (r) − xk−1 (r)) = mf (b − a).
k=1 k=1

Similarly,
n
X n
X
Sf (x(r), ξ(r)) = f (ξk (r))(xk (r) − xk−1 (r)) ≤ Mf (xk (r) − xk−1 (r)) = Mf (b − a).
k=1 k=1

Therefore, for all r,

Z b
mf (b − a) ≤ Sf (x(r), ξ(r)) ≤ Mf (b − a) =⇒ mf (b − a) ≤ f ≤ Mf (b − a).
a

Theorem 5.4. Suppose f ∈ C([a, b]) and g ∈ C([a, b]).

1. If ∀x ∈ [a, b] f (x) ≤ g(x), then
Z b Z b
f≤ g.
a a
Rb Rb
2. (Triangle Inequality for integrals): | a f| ≤ a |f |.
Proof :
1. Let {(x(r), ξ(r)}r be a sequence of tagged partitions such that ∥x(r)∥ → 0. Then, for all
r ∈ N,
n
X
Sf (x(r), ξ(r)) = f (ξj (r))(xj (r) − xj−1 (r))
j=1
Xn
≤ g(ξj (r))(xj (r) − xj−1 (r))
j=1

= Sg (x(r), ξ(r)).
Then, letting r → ∞, we get that
Z b Z b
f≤ g.
a a

2. Notice that ±f (x) ≤ |f (x)| for all x, and thus

Z b Z b Z b Z b Z b
± f≤ |f | =⇒ − f≤ f≤ |f |.
a a a a a
Rb Rb
Therefore, | a f| ≤ a |f |.

49
Remark 5.5. There are some conventions that are worth noting:
Ra Rb
1. a f := 0. This is consistent with our definitions and theorems thus far as limb→a | a f | =
0.
Rb Ra
2. a f = − b f.

6 Fundamental Theorem of Calculus

Theorem 6.1 (Fundamental Theorem of Calculus). Suppose f ∈ C([a, b]).

1. If F : [a, b] → R is differentiable and F ′ = f , then

Z b
f = F (b) − F (a).
a

Rx
2. The function G(x) := a f is differentiable on [a, b] and
(
G′ = f
.
G(a) = 0

Remark 6.2. We sometimes abbreviate the Fundamental Theorem of Calculus to FTC.

Proof :

1. Let {(x(r))}r be a sequence of partitions with ∥x∥ → 0. Then, by the Mean Value Theorem,
∀r∀j, there exists a ξj (r) ∈ [xj−1 (r), xj (r)] such that

F (xj (r)) − F (xj−1 (r)) = F ′ (ξj (r))(xj (r) − xj−1 (r)) = f (ξj (r))(xj (r) − xj−1 (r)).

Thus,
Z b n(r)
X
f = lim f (ξj (r))(xj (r) − xj−1 (r))
a r→∞
j=1
n(r)
X
= lim F (xj (r)) − F (xj−1 (r))
r→∞
j=1

= lim (F (b) − F (a)) = F (b) − F (a).

r→∞

50
 2. Let c ∈ [a, b]. We wish to show that
Rx Rc
a f− af
lim = f (c).
x→c x−c
Let ϵ > 0. Then, since f is continuous at c, ∃δ0 > 0 such that

|t − c| < δ0 =⇒ |f (t) − f (c)| < ϵ/2.

Choose δ = δ0 . Suppose 0 < x − c < δ. If t ∈ [c, x], then

|t − c| ≤ |x − c| < δ = δ0 .

Thus,
x Z t Z c
 Z   
 1   1 1 

x − c f (t) dt − f (c) =  f (t) dt − f (c) dt
c x−c c x−c x
Z x 
1 
 
= (f (t) − f (c)) dt
x−c c 
Z x
1
≤ |f (t) − f (c)| dt
x−c c
Z x
1
≤ ϵ/2 dt
x−c c
1 ϵ ϵ
= · (x − c) = .
x−c 2 2
A similar argument holds for 0 < c − x < δ. Thus,
R x
 a f − ac f
R 
 ϵ
0 < |x − c| < δ =⇒   − f (c) ≤ < ϵ.
x−c 2

Theorem 6.3 (Integration by Parts). Suppose f, g ∈ C([a, b]) and f ′ , g ′ ∈ C([a, b]). Then,
Z b
f ′ g = (f (b)g(b) − f (a)g(a)) − intba f g ′ .
a

Proof : We have
(f g)′ = f ′ g + f g ′ .
Therefore, by the Fundamental Theorem of Calculus,
Z b Z b
f (b)g(b) − f (a)g(a) = f ′g + f g′.
a a

51
Remark 6.4. We sometimes abbreviate Integration By Parts as IBP.

Theorem 6.5 (Change of Variables). Let φ : [a, b] → [c, d] be continuously differentiable with
φ′ > 0 on [a, b], φ(a) = c, and φ(b) = d. Then,
Z d Z b
f (u) du = f (φ(x))φ′ (x) dx.
c a

Proof : Let F : [a, b] → R such that F ′ = f . Then,

F (φ(x))′ = f (φ(x)).

Hence, by the FTC,

Z b Z b
′
f (φ(x))φ (x) dx = F (φ(x))′ dx
a a
= F (φ(b)) − F (φ(a))
= F (d) − F (c).

Furthermore, by the FTC,

Z d Z d
f (u) du = F (u)′ du = F (d) − F (c).
c c

Sequences of Function
Power Series

Remark 6.6. Power series motivate the general discussion of sequences of functions.

Definition 6.1 (Power series). A power series about x0 is a series of the form
∞
X
am (x − x0 )m .
m=0

Theorem 6.7. Suppose

R = lim |am |1/n
m→∞

exists, and let (

1
R R>0
p=
∞ R = 0.
am (x − x0 )m converges absolutely if |x − x0 | < p and diverges if |x − x0 | > p.
P
Then,

52
Definition 6.2 (Radius of Convergence). In the above theorem, we define p to be the radius of
convergence.

Proof : We have
lim |am (x − x0 )m |1/m = R|x − x0 |,
n→∞

and the theorem

P follows bym the Root test.
Suppose am (x − x0 ) is a power series with radius of convergence p. Furthermore, define
f : (x0 − p, x0 + p) → R such that
∞
X
f (x) := am (x − x0 )m .
m=0

Then, f is a limit of a sequence of functions

f (x) = lim fn (x),

n→∞

for x ∈ (x0 − p, x0 + p) and where

n
X
fn (x) = am (x − x0 )m .
m=0

Example 6.1. For example, we have

∞
1 X
f (x) = = xm .
1−x
m=0

This concept begs a number of questions:

1. Is f continuous?

2. Is f differentiable, and does f ′ = limn→∞ fn′ ?

3. If 1. is true, does
Z b Z b
f = lim fn ?
a n→∞ a

Pointwise and Uniform Convergence

Definition 6.3 (Pointwise Convergence). For n ∈ N, let fn : S → R. Let f : S → R. We say
that {fn } converges pointwise to f if for all x ∈ S,

lim fn (x) = f (x).

n→∞

53
Let’s consider some examples.
1. Let fn (x) = xn on [0,1]. Then,
(
0 x ∈ [0, 1)
lim fn (x) = .
n→∞ 1 x=1

Thus, {fn } converges to the above pointwise function. Hence, notice that a sequence of
continuous functions may not converge pointwise to a continuous function!

2. Let fn (x) = nm=0 xm for x ∈ (−1, 1). Then,

P

n
X 1
lim fn (x) = lim xm = .
n→∞ n→∞ 1−x
m=0

Hence, pointwise, this sequence converges to its power series (see the above example).

3. Let fn (x) : [0, 1] → R be defined by

  1
2 x ∈ 0, 2n
4n x

 1 1
fn (x) = 4n − 4n2 x x ∈ 2n , .
  1 n
0 x ∈ n, 1


1
Then, limn→∞ fn (0) = limn→∞ 0 = 0. Let x ∈ (0, 1]. Let N ∈ N such that N < x. Then,
for all n ≥ N ,
fn (x) = 0.
Therefore,
{fn (x)} = f1 (x), . . . , fN −1 (x), 0, 0, 0, . . . .
Hence, limn→∞ fn (x) = 0 for all x ∈ [0, 1]. Thus, {fn } converges pointwise to f (x) = 0 on
[0, 1].
Definition 6.4 (Uniform Convergence). For n ∈ N, let fn : S → R, and let f : S → R. Then,
we say fn converges to f uniformly or converges uniformly to f if ∀ϵ > 0 ∃M ∈ N such that
for all n ≥ M ∀x ∈ S,
|fn (x) − f (x)| < ϵ
Theorem 6.8. If fn : S → R, f : S → R, and fn → f uniformly, then fn → f pointwise.
Proof : Let c ∈ S and let ϵ > 0. Then, fn → f uniformly implies that there exists M0 ∈ N
such that for all n ≥ M, ∀x ∈ S, |fn (x) − f (x)| < ϵ. Choose M = M0 . Then, ∀n ≥ M ,

|fn (c) − f (c)| < ϵ.

Thus, limn→∞ fn (c) = f (c) for all c ∈ S, and therefore fn → f pointwise.

54
 (
0 x ∈ [0, 1)
Theorem 6.9. Let fn (x) = xn , and let f (x) = .
1 x=1

1. ∀0 < b < 1, fn → f uniformly on [0, b].

2. fn does not converge to f uniformly on [0, 1].

Proof :

1. Let ϵ > 0. Since b ∈ (0, 1), bn → 0. Therefore, ∃M0 ∈ N such that for all n ≥ M0 , bn < ϵ.
Choose M = M0 . Then, ∀n ≥ M, ∀x ∈ [0, b],

|fn (x) − f (x)| = |fn (x)| = xn ≤ bn < ϵ.

Thus, fn → f uniformly on [0, b].

Before proving the other part, we first note the following negation:
fn : S → R does not converge to f : S → R uniformly if ∃ϵ0 > 0 such that ∀M ∈ N, ∃n ≥ M
and ∃x ∈ S with |fn (x) − f (x)| ≥ ϵ0 .

1
2. Hence, for our example, choose ϵ0 = 14 . Let M ∈ N and choose x = 1
2
M
∈ (0, 1). Thus,

1
|fM (x) − f (x)| = fM (x) = > ϵ0 .
2

Theorem 6.10 (Weierstrass M-test). let fj : S → R and suppose ∃Mj > 0 such that

1. ∀x ∈ S, |fj (x)| ≤ Mj .
P∞
2. j=1 Mj converges.

Then,
P∞
1. ∀x ∈ S, j=1 fj (x) converges absolutely.
P∞
2. Let f (x) = j=1 fj (x) for x ∈ S. Then,
n
X
fj → f uniformly on S.
j=1

Proof :

1. The first part follows from a), b), and the Comparison Test.

55
 P
2. Let ϵ > 0. Since Mj converges, ∃N0 ∈ N such that ∀n ≥ N0 ,
 
X∞ ∞ n 
X X 
Mj =  Mj − Mj  < ϵ.
j=n+1  j=1 j=1 

Choose N = N0 . Then, for all n ≥ N and ∀x ∈ S,

   
n
  ∞
 X   X 
 
f (x) − fj (x) =  fj (x)

 j=1  j=n+1 
∞
X
≤ |fj (x)|
j=n+1
X∞
≤ Mj < ϵ.
j=n+1
Pn
Thus, j=1 fj → f uniformly on S.

Interchange of Limits
Remark 6.11. In general, limits cannot be interchanged.
Example 6.2. For instance, consider the following example:
n/k 0
lim lim = lim =0
n→∞ k→∞ n/k + 1 n→∞ 0+1
n/k
lim lim = lim 1 = 1.
k→∞ n→∞ n/k + 1 k→∞

Hence, we ask three questions about interchanging limits:

1. If fn : S → R, fn continuous and fn → f pointwise or uniform, then is f continuous?

2. If fn : [a, b] → R, fn differentiable, and fn → f with fn′ → g, then is f differentiable and

does g = f ′ ?

3. If fn : [a, b] → R, with fn and f continuous such that fn → f , then does

Z b Z b
fn = f?
a a

The answer to the above questions are all yes, if the convergence is uniform.
Hence, we ask three questions about interchanging limits:

56
 1. If fn : S → R, fn continuous and fn → f pointwise or uniform, then is f continuous?

2. If fn : [a, b] → R, fn differentiable, and fn → f with fn′ → g, then is f differentiable and

does g = f ′ ?

3. If fn : [a, b] → R, with fn and f continuous such that fn → f , then does

Z b Z b
fn = f?
a a

The answer to the above questions are all no, if the convergence is pointwise as seen by the
following counterexamples:
(
0 x ∈ [0, 1)
1. Let fn (x) = xn on [0, 1] is continuous ∀n. As we noted earlier, fn (x) → f (x) = .
1 x=1
Notice that f is not continuous.
xn+1
2. Let fn (x) = n+1 on [0, 1]. Then, fn → 0 pointwise on [0, 1]. However,
(
0 x ∈ [0, 1)
fn′ (x) → g(x) = .
1 x=1

Thus, g(x) ̸= (0)′ = 0 at x = 1.

3. Consider the functions   1

2 x ∈ 0, 2n
4n x

 1 1
fn (x) = 4n − 4n2 x x ∈ 2n ,
  1 n
0 x ∈ n, 1


as described in the previous lecture. Then, fn (x) → 0 pointwise on [0,1] as we showed last
time. However,
Z 1 Z 1
1 1
fn = (base)(height) = · 2n = 1 ̸→ 0 = 0.
0 2 2n 0

We now prove that the answer to the three questions above is yes if convergence is uniform.

Theorem 6.12. If fn : S → R is continuous for all n, f : S → R, and fn → f uniformly, then

f is continuous.

Proof : Let c ∈ S and let ϵ > 0. Since fn → f uniformly, ∃M ∈ N such that ∀n ≥ M ,

∀y ∈ S,
ϵ
|fn (y) − f (y)| < .
3

57
Since fM : S → R is continuous, ∃δ0 > 0 such that ∀|x − c| < δ0 ,
ϵ
|fM (x) − fM (c)| < .
3
Choose δ = δ0 . If |x − c| < δ, then

|f (x) − f (c)| ≤ |f (x) + fM (x)| + |fM (c) − f (c)| + |fM (x) − fM (c)|
ϵ ϵ ϵ
< + + = ϵ.
3 3 3

Theorem 6.13. If fn : [a, b] → R is continuous for all n, f : [a, b] → R and fn → f uniformly,

then Z b Z b
fn → f.
a a

Proof : Let ϵ > 0. Since fn → f uniform, ∃M0 ∈ N such that ∀n ≥ M0 , ∀x ∈ [a, b],
ϵ
|fn (x) − f (x)| < .
b−a
Then, for all n ≥ M = M0 , we have
Z b Z b  Z b Z b
  ϵ
 fn − f ≤
 |fn − f | < = ϵ.

a a a a b−a

Remark 6.14. Notationally, this states that

Z b Z b Z b
lim fn = lim fn = f.
n→∞ a a n→∞ a

Theorem 6.15. If fn : [a, b] → R is continuously differentiable, f : [a, b] → R, g : [a, b] → R,

and

fn → f pointwise,
fn′ → g uniformly,

then f is continuously differentiable and g = f ′ .

Proof : By the FTC, ∀n∀x ∈ [a, b],

Z x
fn (x) − fn (a) = fn′ .
a

58
Thus, by the previous two theorems,

f (x) − f (a) = lim (fn (x) − fn (a))

n→∞
Z x
= lim fn′
n→∞ a
Z x
= g.
a
Rx Rx
Therefore, f (x) = f (a) + a g. Thus, by the FTC, f is differentiable and f ′ = ( a g)′ = g.
We now return back to our study of power series, answering some questions we asked at the
beginning of Lecture 23.

Theorem 6.16. Let ∞ a (x − x0 )j be a power series of radius of convergence p ∈ (0, ∞].

P
P∞ j=0 j
Then, ∀r ∈ (0, p), j=0 aj (x − x0 )j converges uniformly on [x0 − r, x0 + r].

Proof : Let r ∈ [0, p). Then, ∀j ∈ N ∪ {0}, ∀x ∈ [x0 − r, x0 + r],

|aj (x − x0 )j | ≤ |aj |rj =: Mj .

Now, (
r
1/j 1/j p p<∞
lim j → ∞Mj = lim |aj | r=
j→∞ 0 p=∞

since p−1 = limj→∞ |aj |1/j . Since r < p, we have

∞
1/j
X
lim Mj < 1 =⇒ Mj converges.
j→∞
j=0
P∞
By the Weierstrass M-test, it follows that j=0 aj (x − x0 )j converges uniformly on [x0 − r, x0 +
r].

Theorem 6.17. Let ∞ j

P
j=0 aj (x − x0 ) be a power series with radius of convergence p ∈ (0, ∞].
Then,

1. ∀c ∈ (x0 − p, x0 + p), ∞ j
P
j=0 aj (x − x0 ) is differentiable at c and

∞ ∞
d X X
aj (x − x0 )j = jaj (x − x0 )j−1 .
dx
j=0 j=0

2. ∀a, b such that x0 − p < a < b < x0 + p,

∞
Z bX ∞
(b − x0 )j+1 (a − x0 )j+1
X  
j
aJ (x − x0 ) dx = aj − .
a j=0 j+1 j+1
j=0

59
Remark 6.18. Since
 (j+1)/j
lim ((j + 1)|aj+1 |)1/j = lim ((j + 1)|aj+1 |1/(j+1) = lim |ak |1/k = p,
j→∞ j→∞ k→∞

aj (x − x0 )j is infinitely differentiable and

P
1. implies

dk X
 
j 

k!ak = k
aj (x − x0 ) x=x0 .
dx

60