Real Analysis

Arup Pal

July–November, 2004
(Course given for M. Stat. Ist year)

Indian Statistical Institute, Delhi Centre

7, SJSS Marg, New Delhi–110 016, India
1
Contents
1 Prerequisites 1
1.1 Sets and mappings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Binary operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Real Numbers 2
2.1 Rational powers of positive numbers . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Arbitrary powers of positive numbers . . . . . . . . . . . . . . . . . . . . . . . . 7

3 Countable Sets 8
3.1 Decimal Expansion of a Real Number . . . . . . . . . . . . . . . . . . . . . . . 11

4 Topological Notions 14
4.1 Limit point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2 Extended Real Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Lim sup and lim inf . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

5 Sequences and Series 20

5.1 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.2 Monotone sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.3 Cauchy sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.4 Limit point of a sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.5 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.6 Tests for convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.7 Power series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.8 Product of two series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

6 Limit and continuity 33

6.1 Limit of a function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.2 Computation of some limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.3 Limits via limits of sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.4 Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.5 Properties of continuous functions . . . . . . . . . . . . . . . . . . . . . . . . . 38

7 Some important functions 40

7.1 Trigonometric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.2 Exponential function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2
8 Differentiability 43
8.1 Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
8.2 Mean value theorem and applications . . . . . . . . . . . . . . . . . . . . . . . . 45
8.3 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

9 Integration 49
9.1 Riemann integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
9.2 Improper Riemann integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

10 Sequences and series of functions 61

10.1 Double sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

11 Multivariable calculus 68

12 Misc 69
12.1 Proof of irrationality of π . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

0
 Week 1

1 Prerequisites
1.1 Sets and mappings
1.2 Relations
Let S be a set. A relation R on S is a subset of S × S. It is customery to write x R y instead
of (x, y) ∈ R while talking about relations.
A relation R is said to be

1. reflexive if x R x for all x ∈ S .

2. symmetric if x R y implies y R x .

3. transitive if x R y and y R z imply x R z .

4. an equivalence relation if it is reflexive, symmetric and transitive .

5. an order relation if it is transitive and for any two elements x and y, exactly one of the
following occurs: x = y, x R y, y R x.

An ordered set is a pair (S, R), where S is a set and R is an order relation defined on it.
Notation. Let (S, R) be an ordered set, and let a, b ∈ S. Then

(a, b) := {x ∈ S : a R x, x R b}, [a, b] = (a, b) ∪ {a, b}.

[a, b) = (a, b) ∪ {a}, (a, b] := (a, b) ∪ {b}.

(−∞, b) := {x ∈ S : x R b}, [a, b) = (a, b) ∪ {a}, (a, ∞) := {x ∈ S : a R x},

(−∞, b] := (−∞, b) ∪ {b}, [a, ∞) := (a, ∞) ∪ {a}.

Let S be an ordered set, R being an order relation defined on it. Let A be a subset of S.
An element a ∈ S is said to be an upper bound for A if for any x ∈ A, either x R a or x = a.
An element b ∈ S is said to be a least upper bound (or supremum) for A if

1. b is an upper bound of A, and

1
 2. if c is an upper bound for A, then either b = c or b R c.

Exercise 1.1 Show that a subset A of an ordered set S can not have more than one supremums.

Notation. If x is the least upper bound of a set E, we write x = sup E.

Notation. Let (S, <) be an ordered set and x, y ∈ S. By x > y, we mean y < x, by x ≤ y, we
mean ‘either x < y or x = y’, and by x ≥ y, we mean ‘either y < x or y = x’.

Exercise 1.2 Let (S, <) be an ordered set and let E, F ⊆ S.

1. Suppose a = sup E, and b < a. Show that there is an element x ∈ E such that b < x.
2. If x ≤ y for all x ∈ E and y ∈ F , then show that sup E < sup F or sup E = sup F .

Least upper bound property:

Let S be an ordered set. It is said to have the least upper bound property if whenever a
nonempty subset S has an upper bound, it has a least upper bound in S.

1.3 Binary operations

A binary operation on a set S is a map from S × S to S.

2 Real Numbers
Definition 2.1 The space of real numbers R is a set, equipped with two operations, addition
(+) and multiplication (.), and a relation <, satisfying properties 1–5 below:

1. Properties of addition:

A1 (x + y) + z = x + (y + z),

A2 x + 0 = x = 0 + x,

A3 for each x ∈ R, there is a −x ∈ R such that x + (−x) = 0 = (−x) + x.

A4 x + y = y + x,

In short, R is an abelian group under addition.

2. Properties of multiplication:

M1 (x.y).z = x.(y.z),

M2 x.1 = x = 1.x,

M3 for each x 6= 0 ∈ R, there is a 1/x ∈ R such that x.(1/x) = 1 = (1/x).x.

M4 x.y = y.x,

2
 In short, R\{0} is an abelian group under multiplication.
3. Distributive law holds: (x + y)z = xz + yz.
Any space, equipped with two operations satisfying 1, 2 and 3 above is called a field.

Exercise 2.2 Shw that

1. 0.x = 0 for all x ∈ R.
2. if x + y = x + z, then y = z.
3. if x · y = 0 and x 6= 0, then y = 0.
4. if xy = xz and x 6= 0, then y = z.
5. −(−x) = x for all x ∈ R.
1
6. 1/x
= x.

Definition 2.3 An ordered field is a field, with an order relation < such that
1. a < b ⇒ a + c < b + c,
2. 0 < a, 0 < b ⇒ 0 < ab.

4. (R, <) is an ordered field

Exercise 2.4 Show that

1. 0 < 1.
2. 0 < x < y implies 0 < 1/y < 1/x.
3. What happens if x < y < 0?
4. An ordered field is always infinite.
5. A field that is ordered is not necessarily an ordered field.

5. R has the LUB property.

Remark 2.5 At this point, we can give the definitions of the sets N, Z, Z+ and Q. Call a
subset E of R inductive if whenever it contains a real x, it also contains x + 1. R itself is one
such set. Now define N to be the intersection of all inductive sets that contain 0. Let us first
convince ourselves that this will indeed define what we normally mean by the set of natural
numbers. Let n := |1 + 1 +{z· · · + 1}. First, if a < 0, then a 6∈ N. Because the set R+ is inductive,
n times
and it does not contain a. Next, if 0 < a < 1, then a 6∈ N; for the set {0} ∪ [1, ∞) is inductive
and does not contain a. Similarly, if n < a < n+1, then a 6∈ N, because {0, 1, . . . , n}∪[n+1, ∞)
is inductive and does not contain a.
Having defined N, it is easy to define the other three sets.

Lemma 2.6 Between any two real numbers, there is another.

Proof : ✷

Lemma 2.7 Let x, y ∈ R. If x ≤ y + ǫ for all ǫ > 0, then x ≤ y.

3
Proof : ✷

Lemma 2.8 Let E, F ⊆ R, and G = {x + y : x ∈ E, y ∈ F }. Then sup G = sup E + sup F .

Proof : sup E + sup F will be an upper bound for G. So sup G ≤ sup E + sup F . Therefore it
is enough to show that sup G ≥ sup E + sup F . Take ǫ > 0. We will show that sup G + ǫ ≥
sup E + sup F . Choose x ∈ E, y ∈ F such that x > sup E − ǫ/2 and y > sup F − ǫ/2. Then
x + y > sup E + sup F − ǫ. Hence sup G > sup E + sup F − ǫ. ✷

Exercise 2.9 Show that x > y > 0 implies x2 > y 2 . Using this, show that if x > 0, y > 0 and x2 > y 2 , then
x > y.

Example 2.10 1. Let E = {r ∈ Q : r < 1/2}. Then sup E = 1/2.

2. Let E = {r ∈ Q : r 2 < 2}. Then (sup E)2 = 2.

Sketch of proof : Write y = sup E. We will show that we can not have y 2 < 2 or y 2 > 2.
First, suppose if posssible that y 2 > 2. Choose an h such that (y − h)2 > 2. Then y − h
is an upper bound, which contradicts the definition of y. Next assume if possible that
y 2 < 2. This time choose an h > 0 such that y + h is rational and (y + h)2 < 2. This
would contradict the fact that y is an upper bound. So one must have y 2 = 2.

Theorem 2.11 (Principle of Mathematical Induction) Let E be a set of positive integers

such that
(i) 1 ∈ E,
(ii) if an integer k ∈ E, then k + 1 ∈ E.
Then E contains all positive integers.

Proof : E ∪ {0} is an inductive set containing the element 0. Hence N ⊆ E ∪ {0}. Hence
Z+ = N\{0} ⊆ E. ✷

Theorem 2.12 (Well Ordering Principle) Every nonempty subset of N has a smallest mem-
ber.

Proof : Suppose if possible S is a nonempty subset of N that has no smallest element. Let
E = {x ∈ N : x < n for all n ∈ S}, i. e. E is the set of strict lower bounds of S in N. Then
0 ∈ E, because, otherwise we must have 0 ∈ S, which would mean 0 is the smallest element of
S. Next, if x ∈ E, then x + 1 ∈ E. For, if x + 1 6∈ E then there is a t ∈ S such that t ≤ x + 1.
Since S has no smallest member, there is an s ∈ S such that s < t. But then s < x + 1,
and therefore s ≤ x so that x 6∈ E. Thus E is inductive, and hence contains all the natural
numbers. In particular S ⊆ E = N. Now take a t ∈ S. Since t ∈ E also, this implies t < t,
which can not happen. ✷
Elementary consequences:

4
Theorem 2.13 1. N is not bounded above.

2. If x and y are positive reals, then ∃n ∈ N such that y < nx.

3. If x is a positive real, then ∃n ∈ N such that 0 < 1/n < x.

4. Let E be a bounded subset of N. Then it has a maximum element.

5. Between any two distinct real numbers, there is a rational number.

6. Given a positive real x and a positive integer n, there is one and only one positive real y
such that y n = x.

7. Between any two distinct real numbers, there is an irrational number.

Proof :

1. Suppose if possible N has an upper bound. By the LUB property of R, it has an upper
bound x in R. Since then x − 1 is not an upper bound for N, there is an n ∈ N such that
n > x − 1. But then n + 1 > x, i.e. x can not be an upper bound for N.

2. By (1), there is an n ∈ N such that y/x < n.

3. Take y = 1 in (2).

4. Let t ∈ R be an upper bound for E so that n < t for all n ∈ E. By (1), there is an
integer k ∈ N such that t < k. Define F = {k − n : n ∈ E}. Since for all n ∈ E, n < k,
this is a subset of N. Hence by the well ordering principle, F has a minimum element,
say k − m, where m ∈ E. This means k − m ≤ k − n for all n ∈ E, i. e. m ≥ n for all
n ∈ E. But m ∈ E. So m is the maximum element of E.

5. Consider the case 0 < x < y. By (3), there is an n ∈ N such that 0 < 1/n < y − x.
Now look at the set {m/n : m ∈ N}. Let m0 = max{m : m/n ≤ x}. Then we have
x < m0n+1 < y (the first inequality follows from the definition of m0 , and the second
follows from the fact that m0n+1 − x = 1/n − (x − mn0 ) ≤ 1/n < y − x).
The case x < y < 0 is similar. In the other two cases, viz. x < 0 < y and y < 0 < x, 0
would be a rational between the two.
x 2 1
6. Proof for the case n=2 : Let E = {t ∈ R : t ≥ 0, t2 ≤ x}. Observe that ( 1+x ) = (1+ x1 )2
<
1
1/x = x, and (1 + x)2 > x. Therefore the set E is nonempty and bounded above. Let
y = sup E.
Case I. y 2 < x: For any h, (y + h)2 = y 2 + 2hy + h2 = y 2 + h(2y + h). Choose h in such
2 x−y 2
a way that 0 < h < y, and h < x−y 2 2
3y . Then (y + h) < y + 3y 3y = x. So y can not be
the supremum of E.
Case II. y 2 > x: (y − h)2 = y 2 − 2hy + h2 = y 2 − h(2y − h). Choose h ∈ (0, y) such that

5
 2
h < y 3y−x . Then (y − h)2 = y 2 − h(2y − h) > y 2 − h(2y + h) > y 2 − (y 2 − x) = x. So y − h
is an upper bound, which is not the case.
Therefore we must have y 2 = x.

Proof for general n: As before, let E = {t ∈ R : t ≥ 0, tn ≤ x} and y = sup E. If y n < x,

n
choose h such that 0 < h < y and h < 2x−y
n y n−1 . Then

    
n n n n−1 n n−2 n−1
(y + h) = y +h y + y h + ··· + h
1 2
   
n x − yn n n
< y + n n−1 + ··· + y n−1
2 y 1 n
< x.

But this is not possible.

y n −x
Similarly if y n > x, choose h such that 0 < h < y, h < 2n y n−1
. Then
    
n n−1 n n−2
(y − h)n = y n − h y − y h + · · · ± hn−1
1 2
    
n n n−1 n n−2 n−1
> y −h y + y h + ··· + h
1 2
   
yn − x n n
> y n − n n−1 + ··· + y n−1
2 y 1 n
> x.

This would imply y − h is an upper bound for E.

So we must have y n = x.

7. Use part 5. ✷

6
 Week 2

2.1 Rational powers of positive numbers

For x > 0 and p a positive rational, define xp to be (xm )1/n where p = m/n with m, n ∈ Z+ ,.

Exercise 2.14 Show that the above definition is independent of the choice of m and n.

1
For x > 0 and p a negative rational, define xp to be x−p
. Define x0 to be 1 for all x > 0.

Exercise 2.15 Let x, y > 0 and p, q ∈ Q. Show that

xp xq = xp+q , (xp )q = xpq , (xp )(y p ) = (xy)p .

Show that if x > 1 and 0 < p < q, then xp < xq .

Exercise 2.16 Let E, F ⊆ R+ , with sup E > 0, sup F > 0. Assume EF ⊆ G. Show that (sup E)(sup F ) ≤
sup G.

2.2 Arbitrary powers of positive numbers

Let x > 1, p > 0. Define
Ep,x = {xr : r ∈ Q, 0 ≤ r < p}.

Take some rational s > p. Then y < xs for all y ∈ Ep,x . Thus Ep,x is bounded above. Define
fp (x) = sup Ep,x .

Proposition 2.17 Let x > 1, p, q > 0, and fp (x) be as above. Then

1. fp (x) = xp if p ∈ Q,

2. fp (x)fq (x) = fp+q (x),

3. fq (fp (x)) = fpq (x),

4. fp (x) < fq (x) if p < q,

5. fp (x)fp (y) = fp (xy).

Proof : We will prove the first two parts here. Others would be left as exercises.
Part (1) will follow from the following lemma.

7
Lemma 2.18 Let x > 1, and p be a positive rational. Given any ǫ > 0, there exists a rational
r < p such that xp − ǫ < xr .

Proof : Choose N ∈ N such that xp ( x−1

N ) < ǫ. Now write p in the form m/n where n > N .
m−1
Take r = n . Then
m m−1
xp − xr = x n − x n

m−1 1
= x (x n − 1)
n

x−1
< xp
n
< ǫ.

✷
Next, we will show that fp (x)fq (x) = fp+q (x) for all p, q > 0.
It is easy to see that Ep Eq ⊆ Ep+q for all p, q > 0. Hence it follows that fp (x)fq (x) ≤ fp+q (x)
We now need to prove the reverse inequality.
Take an ǫ > 0. Choose a rational t such that t < p + q and xt > fp+q (x) − ǫ (in the case
p + q is a rational, we have to use the above lemma for this). Let r and s be rationals such that
r ≤ p, s ≤ q and r + s = t. Then xr xs = xt > fp+q (x) − ǫ. Hence fp (x)fq (x) > fp+q (x) − ǫ.
Since this is true for all ǫ > 0, we have fp (x)fq (x) ≥ fp+q (x). ✷
This number fp (x) will be called the pth power of x and will normally be denoted by xp .
1 1
For 0 < x < 1, define xp to be (1/x) p
p . For x > 0 and p < 0, define x to be x−p .

Exercise 2.19 Show that if x > 0 and p, q are any real numbers, then
xp xq = xp+q , (xp )q = xpq .

3 Countable Sets
Suppose we are given two sets E and F and asked to show that E contains en element that
is not contained in F . One way of course would be to check each and every element of E if it
belongs to F or not. But that may not always be possible. If we could somehow show that the
set E is ‘bigger’ than F , that would automatically imply the existence of one such element.
Thus it is natural to try and compare the ‘sizes’ of different sets. For finite sets, this is easy.
One just has to count the number of elements in each set and compare them. But this way
of comparing can not be readily applied for infinite sets. For finite sets, another equivalent
alternative way is to see if there is any bijection between the two sets. They are of the same
size if there is some bijection. This notion extends readily to infinite sets as well, and we say
two sets are of the same size if there exists a bijection between the two. Now the standard
accepted terminology for ‘size’ is cardinality.
Two sets E and F are said to be of the same cardinality if there is a bijective map from
one to the other.

8
 For finite sets, in addition to comparing the cardinality of two sets, we can also talk about
the value of their cardinality, which is given by a natural number. For example, we say that a
finite set is of cardinality n if there is a bijection between that set and the set {1, 2, . . . , n}. For
infinite sets, the way we have defined the concept of cardinals being equal, it is clear that the
‘value’ has to be some known set, rather than any integer, or real number. The most natural
candidate that comes to mind immediately is the set N of natural numbers. There is a special
name for sets whose cardinality ‘equals’ N, or, whose cardinality is the same as that of N.

Definition 3.1 A set E is said to be countable if there is a bijection between E and N. It is

said to be at most countable if it is finite or countable.

Example 3.2 1. Z. We can take the following bijection:


2k if k ≥ 0,
k 7→
−2k − 1 if k < 0.

2. The set of all even integers.

3. The set of all odd integers.

Proposition 3.3 Any subset of a countable set is at most countable.

Proof : It is enough to prove that any infinite subset of a countable set is countable. Let F be
a countable set and E be an infinite subset of F . Let f : N → F be a bijection. Then

F = {f (0), f (1), . . .}.

Let us define a function g : N → E as follows:

g(0) = f (k) if k is the least integer n for which f (n) ∈ E,
(i. e. k = min {n : f (n) ∈ E} = min f −1 (E))
Having chosen g(0), g(1), . . . , g(r), let

g(r + 1) = f (k) where k = min f −1 (E\{g(0), g(1), . . . , g(r)}).

Corollary 3.4 If there is an injective map f from a set E into a countable set F , then E is
at most countable.

Exercise 3.5 Prove the converse of the above corollary.

Proposition 3.6 Countable union of at most countable sets is at most countable.

Proof : Let Ei , i ∈ N be a countable family of sets that are at most countable. Since each of Ei
is at most countable, there are injective maps fi : Ei → N. Define a map f : ∪i∈N Ei → N × N
such that
f (a) = (j, fj (a)) where j = min{k : a ∈ Ek }.
Check that f is 1–1. Since N × N is countable, it follows that ∪i∈N Ei is at most countable. ✷

9
Proposition 3.7 Finite Cartesian products of countable sets is countable.

Proof : It is enough to show that Z+ × Z+ is countable. Define f : Z+ × Z+ → N by

m+n−2
X
f (m, n) = i + m.
i=1

We will show that f is 1–1. Suppose f (m, n) = f (r, s). This means
m+n−2
X r+s−2
X
i+m= i + r. (3.1)
i=1 i=1
Pm+n−2
If m + n > r + s, then we get i=r+s i = r − m. But clearly the left hand side is greater
than or equal to the initial term r + s − 1 and r + s − 1 ≥ r > r − m. Thus we can not have
m + n > r + s. Similarly r + s > m + n is also not possible. Hence r + s = m + n, and
therefore using (3.1), we get m = r and n = s. Thus f is 1–1. Now by corollary 3.4, Z+ × Z+
is countable. ✷

Corollary 3.8 Q is countable.

Proof : Write each element r of Q in the form m/n, where m ∈ Z, n ∈ Z+ and m, n have no
common divisor. Now define f : Q → Z × Z+ by

f (r) = (m, n).

Since Z and Z+ are countable, and f is 1–1, Q is countable. ✷

Remark 3.9 Infinite cartesian products of finite sets may fail to be countable.

For example, take the space S of all sequences each of whose terms are either 0 or 1. Check
that this is just the cartesian product of countably many copies of {0, 1}. Let us show that
this is uncountable. If it is countable, there would exist a bijection between N and S. Now
take any map f : N → S. Let us write all the elements in the range of f in a column, stacked
one below the other:

f (0) = a00 , a01 , a02 , . . . ,

f (1) = a10 , a11 , a12 , . . . ,
f (2) = a20 , a21 , a22 , . . . ,
··· ···

Now choose a sequence {bn } as follows: choose bi so that bi 6= aii for all i. Then {bn } can not
be equal to any of the sequences f (0), f (1), . . . . Thus {bn } ∈ S, but is not in the range of f .
So f can not be onto, and hence can not be bijective.
Notice that in order that you can choose bi different from aii , the set where the terms of
the sequence take values must have more than one element.
The argument employed above is due to mathematician George Cantor, and is known as
the Cantor’s diagonal argument.

10
3.1 Decimal Expansion of a Real Number
A real sequence is just a map from Z+ to R.
Let x ∈ [0, 1). We will associate with x a sequence each term of which is a nonnegative
integer and lies between 0 and 9.
Choose k1 ∈ N such that k101 ≤ x < k110+1 .
k2 k1 k2 +1
Choose k2 ∈ N such that 10 2 ≤ x − 10 < 102 .
kn Pn−1 ki kn +1
Having chosen k1 , . . . , kn−1 , choose kn such that 10n ≤ x − i=1 10i < 10n .
Then {k1 , k2 , . . .} is the sequence that we associate with the number x and one says that
.k1 k2 k3 . . . the decimal expansion of x. It is easy to see that 0 ≤ ki ≤ 9 for each i.
Pn ki
Remark 3.10 1. x is the supremum of the set E = {x1 , x2 , . . .} where xn = i=1 10i .

By our choice of k1 , k2 etc. we have x ≥ xn for all n ≥ 1. Hence sup E ≤ x. Now, given any
P
ǫ > 0, there is an n ∈ Z+ such that 10n > 1/ǫ. But by our choice, x < n−1 ki kn +1
i=1 10i + 10n .
P
Hence x − ǫ < x − 101n < ni=1 10 ki
i = xn . Thus x = sup E.

2. As a corollary of the above remark, it follows that if the decimal expansions of two real
numbers coincide, then they are equal.

3. We now have a way of recovering the number x if we know its decimal expansion. But,
given a sequence whose terms are all integers and lie between 0 and 9, how do we know
in the first place whether it is indeed the decimal expansion of a number or not? First
of all, if a given sequence k1 , k2 , . . . is to be the decimal expansion of a real, then that
P
number has to be x := sup{x1 , x2 , . . .}, where xn = ni=1 10 ki
i . And secondly, we must
1
have xn ≤ x < xn + 10n for all n. The first inequality here is clear from the definition of
x. So all we need to do is, verify whether x < xn + 101n for all values of n.
1
For this, we need the following simple fact: If kj < 9 for some j > n, then x < xn + 10n .

For any m > j, we have

m
X ki kn+1 kj km
xm = = xn + n+1 + . . . + j + . . . + m
10i 10 10 10
i=1
kn+1 kj 9 9
≤ xn + + ... + + j+1 + . . . + m
10n+1 10 j 10 10
kn+1 kj 1 1
≤ xn + n+1 + . . . + + j (1 − m−j )
10 10j 10 10
kn+1 kj + 1
≤ xn + n+1 + . . . + .
10 10j

kn+1 kj +1
Hence x ≤ xn + 10n+1
+ ... + 10j
. But

kn+1 kj + 1
xn + n+1
+ ... +
10 10j
9 9
≤ xn + n+1 + . . . + j
10 10

11
 1 1
= xn + n
(1 − j−n )
10 10
1
< xn + n .
10
1
Hence x < xn + 10n .

Now, if infinitely many of the ki ’s are different from 9, then for any n, there will always
exist a j > n such that kj < 9. Therefore we will have xn ≤ x < xn + 101n for all n. But
this means k1 , k2 , . . . is the decimal expansion of x. Thus the answer to the question is: if
infinitely many of the ki ’s are different from 9, then the sequence k1 , k2 , . . . is the decimal
expansion of a real number x ∈ [0, 1). In other words, we have the following proposition:

Proposition 3.11 Let {k1 , k2 , . . .} be a sequence such that each ki ∈ N, 0 ≤ ki ≤ 9 for

P
all i, and ki 6= 9 for infinitely many i’s. Let xn = ni=1 10 ki
i , n ≥ 1 and let E be the set
Pn 9
{x1 , x2 , . . .}. Then E is bounded above (by sup{ i=1 10i : n ∈ Z+ } = 1) and the decimal
expansion for sup E is .k1 k2 . . ..

4. Converse of the above statement is also true, i. e. if the ki ’s are all 9 from a certain stage
onwards, then the sequence k1 , k2 , . . . is not the decimal expansion of any real.

Exercise 3.12 Let {k1 , k2 , . . .} be a sequence such that each ki ∈ N, 0 ≤ ki ≤ 9 for all i. Let m ∈ Z+
be such that km 6= 9 and ki = 9 for all i > m. Let E be as earlier. Then show that the decimal expansion
of sup E is .k1 . . . km−1 (km + 1)00 . . ..

5. There is nothing special about the number 10 above. One could equally well have chosen
k1 , k2 etc in such a manner that k21 ≤ x < k12+1 , k222 ≤ x − k21 < k22+1 kn
2 , . . . , 2n ≤
Pn−1 ki
x − i=1 2i < kn2+1n , and each ki is either 0 or 1. The resulting expansion then is called

the diadic expansion of x

Theorem 3.13 R is uncountable.

Proof : It is enough to prove that the interval (0, 1) is uncountable. Let f : Z+ → (0, 1) be
a map, and let .ki1 ki2 . . . be the decimal expansion of the number f (i). Thus the range of f
consists of the elements

f (1) = .k11 k12 . . .

f (2) = .k21 k22 . . .
... ...

Now choose a sequence {rn } as follows:


0 if 1 ≤ knn ≤ 9,
rn =
1 if knn = 0.
Then there is a real x whose decimal expansion is .r1 r2 . . .. Clearly this number x is different
from all the numbers f (i) (otherwise their decimal expansions would have been the same).
Thus f can not be onto. Therefore there can not exist any bijection between Z+ and (0, 1). ✷

12
Exercise 3.14 Which of the following sets are countable and which are uncountable:
R, the set of all irrational numbers, the set of all open intervals in R with rational endpoints, the set of all
rectangles in R2 whose vertices have rational endpoints.

13
 Week 3

4 Topological Notions
Let E ⊆ R. A point x ∈ E is said to be an interior point of E if there is an ǫ > 0 such that
Nǫ (x) := (x − ǫ, x + ǫ) ⊆ E. The set of all interior points of E is called the interior of E and
is denoted by int E.
A subset E of R is called open if every point of E is an interior point of E. A subset E of
R is said to be closed if E c is open
A neighbourhood of a point a will always mean an open set of the form (b, c) containing
the point a.

Proposition 4.1 Arbitrary union of open sets is again an open set.

Finite intersection of open sets is open.

Proof : Let Uα be a family of open sets, and let U = ∪α Uα . Take any x ∈ U . Then x ∈ Uα
for some α. Since Uα is open, there is an ǫ > 0 such that Nǫ (x) ⊆ Uα . But then Nǫ (x) ⊆ U .
Therefore x is an interior point of U . Thus U is open.
Let U1 , U2 , . . . , Un be open, and let U = ∩ni=1 Ui . Take any x ∈ U . Then x ∈ Ui for all
i ∈ {1, 2, . . . , n}. Since each Ui is open, there exist positive numbers ǫ1 , ǫ2 , . . . , ǫn such that
Nǫi (x) ⊆ Ui . Now take ǫ = min{ǫ1 , ǫ2 , . . . , ǫn }. Then Nǫ (x) ⊆ Nǫi (x) ⊆ Ui for each i. Therefore
Nǫ (x) ⊆ U . Thus x is an interior point of U . Since this is true for any point x in U , U is open.
✷
1 1
Infinite intersection of open sets may fail to be open. For example, [0, 1] = ∩∞
n=1 (− n , 1+ n ).

Proposition 4.2 Let E ⊆ R and let V be the union of all open sets contained in E. Then
V = int E.

Proof : Let us first show that int E is open. Take an x ∈ int E. Then there is some ǫ > 0
such that (x − ǫ, x + ǫ) ⊆ E. We need to show that any point in (x − ǫ, x + ǫ) is an interior
point of E. Take y ∈ (x − ǫ, x + ǫ). Let δ = min{x + ǫ − y, y − x + ǫ}. Then δ > 0 and
(y − δ, y + δ) ⊆ (x − ǫ, x + ǫ) ⊆ E. Hence y ∈ int E. Thus (x − ǫ, x + ǫ) ⊆ int E, i. e. int E is
open.

14
 Since int E is open and is contained in E, one has int E ⊆ V . On the other hand, since V
itself is open, for any point x of V , there is an ǫ > 0 such that (x − ǫ, x + ǫ) ⊆ V . Also, V ⊆ E.
Hence one has (x − ǫ, x + ǫ) ⊆ E. Thus x ∈ int E. Thus V ⊆ int E. ✷

Proposition 4.3 Arbitrary intersection of closed sets is closed.

Finite union of closed sets is closed.

Proof : Exercise. ✷

Exercise 4.4 Show that infinite union of closed sets may fail to be closed.

Definition 4.5 For a subset E of R, the closure Ē of E is defined to be the intersection of

all closed sets F such that E ⊆ F .

The following proposition says that the interior of a set is the maximal open set contained
in it and the closure of a set is the minimal closed set containing it.

Proposition 4.6 Let E ⊆ R.

1. Let V be an open subset of R such that (1) V ⊆ E and (2) if U is open and U ⊆ E then
U ⊆ V . Then V = int E.

2. Let F be a closed subset of R satisfying (1) E ⊆ F , and (2) if G closed and E ⊆ G then
F ⊆ G, then F = Ē.

Proof : Exercise. ✷

4.1 Limit point

Let E ⊆ R. A point x ∈ R is called a limit point of E if for any ǫ > 0, the neighbourhood Nǫ (x)
contains infinitely many points of E. A limit point is also called an accumulation point or
a cluster point.
We denote the set of all limit points of a set E by E ′ and call it the first derived set of E.

Lemma 4.7 A number x ∈ R is a limit point of E if and only if for every ǫ > 0, Nǫ (x)
contains a point of E other than x..

Proof : Take any ǫ > 0. If Nǫ (x) contains infinitely many points of E, then it has to contain
at least one point of E other than x. So if x is a limit point then the given condition holds.
For the converse, assume that x is not a limit point of E. Then there is an ǫ > 0 for which
there are only finitely many points, say x1 , x2 , . . . xn of E in Nǫ (x). Let δ = min{|x − xi | : 1 ≤
i ≤ n, |x − xi | > 0}. Then the neighbourhood Nδ (x) does not contain any point xi for which
x 6= xi , i. e. does not contain any point of E other than x. ✷

15
Exercise 4.8 A finite subset of R can not have any limit point.

Example 4.9 1. 0 is a limit point of the set { n1 : n ∈ Z+ }.

2. E = (0, 1). Any point of E is a limit point of E.

3. E = { n1 : n ∈ Z+ , n ≥ 2} ∪ {1 − 1
n : n ∈ Z+ , n ≥ 2}. 0 and 1 are two limit points of E.
1
4. E = [0, 1) ∪ {1 + n : n ∈ Z+ }. Any point in [0, 1] is a limit point of E.

5. E = Q. Any real number is a limit point of E.

Thus there can be several limit points of a set.

Proposition 4.10 Let E and F be subsets of R such that E ⊆ F . Then E ′ ⊆ F ′ .

Proof : Exercise. ✷

Proposition 4.11 Let E1 , E2 ⊆ R. Then

1. (E1 ∪ E2 )′ = E1′ ∪ E2′ ,

2. (E1 ∩ E2 )′ ⊆ E1′ ∩ E2′ .

Proof : Take x ∈ E1′ ∪ E2′ . Then either x ∈ E1′ or x ∈ E2′ . In the first case, for every ǫ > 0,
Nǫ (x) contains infinitely many points of E1 , and in the second case, for every ǫ > 0, Nǫ (x)
contains infinitely many points of E2 . In either case, Nǫ (x) contains infinitely many points
from E1 ∪ E2 . So x ∈ (E1 ∪ E2 )′ . Thus E1′ ∪ E2′ ⊆ (E1 ∪ E2 )′ .
Observe that to prove the reverse inclusion, it is enough to show that if x ∈ (E1 ∪ E2 )′ , and
x 6∈ E1′ , then x ∈ E2′ . So assume x ∈ (E1 ∪ E2 )′ , and x 6∈ E1′ . Since x 6∈ E1′ , there is an ǫ0 > 0
for which Nǫ0 (x) ∩ E1 is at most finite. But then for any ǫ ≤ ǫ0 , Nǫ (x) ∩ E1 is finite. But since
Nǫ (x) ∩ (E1 ∪ E2 ) is infinite for all ǫ > 0, we must have Nǫ (x) ∩ E2 infinite for all ǫ ≤ ǫ0 . This
implies Nǫ (x) ∩ E2 is infinite for all ǫ > 0. Hence x ∈ (E2 )′ .
This proves part (1). Proof of part (2) is left as an exercise. ✷
The following examples show that one does not have equality in part (2) in general.
Take E1 = (0, 1), E2 = (1, 2), or E1 = Q, E2 = R\Q.

Proposition 4.12 Let Eα be a family of subsets of R. Then

∪α Eα′ ⊆ (∪α Eα )′ , (∩α Eα )′ ⊆ ∩α Eα′ .

Proof : Exercise. ✷
The following example illustrates that one does not have equality in the above.
Let Q = {r1 , r2 , . . .}. Let En = {rn }.

Proposition 4.13 E ∪ E ′ is closed.

16
Proof : Take x ∈ (E ∪ E ′ )c . Consider Nǫ (x), where ǫ > 0. Our claim now is that if Nǫ (x)
contains a point of E ∪ E ′ , then it contains a point of E. Suppose y ∈ E ′ be a point in Nǫ (x).
Since Nǫ (x) is open, there is a δ > 0 such that Nδ (y) ⊆ Nǫ (x). Since y ∈ E ′ , Nδ (y) contains a
point of E, say z, that is different from x. But then z ∈ Nǫ (x).
From the above, it now follows that if every neighbourhood of x intersects E ∪ E ′ , then
every neighbourhood will contain an element of E (other than x) and hence x ∈ E ′ , which is
not the case. So there exists a neighbourhood Nǫ (x) such that Nǫ (x) ∩ (E ∪ E ′ ) = Φ. But this
means Nǫ (x) ⊆ (E ∪ E ′ )c . ✷

Proposition 4.14 Ē = E ∪ E ′ .

Proof : Let F be a closed set containing E. We will show that E ∪ E ′ ⊆ F . Take x ∈ F c .

Since F c is open, there is a neighbourhood Nǫ (x) of x such that Nǫ (x) ⊆ F c . Since E ⊆ F ,
Nǫ (x) does not contain any point of E, i. e. x 6∈ E ′ . Thus F c ⊆ E ′ c , so that E ′ ⊆ F . Thus
E ∪ E′ ⊆ F . ✷

Definition 4.15 Let E ⊆ R. The boundary of E is defined to be the set Ē\ int E and is
denoted by ∂E.

Proposition 4.16 Let E ⊆ R. A point x is in the boundary of E if and only if for any ǫ > 0,
the neighbourhood Nǫ (x) intersects both E and E c .

Proof : First assume that x ∈ ∂E = Ē\int E. Since x ∈ Ē = E ∪ E ′ , every neighbourhood

Nǫ (x) intersects E. And since x 6∈ int E, there does not exist any ǫ > 0 for which Nǫ (x) ⊆ E;
in other words, for all ǫ > 0, Nǫ (x) intersects E c .
Conversely, let x be a real satisfying the condition of the proposition. Since Nǫ (x) intersects
E for every ǫ > 0, either x ∈ E or x ∈ E ′ , i.e. x ∈ Ē. Next, since Nǫ (x) always intersects E c ,
x is not an interior point of E. ✷

Exercise 4.17 Let E be a subset of R. Show that sup E (resp. inf E) is either a point of E or a limit point
of E.
Show that E ′′ ⊆ E ′ . Hence prove that E ′ is closed.

Theorem 4.18 (Bolzano-Weierstrass Theorem) Any bounded infinite subset of R has a

limit point.

Proof : Let E be a bounded subset of R. Let a0 , b0 ∈ R be such that E ⊆ [a0 , b0 ]. Consider

the two intervals [a0 , a0 +b a0 +b0
2 ] and [ 2 , b0 ]. At least one will contain infinitely many points of
0

E. If exactly one of them contain infinitely many points of E, call it [a1 , b1 ]. If both contain
infinitely many points of E, then write [a1 , b1 ] for the left interval. Now repeat the procedure
with the interval [a1 , b1 ] to get [a2 , b2 ]. Having obtained the intervals [a0 , b0 ], . . . , [an , bn ], get
[an+1 , bn+1 ] by dividing [an , bn ] into two equal parts.

17
 Observe the following now: a0 ≤ a1 ≤ . . . ≤ an ≤ . . . ≤ bn ≤ . . . ≤ b1 ≤ b0 . Hence the
set L = {a0 , a1 , . . .} has a supremum, and U = {b0 , b1 , . . .} has an infimum. Also, observe that
bn − an = b02−an
0
for all n ∈ N. We will show that sup L = inf U , and this number is a limit
point for the set E.
Let us write a = sup L, and b = inf U . Suppose if possible, a < b. Since an ≤ a and bn ≥ b
for all n, we have bn − an ≥ b − a for all n. By our observation made earlier, this implies
b0 −a0 n
2n ≥ b − a for all n. But this is not possible, as the set {2 : n ∈ N} is not bounded above.
Next, assume a > b. Then there exists c such that a > c > b. Since a = sup L, an > c for some
n, and bm < c for some m. Let k = max{m, n}. Then we have ak > c and bk < c; in other
words, bk < ak . This is not possible. So we can not have a > b either. Thus we have a = b.
Take an interval Nǫ (a) around a. Choose n such that an > a−ǫ, and m such that bm < a+ǫ.
Taking k = max{m, n}, we get [ak , bk ] ⊆ Nǫ (a). Hence Nǫ (a) contains infinitely many points
of E. ✷

Remark 4.19 In the course of the above proof, we have almost proved another very important
theorem, the Cantor intersection theorem, which says that if we have a decreasing sequence
of closed intervals, with lengths shrinking to zero, then their intersection is a singleton. We
will study it later, once we have introduced the notion of convergence of a sequence.

18
 Week 4

4.2 Extended Real Numbers

Often it is helpful to deal with the following set rather than R: R̄ := R ∪ {∞, −∞}. Define
addition and multiplication on this new set as follows: if a, b ∈ R, define a + b and ab to be
their sum and product respectively in R. For a ∈ R, define

a + ∞ = ∞ = ∞ + a, a + (−∞) = −∞ = (−∞) + a,

∞ + ∞ = ∞, (−∞) + (−∞) = −∞,


±∞ if a > 0
a · (±∞) =
∓∞ if a < 0
∞ · ∞ = ∞, (−∞) · (−∞) = ∞, ∞ · (−∞) = −∞, (−∞) · ∞ = −∞

∞ if a > 0
a/0 = , a/ ± ∞ = 0
−∞ if a < 0
We leave the following quantities undefined: ∞ + (−∞), 0 · ±∞, ±∞/ ± ∞.
By a neighbourhood of +∞ in R̄, we mean a set of the form (a, ∞], and by a neighbourhood
of −∞, we mean a set of the form [−∞, b).
For unbounded subsets of R, we agree on the following convention: if E ⊆ R is not bounded
above, then we say that +∞ is a limit point of E, and if E is not bounded below, then we say
that −∞ is a limit point of E.

Remark 4.20 Let E ⊆ R. Then x ∈ R̄ is a limit point of E if and only if every neighbourhood
of x intersects E at a point other than x. Equivalently, x ∈ R̄ is a limit point of E if and only
if every neighbourhood of x contains infinitely many points of E.

If we agree to use the extended real number system, then we can talk about supremum and
infimum of unbounded sets also.

4.3 Lim sup and lim inf

Definition 4.21 Let E be a subset of R. Define the lim sup (limit superior) and lim inf (limit
inferior) of the set E as follows:

sup E ′ if E ′ is nonempty,
lim sup E =
−∞ if E ′ is empty.

19
 
inf E ′ if E ′ is nonempty,
lim inf E =
∞ if E ′ is empty.

Observe that from Exercise 4.17 it follows that sup E ′ and inf E ′ both are limit points of
E.

Proposition 4.22 Let E ⊆ R be bounded and infinite and let a = lim sup E. Then for any
ǫ > 0, Nǫ (a) contains infinitely many points of E, but there are at most finitely many points x
in E with x ≥ a + ǫ.
Similarly, if b = lim inf E, then there are infinitely many points of E in Nǫ (b), but only
finitely many points x satisfying x ≤ b − ǫ.

Proof : Observe that from Exercise 4.17 it follows that lim sup E := sup E ′ and lim inf E :=
inf E ′ both are limit points of E. Since a = lim sup E is a limit point of E, for any ǫ > 0,
the neighbourhood (a − ǫ, a + ǫ) contains infinitely many points of E. In particular, there are
infinitely many points x ∈ E with x > a − ǫ.
Suppose if possible, there is an ǫ > 0 such that there are infinitely many points x ∈ E with
x > a + ǫ. Let M be an upper bound of E. Then the set F := [a + ǫ, M ] ∩ E is infinite and
bounded. Hence by Bolzano-Weirstrass theorem, it has a limit point, say, b. Since F ⊆ E, b is
a limit point of E as well. But b ≥ a + ǫ, which contradicts the fact that a = lim sup E. ✷
The next proposition shows that we could have used the above conditions as the defining
conditions for lim sup and lim inf.

Proposition 4.23 Let E be a subset of R, and let a (resp. b) be a real satisfying the following
condition: for any ǫ > 0, Nǫ (a) (resp. Nǫ (b)) contains infinitely many points of E, but there
are at most finitely many points x in E with x ≥ a + ǫ (resp. x ≤ b − ǫ).
Then a = lim sup E (resp. b = lim inf E).

Exercise 4.24 Prove/disprove the following:

lim sup(E + F ) ≥ lim sup E + lim sup F,
lim sup(EF ) ≥ (lim sup E)(lim sup F ) if E ⊆ R+ , F ⊆ R+ .

5 Sequences and Series

5.1 Sequences
A real sequence is just a map from N (or Z+ ) to R. A sequence f : n 7→ an is usually denoted
by {an }n∈N , or by simply {an }. Let {an } be a sequence and let {kn }n be a sequence such that
k1 < k2 < . . . < kn < . . .. Then we say {bn }n is a subsequence of {an } if bn = akn for all n.
A sequence {an } is said to converge to a limit a if for any neighbourhood Nǫ (a), from a
certain stage onwards, all the terms of the sequence lie in Nǫ (a). In other words, given any
ǫ > 0, if there exists an N ∈ N such that |an − a| < ǫ for all n > N , then we say that the

20
sequence {an } converges to a (−∞ < a < ∞). In such a case, we write limn→∞ an = a. For
example, check that the sequence { n1 } converges to zero. The number N may and in general
will depend on the value of ǫ.
We say a sequence is convergent if it converges to a finite limit. Note that a convergent
sequence can have at most one limit.
A sequence {an } is said to diverge to ∞ if for any M > 0, there exists an N ∈ N such
that an > M for all n > N . In such a case, one writes limn→∞ an = ∞. It is said to diverge
to −∞ if for any M > 0, there is an N ∈ N such that an < −M for all n > N . We write
limn→∞ an = −∞.

Exercise 5.1 Give examples of sequences that diverge to ∞ or −∞.

A sequence {an } is said to be bounded above if there is an M > 0 such that an ≤ M for
all n. It is said to be bounded below if there is an M > 0 such that an > −M for all n. A
sequence that is bounded below as well as above is said to be bounded.

Lemma 5.2 Any convergent sequence is bounded.

The converse need not be true. (Give examples)

Proposition 5.3 Let {an } and {bn } be two convergent sequences, with lim an = a, lim bn = b.
Then

1. lim(an + bn ) = a + b,

2. lim(an bn ) = ab,

3. if b 6= 0, then lim( abnn ) = ab ,

4. for any λ ∈ R, lim(λan ) = λa,

5. if an ≤ bn for all n, then a ≤ b.

Examples

1. If p > 0, then lim n1p = 0.

√
2. If p > 0, then lim n p = 1.

√ √ n(n−1) √
3. lim n
n = 1. (Use n = (1 + n
n − 1)n > 2 ( n n− 1)2 )

4. If p > 0 and α ∈ R, then

nα
lim = 0.
(1 + p)n

k k
((1 + p)n > nk pk = n(n − 1) . . . (n − k + 1) pk! > ( n2 )k pk! if n > 2k. Choose k > α. Then
nα 2k k! 1
for all n > 2k, (1+p)n < pk nk−α .)

21
5.

0 if −1 < x < 1,
n
lim x = 1 if x = 1,

∞ if x > 1,
and the limit does not exist if x ≤ −1.

22
 Week 5

Proposition 5.4 If an converges to a, then any subsequence converges to a.

If an diverges to ∞, then any subsequence diverges to ∞.
If an diverges to −∞, then any subsequence diverges to −∞.

Proposition 5.5 If a is a limit point of the set E, then there is a sequence {an } of distinct
points from E such that lim an = a.

Proof : Since a is a limit point of E, the neighbourhood N1 (a) contains a point, say x of E
such that x 6= a. Define a1 = x. Next, let ǫ = min{ 12 , |a1 − a|}, and let a2 be a point of E lying
in Nǫ (a) such that a2 6= a. Having chosen a1 , a2 , . . . , an , let an+1 be a point of E different from
1
a such that it lies in Nǫ (a) where ǫ = min{ n+1 , |a1 − a|, . . . , |an − a|}.
First, observe that by our choice, |an+1 − a| < |ai − a| for 1 ≤ i ≤ n, so that an+1 is
different from a1 , a2 , . . . , an . Thus all the elements of the sequence {an } are distinct. We also
have |an − a| < n1 for all n. Hence given any ǫ > 0, if we take N to be an integer greater than
1/ǫ, then for all n > N , we will have |an − a| < n1 < N1 < ǫ. Thus lim an = a. ✷
Observe that if a is a limit point of a set E. then for any b ∈ E, a is a limit point of E\{b}
also.

Exercise 5.6 Prove that the converse of the above proposition is also true.

Proposition 5.7 Let E ⊆ R. Then a ∈ Ē if and only if there is a sequence (an )n such that
an ∈ E for all n and lim an = a.

5.2 Monotone sequences

A sequence {an } is said to be monotone increasing if for all n ∈ N, we have an ≤ an+1 for all
n ∈ N. It is said to be monotone decreasing if an ≥ an+1 for all n ∈ N. A monotone sequence
is one that is either monotone increasing or monotone decreasing.

Proposition 5.8 A monotone increasing sequence either converges to a finite limit or diverges
to ∞.
Similarly, a monotone decreasing sequence either converges to a finite limit or diverges to
−∞.

23
Proof : Let us write a = sup an . If a = ∞, then it follows immediately that the sequence an
diverges to ∞. So assume a < ∞. By the definition of a supremum, for any ǫ > 0, there is an
ak such that ak > a − ǫ. Since the sequence is increasing, we have an > a − ǫ for all n ≥ k.
And we of course have an ≤ a < a + ǫ for all n. Thus a = lim an .
Proof for monotone decreasing sequences is similar. ✷

Corollary 5.9 Any bounded monotone sequence is convergent.

Exercise 5.10 Show that the sequence an = {(1 + n1 )n } converges.

The limit of the above sequence is denoted by e.

Proposition 5.11 Every bounded sequence has a convergent subsequence.

Proof : Let {an } be a bounded sequence. Let E = {x ∈ R : x = an for some n}. Let us
consider the following two cases now:
Case I. E is finite. In this case, there is an a ∈ E such that a = an for infinitely many n’s. Hence
the sequence {bn } where bn = a for all n, is a subsequence of {an }. And clearly lim bn = a.
Thus {an } has a subsequence converging to a.
Case II. E is infinite. Since the sequence {an } is bounded, so is the set E. By Bolzano-
Weierstrass theorem, it has a limit point, say a. Take ǫ1 = 1. Choose n1 such that an1 ∈
(a − ǫ1 , a + ǫ1 )\{a}. Next, take ǫ2 = min{ 21 , |a − an1 |} and choose n2 such that an2 ∈ (a − ǫ2 , a +
1
ǫ2 )\{a}. Having chosen n1 , . . . , nk , take ǫk+1 = min{ k+1 , |a − an1 |, . . . , |a − ank |} and choose
nk+1 such that ank+1 ∈ (a − ǫk+1 , a + ǫk+1 )\{a}. The resulting subsequence {ank }k converges
to a. ✷

5.3 Cauchy sequences

Let {an } be a convergent sequence converging to a. Then, given any ǫ > 0, we can choose an
N > 0 such that |an − a| < ǫ/2 for all n > N . But then, whenever m and are both greater
than N , we have |am − an | ≤ |am − a| + |an − a| < ǫ. Thus the distance between the mth and
nth terms shrinks to zero as both m and n grow. Such sequences are called Cauchy sequences.

Lemma 5.12 Let {an } be a cauchy sequence. If a subsequence of {an } converges, then {an }
also converges and to the same limit.

Proposition 5.13 Let {an } be a Cauchy sequence in R. Then it is convergent.

Proof : Let us first show that {an } is bounded. Choose N > 0 such that |am − an | < 1
whenever m, n > N . Now let M = max{|a1 |, |a2 |, . . . , |aN |, |aN +1 | + 1}. Then for any n, we
have −M < an < M . (for n ≤ N , it is clear. for n > N , it follows from the fact that
|an − aN +1 | < 1).

24
 By the previous proposition, {an } has a subsequence, say {bn } where bn = akn for all n,
such that it converges. But if any subsequence of a Cauchy sequence converges, the Cauchy
sequence itself also converges and to the same limit. ✷

Remark 5.14 In many cases, it is easier to show that a given sequence is a Cauchy sequence
than showing that it is convergent, because the latter involves the limit of the sequence.
P n1
Example: an = n (−1) n .

5.4 Limit point of a sequence

Definition 5.15 Let {an } be a sequence. A point a is called a limit point if any neighbour-
hood of it contains infinitely many terms of the sequence.

Thus if −∞ < a < ∞, then a is a limit point if for any ǫ > 0, and any N > 0, there exists
n ≥ N such that an ∈ Nǫ (a). The point ∞ is a limit point if for any N > 0 and M > 0, there
exists n ≥ N such that an > M . Similarly −∞ is a limit point if for any N > 0 and M > 0,
there exists n ≥ N such that an < −M .

Proposition 5.16 Let {an } be a sequence of real numbers. Then

1. a real number a is a limit point of {an } if and only if there is a subsequence that converges
to a.

2. ∞ (respectively −∞) s a limit point of {an } if and only if there is a subsequence that
diverges to ∞ (respectively −∞).

Proof : 1. Assume {an } has a subsequence, say {ank }k that converges to a. Take any ǫ > 0. By
definition, there is an integer N such that a − ǫ < ank < a + ǫ for all k ≥ N . This implies that
there are infinitely many terms of the subsequence contained in the interval (a − ǫ, a + ǫ). But
these are terms of the original sequence. Thus there are infinitely many terms of the sequence
{an } in the interval (a − ǫ, a + ǫ). So a is a limit point.
Conversely, assume that a is a limit point of the sequence {an }. We will now produce a
subsequence of {an } that converges to a. Take ǫ = 1. Look at the set E1 := {n ∈ N : an ∈
(a − 1, a + 1)}. Since the interval (a − 1, a + 1) contains infinitely many terms of the sequence
{an }, this is nonempty. Let n1 = min E1 . Then

|an1 − a| < 1.

Next look at E2 := {n ∈ N : an ∈ (a− 12 , a+ 21 )}. Again, this is infinite. Hence E2 \{1, 2, . . . , n1 }

is nonempty. Let n2 = min E2 \{1, 2, . . . , n1 }. Then n2 > n1 and
1
|an2 − a| < .
2

25
 1 1
Having chosen n1 , . . . , nk , choose nk+1 to be min{n ∈ N : an ∈ (a− k+1 , a+ k+1 )\{1, 2, . . . , nk }.
Then {ank } is a subsequence of {an } and
1
|ank − a| < ∀ k.
k
It now follows that limk→∞ ank = a.
2. Assume there is a subsequence {ank } that diverges to ∞. Then given any M ∈ R, there
is an integer N such that ank > M for all k ≥ N . This implies that there are infinitely many
terms of {ank } and hence of the original sequence {an } that are greater than M . Since this is
true for any M , the point ∞ is a limit point of {an }.
Conversely assume that ∞ is a limit point of {an }. Chhose a subsequence {ank } in the
following manner. Let

n1 = min{n ∈ N : an > 1},

n2 = min{n ∈ N : n > n1 , an > 2},
··· ···
nk+1 = min{n ∈ N : n > nk , an > k + 1},
··· ···

Then the subsequence {ank }k diverges to ∞ (Show this!).

The proof of the statement involving −∞ is similar. ✷

Corollary 5.17 Let {an } be a sequence and let {ank } be a subsequence. Then any limit point
of {ank } is also a limit point of the sequence {an }.

Proof : Exercise. ✷

Exercise 5.18 Let {an } be a sequence of real numbers. Let E be the set of its limit points (thus E ⊆ R̄).
Show that
1. E is nonempty,
2. both sup E and inf E are limit points of {an } (i. e. points of E)

Exercise 5.19 Let {an } be a sequence. Show that

1. if {an } is not bounded above, then ∞ is a limit point of {an }.
2. if {an } is not bounded below, then −∞ is a limit point of {an }.

Let {an } be a sequence of real numbers. Define lim sup an and lim inf an to be the numbers
sup E and inf E respectively, where E is the set of all limit points of {an }.
By the earlier exercise, lim sup and lim inf of a sequence always exist.

Proposition 5.20 Let a ∈ R. Then lim sup an = a if and only if for any ǫ > 0, one has
an > a − ǫ for infinitely many n’s, but an ≥ a + ǫ for at most finitely many n’s.
lim sup an = ∞ if and only if there are infinitely many terms of the sequence in any neigh-
bourhood of ∞.
lim sup an = −∞ if and only if the sequence diverges to −∞.

26
Proof : Assume lim sup an = a. By defintion, for any ǫ > 0, the interval (a − ǫ, a + ǫ) contains
infinitely many terms of the sequence. So it is now enough to show that {n ∈ N : an ≥ a + ǫ} is
finite. Suppose if possible the above set is infinite. Then one can choose a subsequence {ank }k
such that ank ≥ a + ǫ for all k. Therefore for any limit point ℓ of this subsequence, one must
have ℓ ≥ a + ǫ > a = lim sup an . But ℓ has to be limit point of the original sequence {an } also,
in which case one must have ℓ ≤ a. Thus one gets a contradiction.
Conversely, assume that a satisfies the condition stated in the proposition. Take any ǫ > 0.
Since infinitely many terms are greater than a − ǫ, but only finitely many are greater than or
equal to a + ǫ, there are infinitely many terms in the interval (a − ǫ, a + ǫ). Thus a is a limit
point. It is enough now to show that if b is a limit point of the sequence, then one must have
b ≤ a. Suppose if possible, b is a limit point and b > a. Choose a real c such that a < c < b.
Since b is a limit point, there must be infinitely many terms that are greater than c, but by
the given condition, there are at most finitely many terms greater than c. Thus one gets a
contradiction. So one must have b ≤ a.
Proof of the remaining two parts are left as exercises. ✷

Proposition 5.21 lim inf an = a if and only if for any ǫ > 0, there are infinitely many terms
strictly less than a + ǫ, but at most finitely many terms less than a − ǫ.
lim inf an = ∞ if and only if the sequence diverges to ∞.
lim inf an = −∞ if and only if there are infinitely many terms of the sequence in any
neighbourhood of −∞.

Proof : Similar to the proof of the earlier proposition. ✷

Proposition 5.22 lim sup(an + bn ) ≤ lim sup an + lim sup bn ,

lim inf(an + bn ) ≥ lim inf an + lim inf bn ,
If an ≤ bn for all n then lim sup an ≤ lim sup bn and lim inf an ≤ lim inf bn .
If an ≥ b for infinitely many n, then lim sup an ≥ b.
If an ≤ b for all but finitely many n, then lim sup an ≤ b.

Proof : Exercise. ✷

Proposition 5.23 A sequence {an } converges (or diverges to ±∞) if and only if lim sup an =
lim inf an (= ±∞).

Proof : Exercise. ✷

27
 Week 6

5.5 Series
P
By a series, one means an expression of the form ∞ n=k an , where {an } is a sequence of real
P
numbers. In most situations, the integer k is either 0 or 1 and one writes just an instead
P∞ P∞ Pn
of n=0 an or n=1 an . The number sn := r=k ar is called the nth partial sum of the
P P
series an . The series an is said to converge if the sequence {sn }n converges. In such a
P
case, the number limn→∞ sn is called the sum of the series and is denoted by ∞ r=k ar . The
P
series an is said to diverge to ∞ (respectively −∞) if the sequence {sn }n diverges to ∞
(respectively −∞). In this case one says that the sum of the series is ∞ (respectively −∞).

Exercise 5.24 Asume that the sums of the two series

P P
an and bn are L and M respectively (here L and
P
M are extended reals). Let cn = an + bn for all n ∈ N. Show that the sum of the series cn is L + M whenever
L + M is defined.

P∞
Lemma 5.25 Let {an } be a sequence of real numbers. The series n=k an converges if and
P
only if the series ∞
n=0 an converges.

Proof : Let sn and tn denote the nth partial sums of the two series. Then for n ≥ k, one has
P
tn = sn + k−1r=0 ar . Therefore limn→∞ tn exists if and only if limn→∞ sn exists, and if the limits
P
exist, then limn→∞ tn = limn→∞ sn + k−1 r=0 ar . ✷

Exercise 5.26 Let {an } be a sequence of real numbers. Show that the series
P∞
n=k an diverges to ∞
P∞
(respectively −∞) if and only if the series n=0 an diverges to ∞ (respectively −∞).

P
Lemma 5.27 Let an be a convergent series. Then lim an = 0.

Proof : Let s be the sum of the series. Then limn→∞ sn = s = limn→∞ sn−1 . Hence limn→∞ an =
limn→∞ (sn − sn−1 ) = limn→∞ sn − limn→∞ sn−1 = s − s = 0. ✷
P1
The converse is not true. One could for example look at the series n.
P P
A series an is said to converge absolutely if the series |an | converges. It is said to
P P
converge conditionally if an converges but |an | does not.
P
Lemma 5.28 If an converges absolutely, then it converges.

28
 P
Proof : Let sn denote the nth partial sum of the series |an | and let tn denote the nth partial
P
sum of the series an . Then for any m and n, one has |tn −tm | ≤ |sm −sn |. Since the sequence
{sn } is Cauchy, it now follows (how?) that the sequence {tn } is also Cauchy. ✷

Proposition 5.29 (Leibneiz’s theorem) Let {cn } be a sequence such that ck ↓ 0, i.e. ck ≥ 0
P
for all k, ck ’s are decreasing, and lim ck = 0. Then the series (−1)n cn converges.
Pn k
Proof : It is enough to show that the sequence sn = k=1 (−1) ck is cauchy. Let m > n. Then

|sm − sn | = |(−1)n+1 cn+1 + (−1)n+2 cn+2 + . . . + (−1)m cm |

= |cn+1 − cn+2 + . . . + (−1)m−n−1 cm |.

Now the sequence {cn } is decreasing, therefore ck − ck+1 ≥ 0 for all k. Also, observe that

m−n−1 (cn+1 − cn+2 ) + . . . + (cm−1 − cm ) if m − n is even,
cn+1 −cn+2 +. . .+(−1) cm =
(cn+1 − cn+2 ) + . . . + (cm−2 − cm−1 ) + cm if m − n is odd.

Since the right hand side in nonnegative, we have

|cn+1 − cn+2 + . . . + (−1)m−n−1 cm | = cn+1 − cn+2 + . . . + (−1)m−n−1 cm .

Next,

m−n−1 cn+1 − (cn+2 − cn+3 ) − . . . − cm if m − n is even,
cn+1 −cn+2 +. . .+(−1) cm =
cn+1 − (cn+2 − cn+3 ) − . . . − (cm−1 − cm ) if m − n is odd.

Therefore we have |sm − sn | ≤ cn+1 . Since the sequence {cn } converges to 0, it follows that the
sequence {sn } is Cauchy. ✷

5.6 Tests for convergence

Exercise 5.30 Show that the series
< converges absolutely if |x| < 1,
8
X n
x diverges to ∞ if x ≥ 1,
:
neither converges nor diverges to ± ∞ if x ≤ −1.
P
Theorem 5.31 (Comparison test) 1. If |an | ≤ bn for all but finitely many n’s and bn
P
converges, then an converges.
P P
2. If an ≥ bn ≥ 0 for all n, and bn diverges, then an diverges.
P P P
Proof : Write sn = nk=1 |ak |, tn = nk=1 bk and un = nk=1 ak .
1. By the given condition, there is an integer N such that for all n ≥ N , one has |an | ≤ bn .
Hence whenever m and n are both greater than or equal to N , then |sm − sn | ≤ |tn − tm|. Since
the sequence {tn } is cauchy, it follows that {sn } is also cauchy. This means that the series
P
an converges absolutely. Hence it converges.

29
 2. By the given condition, we have un ≥ tn for all n. Also, since {tn } is increasing, and
diverges (to ∞), it follows that given any M ∈ R, there is an integer N such that tn > M for
all n ≥ N . By the inequality above, it then follows that un > M for all n ≥ N . Thus the series
P
an diverges. ✷
This is the main test for convergence or divergence of a series. By making explicit choices
of the sequence {bn }, one can derive various other tests. Indeed all the other tests that we will
learn are derived from the above in this manner.
P
Theorem 5.32 (Root test) If lim sup |an |1/n < 1, then an converges absolutely.
P
If lim sup |an |1/n > 1, then an does not converge.

Proof : Write a = lim sup |an |1/n . If a < 1, we can choose an ǫ > 0 such that a + ǫ < 1. Since
a = lim sup |an |1/n , we have |an |1/n ≥ a + ǫ for at most finitely many terms. In other words,
there is an integer N such that |an |1/n < a + ǫ for all n ≥ N . But this means |an | < (a + ǫ)n
P
for all n ≥ N . Since 0 < a + ǫ < 1, it follows that the series ∞ n=N an converges, which in turn
P∞
implies that n=1 an converges.
If a > 1, then choose an ǫ > 0 such that a − ǫ > 1. Then |an |1/n > a − ǫ for infinitely
many n’s. This means |an | > (a − ǫ)n > 1 for infinitely many n’s. But then one can not have
P
limn→∞ an = 0. So the series an can not converge. ✷
  P
Theorem 5.33 (Ratio test) If lim sup  an+1
 
an  < 1, then an converges absolutely.
  P
If  an+1
 
an  > 1 for all but finitely many n’s, then an does not converge.
 
Proof : Write a = lim sup  an+1
 
an . If a < 1, then we can choose an ǫ > 0 such that a + ǫ < 1.
   
Since a = lim sup  an+1
   an+1 
an  , there is an integer N such that  an  < a + ǫ for all n ≥ N . It
then follows that for all n ≥ N , one has |an | ≤ |aN |(a + ǫ)n−N = const · (a + ǫ)n . Hence by
P P∞
comparison test, the series ∞ n=N |an | converges. Therefore so does
 the series n=1 |an |.
By the given condition, there is an integer N such that  an+1
 
an  > 1 for all n ≥ N . This in
P
particular implies that |an | > |aN | for all n > N . Therefore lim an can not be 0. So an does
not converge. ✷

xn
P `m+n´
Exercise 5.34 Investigate the behaviour of the following series: { xn , where m ∈ N,
P
n!
}, n
nα xn .
P

5.7 Power series

P
A series of the form an xn is known as a power series. An application of the root test
tells us that this series converges absolutely if lim sup |an xn |1/n < 1, and does not converge if
lim sup |an xn |1/n > 1. In other words, the series converges whenever |x| < lim sup1|a |1/n , and
n
fails to converge if |x| > lim sup1|a |1/n .
n

30
Exercise 5.35 Show that the following power series converge for any value of x:
xn
P∞
1. n=0 n! ,
n x2n
P∞
2. n=0 (−1) (2n)! ,
P∞ n x2n+1
3. n=0 (−1) (2n+1)!
.

Define
∞
X xn
exp x = ,
n=0
n!
∞
X x2n
cos x = (−1)n ,
(2n)!
n=0
∞
X x2n+1
sin x = (−1)n ,
n=0
(2n + 1)!

where the right hand sides are the sums of the corresponding series.

5.8 Product of two series

P P∞ P
Proposition 5.36 Let ∞ n=0 an and n=0 bn be two series such that an is absolutely con-
P P P P
vergent and bn is convergent. Let α = an and β = bn . Then the product series cn ,
Pn
given by cn = k=0 ak bn−k , converges and the sum is αβ.
P P P
Proof : Write An = nk=0 ak , Bn = nk=0 bk , and Cn = nk=0 ck . Then
n
X
Cn = ck
k=0
n X
X k
= ar bk−r
k=0 r=0
Xn X n
= ar bk−r
r=0 k=r
Xn n−r
X
= ar bj
r=0 j=0
Xn
= ar Bn−r
r=0
Xn
= ar (Bn−r − β) + βAn .
r=0
P
Since limn→∞ An = α, it is now enough to show that limn→∞ nr=0 ar (Bn−r − β) = 0.
P
Take an ǫ > 0. Let ξ be the sum of the series |an |. Choose an integer N such that
ǫ
|Bk − β| < ξ for k ≥ N . Then for n > N , we have
n
X n−N
X n
X
| ar (Bn−r − β)| ≤ |ar (Bn−r − β)| + |ar | · |Bn−r − β|
r=0 r=0 r=n−N +1

31
 n
X N
X −1
≤ |an−s | · |Bs − β| + |an−s | · |Bs − β|
s=N s=0
n N −1
ǫ X X
≤ |an−s | + |an−s | · |Bs − β|
ξ s=0
s=N
N −1
ǫ X
≤ ·ξ+ |an−s | · |Bs − β|
ξ s=0
N
X −1
≤ ǫ+ |an−s | · |Bs − β|.
s=0

The second term on the right hand side converges to 0 as n tends to ∞. Hence
n
X
lim sup | ar (Bn−r − β)| ≤ ǫ.
n→∞
r=0
Pn
Since this is true for any ǫ > 0, it follows that lim supn→∞ | r=0 ar (Bn−r − β)| = 0, which
P
implies that limn→∞ | nr=0 ar (Bn−r − β)| = 0. ✷

Exercise 5.37 Show that the nth term of the product of the two series
P P
an and bn , where an = bn =
(−1)n √1n , does not converge to 0.

P P
The above example illustrates that if both the series an and bn converge conditionally,
then the product series need not converge.

Exercise 5.38 Show that

1. exp(x + y) = exp x exp y.
2. sin(x + y) = sin x cos y + sin y cos x,
3. sin2 x + cos2 x = 1.

32
 Week 7

6 Limit and continuity

6.1 Limit of a function
Let E ⊆ R, and let f : E → R be a function. Let a ∈ R be a limit point of E. One says that
that the limit of f as x approaches a (in symbol x → a) exists and is equal to ℓ (in
symbols limx→a f (x) = ℓ) if for every ǫ > 0, there exists a δ > 0 such that |f (x) − ℓ| < ǫ
whenever 0 < |x − a| < δ and x ∈ E (i. e. whenever x ∈ (E\{a}) ∩ (a − δ, a + δ)).
One says that the limit of f as x → a is ∞ (and writes limx→a f (x) = ∞) if given any
M ∈ R, there exists a δ > 0 such that f (x) > M whenever x ∈ (E\{a}) ∩ (a − δ, a + δ).
Similarly one writes limx→a f (x) = −∞ if given any M ∈ R, there exists a δ > 0 such that
f (x) < M whenever x ∈ (E\{a}) ∩ (a − δ, a + δ).
Next assume that the set E is not bounded above. Then one says that the limit of f as x
approaches ∞ (in symbol x → ∞) exists and is equal to ℓ (in symbols limx→∞ f (x) = ℓ)
if for every ǫ > 0, there exists a N ∈ R such that |f (x) − ℓ| < ǫ whenever x ∈ E ∩ [N, ∞).
One writes limx→∞ f (x) = ∞ if given any M ∈ R, there exists a N ∈ R such that f (x) > M
whenever x ∈ E ∩ [N, ∞). One writes limx→∞ f (x) = −∞ if given any M ∈ R, there exists a
N ∈ R such that f (x) < M whenever x ∈ E ∩ [N, ∞).

Exercise 6.1 Define the meaning of the following expressions: limx→−∞ f (x) = ℓ, limx→−∞ f (x) = ∞
limx→−∞ f (x) = −∞

Exercise 6.2 Let c ∈ R and define f by f (x) = c for all x. Show that limx→a f (x) = limx→∞ f (x) =
limx→−∞ f (x) = c.

Proposition 6.3 Let f and g be two functions defined on a set E, with limx→a f (x) = ℓ and
limx→a g(x) = m. Then

1. limx→a(f (x) + g(x)) = ℓ + m,

2. limx→a(f (x)g(x)) = ℓm,

3. if m 6= 0, then there is a neighbourhod (b, c) containing the point a such that g(x) 6= 0 for
all x ∈ (b, c) ∩ (E\{a}), and limx→a fg(x)
(x)
= mℓ .

33
Proof : We will prove part 3 here. Parts 1 and 2 will be left as exercises.
Since limx→a g(x) = m 6= 0, there is a δ1 > 0 such that whenever x ∈ (a − δ1 , a + δ1 )\{a},
one has |g(x)−m| < |m| |m| f (x)
2 . But then |g(x)| > 2 > 0. Thus g(x) is defined on (a−δ1 , a+δ1 )\{a}.

 f (x) ℓ
Take any ǫ > 0. We have to choose a δ > 0 such that  g(x) − m  < ǫ whenever 0 < |x − a| <
δ. Now    
 f (x) ℓ   mf (x) − ℓg(x) 
 g(x) − m  =  .

mg(x)
Hence for 0 < |x − a| < δ1 , one has
 
 f (x) ℓ  2
 g(x) − m  ≤ |mf (x) − ℓg(x)|

|m|2
2
≤ (|m||f (x) − ℓ| + |ℓ||g(x) − m|).
|m|2

Let δ2 > 0 and δ3 be such that 0 < |x − a| < δ2 implies |f (x) − ℓ| < |m|ǫ
4 and 0 < |x − a| < δ3
|m|2 ǫ
implies |g(x) − m| < 4|ℓ| . Let δ := min{δ1 , δ2 , δ3 }. Then for 0 < |x − a| < δ we have
 
 f (x)
 g(x) − mℓ  < ǫ.

✷

Exercise 6.4 Prove that the above proposition continues to hold even when a, ℓ and m are allowed to take
values in the set of extended reals, provided the right hand sides are defined.

Proposition 6.5 Let f be a function defined on a set E and let limx→a f (x) = ℓ. If f (x) ≥ 0
for all x ∈ E, then ℓ ≥ 0.

Proof : Exercise. ✷

Exercise 6.6 Let f and g be two functions defined on a set E, with limx→a f (x) = ℓ and limx→a g(x) = m.
If f (x) ≤ g(x) for all x ∈ E, then show that ℓ ≤ m.

Proposition 6.7 Let f , g and h be functions defined on a set E, with limx→a f (x) = ℓ =
limx→a g(x). If f (x) ≤ h(x) ≤ g(x) for all x ∈ E, then limx→a h(x) = ℓ.

Proof :
✷

Exercise 6.8 Prove or disprove: Let f : R → R and g : R → R be two functions. Assume that limx→a g(x) =
b and limx→b f (x) = ℓ. Then limx→a f (g(x)) = ℓ.

6.2 Computation of some limits

Lemma 6.9 limx→0 sin x = 0.

34
Proof : Let sn (x) denote the nth partial sum of the series for sin x. Then for |x| < 1,
n
X |x|2r+1
|sn (x)| ≤
(2r + 1)!
r=0
n
X 1
≤ |x|
r=0
(2r + 1)!
≤ |x| exp(1).
Therefore
| sin x| ≤ |x| exp(1) for |x| < 1.
ǫ
Now take any ǫ > 0. Choose δ = min{ exp(1) , 1}. Then for |x| < δ, one has | sin x| < ǫ. ✷

Lemma 6.10 limx→0 cos x = 1.

Proof : Let sn (x) denote the nth partial sum of the series for cos x. Then for |x| < 1,
n
X |x|2r
|sn (x) − 1| ≤
r=1
(2r)!
n
2
X 1
≤ |x|
(2r)!
r=1
2
≤ |x| exp(1).
Therefore
| cos x − 1| ≤ |x|2 exp(1) for |x| < 1.
ǫ
Now take any ǫ > 0. Choose δ = min{ exp(1) , 1}. Then for |x| < δ, one has | cos x − 1| < ǫ. ✷
 
Lemma 6.11 limx→0  sinx x − 1 = 0.

Proof : As before, let sn (x) denote the nth partial sum of the series for sin x. Then for 0 <
|x| < 1,
n
sn (x) X |x|2r
| − 1| = (−1)r
x (2r + 1)!
r=1
n
2
X 1
≤ |x|
r=1
(2r + 1)!
2
≤ |x| exp(1).
Therefore
sin x
| − 1| ≤ |x|2 exp(1) for 0 < |x| < 1.
x
Hence the result. ✷

Exercise 6.12 Show that

1. limx→0 exp x = 1.
2. limx→0 ˛ exp x−1−x
˛ ˛
˛ = 0.
x
˛ exp x ˛
3. limx→0 ˛ xn ˛ = ∞.

35
6.3 Limits via limits of sequences
Proposition 6.13 Let f be a function defined on a set E. Then limx→a f (x) = ℓ if and only if
for any sequence {xn } in E (with xn 6= a for all n) converging to a, one has limn→∞ f (xn ) = ℓ.

Proof : Observe that a is a limit point of the set E and hence there will always exist sequences
with terms in E and different from a that converge to a.
Assume limx→a f (x) = ℓ. Let {xn } be a sequence in E converging to a with none of the
terms equal to a. Let ǫ > 0 be given. Since limx→a f (x) = ℓ, there exists a δ > 0 such that
|f (x) − ℓ| < ǫ for 0 < |x − a| < δ. Since limn→∞ xn = a, corresponding to this δ, there exists an
integer N such that |xn − a| < δ if n ≥ N . Since none of the terms of the sequence {xn } is a,
it follows that 0 < |xn − a| < δ for n ≥ N . Therefore whenever n ≥ N , one has |f (xn ) − ℓ| < ǫ.
Thus limn→∞ f (xn ) = ℓ.
Now let us prove the converse. Assume limx→a f (x) 6= ℓ. This means that there is an ǫ > 0
such that for every δ > 0, there is an x ∈ E with 0 < |x − a| < δ and |f (x) − ℓ| ≥ ǫ. In
particular, for each n ∈ Z+ , there is an xn with 0 < |xn − a| < n1 and |f (xn ) − ℓ| ≥ ǫ. But then
{xn } is a sequence with terms from E different from a, and limn→∞ xn = a, but |f (xn ) − ℓ| ≥ ǫ
for all n, so that limn→∞ f (xn ) 6= ℓ. ✷

Exercise 6.14 State and prove analogous statements where the quantities a and ℓ are allowed to be ±∞.

6.4 Continuity
A function f defined on a set E is said to be continuous at a point a ∈ E if limx→a f (x) =
f (a). It is said to be continuous everywhere (or just continuous) if it is continuous at every
point a ∈ E.

Proposition 6.15 A function f defined on a set E is continuous at a point a ∈ E if and only

if for any sequence {xn } in E converging to a, one has limn→∞ f (xn ) = f (a).

Proof : If for any sequence {xn } in E converging to a, one has limn→∞ f (xn ) = f (a), then it
follows from the earlier result that limx→a f (x) = f (a), i. e. f is continuous at a.
Now to prove the converse, assume that f is continuous at a. Take a sequence {xn } such
that xn ∈ E for all n and limn→∞ xn = a. Let ǫ > 0 be given. We want to show that there
is an integer N such that for n ≥ N , one has |f (xn ) − f (a)| < ǫ. Since limx→a f (x) = f (a),
there is a δ > 0 such that |f (x) − f (a)| < ǫ whenever |x − a| < δ. Again, since limn→∞ xn = a,
corresponding to this δ, there is an integer N such that for n ≥ N , we have |xn − a| < δ.
Combining these two observations, it follows that for n ≥ N , we have |f (xn ) − f a)| < ǫ. ✷

Exercise 6.16 Show that a constant function and the function x 7→ x are continuous everywhere.

Proposition 6.17 Let f and g be two functions defined in some neighbourhood of a point
a ∈ R. Assume f and g are continuous at a. Then f + g and f g are continuous at a.
If, moreover, g(a) 6= 0, then f /g is continuous at a.

36
Proof : The proof follows from the corresponding properties of limits. ✷

Proposition 6.18 Let f, g : R → R be two functions. Assume that g is continuous at a and f

is continuous at g(a). Then f ◦ g is continuous at a.

Proof : Take an ǫ > 0. Since f is continuous at g(a), there is a δ′ > 0 such that |f (y)−f (g(a))| <
ǫ whenever |y − g(a)| < δ′ . Again, since g is continuous at a, given this δ′ , there is a δ > 0
such that |g(x) − g(a)| < δ′ whenever |x − a| < δ. Therefore for |x − a| < δ, we have
|f (g(x)) − f (g(a))| < ǫ. ✷

Exercise 6.19 Show that a polynomial function is continuous everywhere.

Exercise 6.20 Show that the function exp is continuous at the point 0. Now use the equality exp(x + y) =
exp(x) exp(y) for all x and y to prove that the exponential function is continuous everywhere.

Exercise 6.21 Show that the function sin is continuous at the point 0. Now use the equality sin(x + y) =
sin x cos y + cos x sin y for all x and y to prove that the sine function is continuous everywhere.
Arguing along similar lines, show that the cosine function is also continuous everywhere.

Proposition 6.22 Let p ≥ 0. Then the function x 7→ xp is continuous on [0, ∞).

If p < 0, then it is continuous on (0, ∞).

Proof : It is enough to prove continuity at the point x = 0 and x = 1 (why?), i. e. it is enough

to show that limx→0+ xp = 0 and limx→0 (1 + x)p = 1. Let us first prove the second equality.
If p ∈ N, then it follows by using the binomial expansion of (1 + x)p . Since p = [p] + (p − [p]),
it is now enough to prove the above for 0 ≤ p < 1. So assume 0 ≤ p < 1. Then for x > 0,
1 < (1 + x)p < 1 + x and for −1 < x < 0, one has 1 + x < (1 + x)p < 1. Combining these, one
gets 1 − |x| < (1 + x)p < 1 + |x| whenever x > −1. Therefore it follows that limx→0 (1 + x)p = 1.
1
Let us now show that limx→0+ xp = 0. Take ǫ > 0. Choose δ = ǫ p . Then for 0 < x < δ,
one has 0 < xp < δp = ǫ.
1
The case p < 0 now follows from the observation that xp = x−p . ✷

Proposition 6.23 Let p > 0. Then the function x 7→ px is continuous.

Proof : It is enough to prove the result for p > 1. Let a ∈ R. Then |px − pa | = pa · |px−a − 1|.
Therefore it is enough to prove continuity at x = 0.
1
Take any ǫ > 0. Choose an ǫ′ > 0 such that 1 − ǫ < 1+ǫ ′
′ and ǫ < ǫ. Since limn→∞ p
1/n = 1,

it follows that there is an integer N such that |p1/N − 1| < ǫ′ . Since p > 1, one in fact has
1 < p1/N < 1 + ǫ′ . Hence for any x ∈ (0, N1 ), one has 1 < px < p1/N < 1 + ǫ′ . Consequently, for
x ∈ (− N1 , 0), we have 1+ǫ
1 x 1 x
′ < p < 1. Thus whenever |x| < N , we have p ∈ (1 − ǫ, 1 + ǫ). ✷

Exercise 6.24 Show that the function f given by

exp(− x12 )

if x 6= 0,
f (x) =
0 if x = 0
is continuous

37
6.5 Properties of continuous functions
Proposition 6.25 Let f : R → R be continuous, and assume that f (x) = 0 for all x ∈ Q.
Then f (x) = 0 for all x ∈ R.

Proof : Take an x ∈ R. Since the closure of Q is R, there is a sequence {xn } such that xn ∈ Q
for all n and limn→∞ xn = x. Since f is continuous at x, we have f (x) = limn→∞ f (xn ). But
f (xn ) = 0 for all n. Hence f (x) = 0. ✷

Exercise 6.26 Let f and g be two continuous functions such that f (x) = g(x) for all x ∈ Q. Show that
f (x) = g(x) for all x ∈ R.

Exercise 6.27 Let f : R → R be continuous. Let E be a subset of R such that Ē = R. If f (x) = 0 for all
x ∈ E, then show that f (x) = 0 for all x ∈ R.

Proposition 6.28 Let f be a function defined on a set E. Assume f is continuous at a point

a ∈ E and f (a) 6= 0. Then there is a neighbourhood (b, c) around a such that the sign of f (x)
for every x ∈ (b, c) ∩ E is same as that of f (a).

Proof : Take ǫ = |f (a)|

2 . Since f (a) 6= 0, ǫ is positive. Using continuity of f at a, there is a
δ > 0 such that |f (x) − f (a)| < ǫ whenever |x − a| < δ and x ∈ E. Thus if f (a) > 0 then
for all x ∈ (a − δ, a + δ) ∩ E, one has f (a)2 < f (x) < 3f2(a) , and if f (a) < 0, then for all
x ∈ (a − δ, a + δ) ∩ E, one has 3f2(a) < f (x) < f (a)
2 . ✷

Theorem 6.29 Let f be continuous on an interval [a, b] and let f (a) and f (b) be of opposite
signs. Then there is a point c ∈ (a, b) such that f (c) = 0.

Proof : Assume f (a) < 0 and f (b) > 0. Let E = {x ∈ [a, b] : f (x) > 0}. Since f (b) > 0, this is
nonempty. Let c := inf E.
First, observe that by the sign-preserving property, there is a δ > 0 such that f (x) > 0 for
all x ∈ (b − δ, b], so that c ≤ b − δ < b. Similarly, there is a δ′ > 0 such that f (x) < 0 for all
x ∈ [a, a + δ′ ). Hence for any y ∈ E, we have y ≥ a + δ′ . Therefore c ≥ a + δ′ > a. Thus
c ∈ (a, b).
Since c = inf E, there is a sequence {cn } converging to c such that cn ∈ E for all E. By
continuity of f at c, it follws that f (c) ≥ 0. If f (c) > 0, by the sign-preserving property, there
is a δ > 0 such that f (x) > 0 for all x ∈ (c − δ, c + δ) ∩ [a, b]. Since c ∈ (a, b), this in particular
implies that there is a c′ < c such that c′ ∈ [a, b] ∩ E. But then c can not be inf E. Therefore
we must have f (c) = 0. ✷

Theorem 6.30 Let f be continuous on a closed bounded interval [a, b]. Given any ǫ > 0, there
is a partition
x0 = a < x1 < . . . < xn = b
of [a, b] such that

over each subinterval of the form [ak , ak+1 ], one has supx,y |f (x) − f (y)| < ǫ. (6.2)

38
Proof : Assume that the conclusion of the theorem is not true. Then there exists an ǫ > 0 for
which there is no partition satisfying the condition stated above. Now apply an argument sim-
ilar to the one used in the proof of the Bolzano-Weierstrass theorem to produce two sequences
{an } and {bn } such that

an − bn
a0 = a, b0 = b, an ≤ an+1 , bn+1 ≤ bn , bn+1 − an+1 = ,
2
and for each n, [an , bn ] does not admit any partition that satisfies the condition stated in the
theorem. It follows that the sequences an and bn both converge to a common point c. Use
continuity of f at c to get a δ > 0 such that |f (x) − f (c)| < 3ǫ for x ∈ (c − δ, c + δ). Now choose
n large enough so that |an − c| < δ and |bn − c| < δ. Then [an , bn ] ⊆ (c − δ, c + δ). Hence for any
x ∈ [an , bn ], we have |f (x) − f (c)| < 3ǫ . Therefore sup{|f (x) − f (y)| : x, y ∈ [an , bn ]} ≤ 2ǫ
3 < ǫ.
This contradicts the choice of the an ’s and bn ’s. ✷

Corollary 6.31 A continuous function on a closed bounded interval is bounded.

Proof : Get a partition x0 = a < x1 < . . . < xn = b of the interval [a, b] on which f is
continuous such that one has sup{|f (x) − f (y)| : x, y ∈ [xk , xk+1 ]} < 1 for each k. Write
m = min{f (xi ) : 1 ≤ i ≤ n} and M = max{f (xi ) : 1 ≤ i ≤ n}. Then it follows that
m − 1 ≤ f (x) ≤ M + 1 for all x ∈ [a, b]. ✷

Proposition 6.32 Let f be continuous on a closed bounded interval [a, b]. Then there are
points c, d ∈ [a, b] such that f (c) ≤ f (x) ≤ f (d) for all x ∈ [a, b].

Proof : Let E := {f (x) : x ∈ [a, b]}, m := inf E and M := sup E. Then m ≤ f (x) ≤ M for all
x ∈ [a, b]. We will prove the existence of the point c with f (c) = m. Proof of existence of d
with f (d) = M is similar and is left as an exercise.
There is a sequence with terms in E that converges to m. The sequence will be of the
form {f (xn )}, where each xn comes from [a, b]. Since the sequence {xn } is bounded, it has a
convergent subsequence, say {xnk }k . Write c = limk→∞ xnk . Then clearly c ∈ [a, b] and since
f is continuous, limk→∞ f (xnk ) = f (c). Bt we know that the sequence {f (xn )} converges to
m. So we must have f (c) = m. ✷

39
 Week 8

7 Some important functions

7.1 Trigonometric functions
1. Show that the functions sin and cos are continuous everywhere.

2. Let Z := {x ∈ R+ : cos x = 0}. Show that this is nonempty by showing that cos 2 < 0.
Let ξ := inf Z. Show that cos ξ = 0. Show that cos x > 0 for 0 ≤ x ≤ 1. Hence conclude
that 1 < ξ < 2.

3. Prove that
cos 2nξ = (−1)n , sin 2nξ = 0 for all n ∈ Z.

cos(2ξ + x) = − sin x, sin(2ξ + x) = − cos x.

cos(4ξ + x) = cos x, sin(4ξ + x) = sin x.

4. Show that 
 >0 for 0 ≤ x < ξ,


= 0 for x = ξ,


cos x < 0 for ξ < x < 3ξ,




 =0 for x = 3ξ,

>0 for 3ξ < x ≤ 4ξ.

5. Since sin2 x + cos2 x = 1 for all x, one has sin2 ξ = 1. Hence sin ξ is either +1 or −1.
Suppose if possible sin ξ = −1. Then cos(ξ+x) = cos ξ cos x−sin ξ sin x = sin x. Therefore
whenever 0 < x < 2ξ, one must have sin x < 0. But it is easy to see that sin 1 > 0. Sine
1 < ξ, this is impossible. Hence we must have sin ξ = 1. It now follows from the equality
cos(ξ + x) = sin x and the properties of the cosine function that

 = 0 if x = 0,


 > 0 for 0 < x < 2ξ,


sin x = 0 for x = 2ξ,




 < 0 for 2ξ < x < 4ξ,

= 0 for x = 4ξ.

40
 6. Show that < 1
 for 0 < x < 4ξ,

cos x 
0 ≤ x < 2ξ,

 > −1 for
2ξ < x ≤ 4ξ
 
 0 ≤ x < ξ,
< 1 for ,


ξ < x ≤ 4ξ
sin x 

 0 ≤ x < 3ξ,
 > −1 for

3ξ < x ≤ 4ξ
(Since | sin x| > 0 for x ∈ (0, 4ξ)\ {2ξ}, one must have | cos x| < 1 for these values of x)

7. Use the above parts to show that



 =1 for x = 0,

 ∈ (0, 1) for 0 ≤ x < ξ,






 =0 for x = ξ,


 ∈ (−1, 0) for ξ < x < 2ξ,
cos x

 = −1 for x = 2ξ,




 ∈ (−1, 0) for 2ξ < x < 3ξ,




 =0 for x = 3ξ,


∈ (0, 1) for 3ξ < x < 4ξ.


 =0 for x = 0,



 ∈ (0, 1) for 0 ≤ x < ξ,




 =1 for x = ξ,


 ∈ (0, 1) for ξ < x < 2ξ,
sin x

 =0 for x = 2ξ,




 ∈ (−1, 0) for 2ξ < x < 3ξ,




 = −1 for x = 3ξ,


∈ (−1, 0) for 3ξ < x < 4ξ.

Standard notation for the number 2ξ in mathematics is π.

Exercise 7.1 Show that for 0 < |x| < 1, one has cos x < sin x
x
< 1.

7.2 Exponential function

Show that

1. Write e := exp(1). Show that en = exp(n) for all n ∈ Z. Conclude that er = exp(r) for
any r ∈ Q.
Now use continuity of the two functions x 7→ ex and x 7→ exp(x) to prove that ex = exp(x)
for all x ∈ R.

2. limx→∞ exp x = ∞.

41
 3. exp x > 0 for x ≥ 0. Conclude that exp x > 0 for all x ∈ R.

4. limx→−∞ exp x = 0.

5. exp x < exp y whenever x < y.

Thus exp is a bijection from R onto (0, ∞). The inverse of this function, from (0, ∞) to R, is
called the logarithm function, and is denoted by log.
Show that

1. log(xy) = log x + log y for all x, y ∈ (0, ∞),

2. if 0 < x < y then log x < log y,

3. limx→0+ log x = −∞,

4. limx→∞ log x = ∞.

5. Let p > 0. Then px = exp(x log p) for all x ∈ R.

42
 Week 9

8 Differentiability
8.1 Derivatives
Let f : R → R be a function defined on some neighbourhood of a point a. If the limit
limx→a f (x)−f
x−a
(a)
exists and is finite, then we say the function f is differentiable at the point
a. The number limx→a f (x)−f x−a
(a)
is then called the derivative of the function f at the point
′
a and is denoted by f (a). If a function f is differentiable at every point x in a set E, one
says that f is differentiable on E. If a function f is differentiable on R, one says that f is
differentiable.
If a function f is defined on an interval of the form [a, b) and the limit limx→a+ f (x)−f x−a
(a)

exists and is finite, then this number is called the right derivative of f at a. Similarly if f
is defined on an interval of the form (c, a] and the limit limx→a− f (x)−f
x−a
(a)
exists and is finite,
then it is called the left derivative of f at a. Thus a function is differentiable at a point if
and only if the right and the left derivatives at that point exist and are equal.

Exercise 8.1 Let n ∈ N and f be the function given by f (x) = xn . Show that
nan−1

if n > 0,
f ′ (a) =
0 if n = 0
for any a ∈ R.

Proposition 8.2 If a function f is differentiable at a point a, then it is continuous at a.

Proof : Since the limit limx→a f (x)−f

x−a
(a)
is finite, there is a δ > 0 such that | f (x)−f
x−a
(a)
| is bounded
(by, say, M ) on (a − δ, a + δ)\{a}. Hence for any x ∈ (a − δ, a + δ), we have |f (x) − f (a)| ≤
M |x − a|. Given any ǫ > 0 now, choose δ′ = min{δ, ǫ/M }. Then for x ∈ (a − δ′ , a + δ′ ), one
has |f (x) − f (a)| ≤ M |x − a| < M δ′ ≤ ǫ. ✷
A function that is continuous at a point need not be differentiable there. Look at the
function x 7→ |x| for example. It is continuous but not differentiable at 0.

Proposition 8.3 Let f and g be two functions, both differentiable at a point a. Then f ± g
and f g are also differentiable at a, and one has

(f ± g)′ (a) = f ′ (a) ± g′ (a), (f g)′ (a) = f ′ (a)g(a) + f (a)g′ (a).

43
 f
If g(a) 6= 0, then g is also differentiable at a, and
 ′
f f ′ (a)g(a) − f (a)g′ (a)
(a) = .
g g(a)2

Proof : Exercise. ✷

Proposition 8.4 Let f, g : R → R be two functions such that f is differentiable at a and g

is differentiable at f (a). Then the composition g ◦ f is differentiable at a and (g ◦ f )′ (a) =
g′ (f (a))f ′ (a).

Proof : Since g is differentiable at f (a), there is an ǫ > 0 such that g is defined over (f (a) −
ǫ, f (a) + ǫ). Since f is differentiable at a, in particular, it is defined in some neighbourhood of
a and is also continuous at a. Therefore one can get hold of a δ > 0 such that f is defined over
(a − δ, a + δ) and for any x ∈ (a − δ, a + δ), one has f (x) ∈ (f (a) − ǫ, f (a) + ǫ). Thus g ◦ f is
defined over (a − δ, a + δ).
Next, define a function h on (f (a) − ǫ, f (a) + ǫ) as follows:
(
g(y)−g(f (a))
y−f (a) if y 6= f (a),
h(y) =
′
g (f (a)) if y = f (a).

The function h is then continuous on (f (a) − ǫ, f (a) + ǫ), and one has
g(f (x)) − g(f (a)) f (x) − f (a)
= h(f (x)) ·
x−a x−a
for all x ∈ (a − δ, a + δ). Hence
g(f (x)) − g(f (a))
(g ◦ f )′ (a) = lim
x→a x−a
= h(f (a))f ′ (a)
= g′ (f (a))f ′ (a).

Proposition 8.5 Let f : R → R be a continuous function. Assume that f has a local maximum
at a point c, and that f is differentiable at c. Then f ′ (c) = 0.

Proof : Assume if possible that f ′ (c) > 0. Then there is a δ > 0 such that for 0 < |x − c| < δ,
one has | f (x)−f
x−c
(c)
− f ′ (c)| < 12 f ′ (c). Thus for x ∈ (c, c + δ), one has f (x)−f
x−c
(c)
> f ′ (c) − 12 f ′ (c) =
1 ′
2 f (c) > 0. Therefore f (x) > f (c) for all x ∈ (c, c + δ). This contradicts the fact that f has a
local maximum at c.
A similar argument shows that one can not have f ′ (c) < 0. Hence f ′ (c) = 0. ✷

Exercise 8.6 Let f : R → R be a continuous function. Assume that f has a local minimum at a point c,
and that f is differentiable at c. Then show that f ′ (c) = 0.

44
8.2 Mean value theorem and applications
Theorem 8.7 (Rolle’s theorem) Let f : [a, b] → R be continuous. Assume that f ′ (x) exists
for x ∈ (a, b) and f (a) = f (b). Then there is a c ∈ (a, b) such that f ′ (c) = 0.

Proof : Let m = inf{f (x) : x ∈ [a, b]} and M = sup{f (x) : x ∈ [a, b]}. If m = M , then f has to
be a constant function. Therefore for any c ∈ (a, b), f ′ (c) = 0. So let us now look at the case
m 6= M . Since f is continuous, there are points c, d ∈ [a, b] such that f (c) = m and f (d) = M .
Since m 6= M , at least one of m and M has to be different from f (a) = f (b). Assume m 6= f (a).
Then c 6= a and c 6= b. So c ∈ (a, b). Since f ′ (c) exists, we must have f ′ (c) = 0. Similarly if
M 6= f (a), then d ∈ (a, b) and f ′ (d) = 0. ✷

Theorem 8.8 (Mean value theorem) Let f : [a, b] → R be continuous. Assume that f ′ (x)
exists for x ∈ (a, b). Then there is a c ∈ (a, b) such that f (b) − f (a) = f ′ (c)(b − a).

Proof : Apply the previous result to the function g(x) = (x−a)(f (b)−f (a))−(b−a)(f (x)−f (a)).
✷

Theorem 8.9 (Generalized mean value theorem) Let f, g : [a, b] → R be continuous.

Assume that both are differentiable on (a, b). Then there is a c ∈ (a, b) such that g′ (c)(f (b) −
f (a)) = f ′ (c)(b − a).

Proof : Apply Rolle’s theorem to the function g(x) = (g(x) − g(a))(f (b) − f (a)) − (g(b) −
g(a))(f (x) − f (a)). ✷

Maxima/minima of a function.

Theorem 8.10 Let f : R → R be a function. Assume that f ′ (x) ≥ 0 for x ∈ (a, b) and
f ′ (x) ≤ 0 for x ∈ (b, c). Assume also that f is continuous at b. Then f (x) ≤ f (b) for all
x ∈ (a, c).

Proof : Take any x ∈ (a, b). Use the mean value theorem over the interval [x, b] to get a
s ∈ (x, b) ⊆ (a, b) such that f (b) − f (x) = f ′ (s)(b − x). Since x < b and 0 ≤ f ′ (s), we have
f (x) ≤ f (b). Next take any y ∈ (b, c). This time use mean value theorem over the interval [b, y]
to get a point t ∈ (b, y) ⊆ (b, c) such that f (y) − f (b) = f ′ (t)(y − b). Since b < y and f ′ (t) ≤ 0,
one has f (y) ≤ f (b). ✷

L’Hopital’s rule. The next two theorems are very useful for evaluating limits in many
situations and go by the name L’Hopital’s rule.

Theorem 8.11 Let f and g be two functions from (a, b) to R such that

1. they are differentiable on (a, b),

45
 2. g′ (x) 6= 0 for x ∈ (a, b),
f ′ (x)
3. limx→a+ g ′ (x) = L.

4. limx→a+ f (x) = 0 = limx→a+ g(x).

f (x)
Then limx→a+ g(x) = L.

Proof : First let us look at the case when L ∈ R. Since g′ (x) 6= 0 for x ∈ (a, b), it follows from
′ (x)
the mean value theorem that g(x) 6= g(y) whenever x 6= y. Since limx→a+ fg′ (x) = L, given
′
ǫ > 0, there there is a δ > 0 such that | fg′ (x)
(x)
− L| < 2ǫ for a < x < a + δ. Take any x ∈ (a, a + δ).
Then by the generalized mean value theorem, for any y ∈ (a, a + δ), there is a z between x and
y and hence in (a, a + δ) such that f ′ (z)(g(x) − g(y)) = g′ (z)(f (x) − f (y)). Since g′ (x) 6= 0
for x ∈ (a, b), it follows from the mean value theorem that g(x) 6= g(y) whenever x 6= y. Thus
′ (z)
we have fg(x)−g(y)
(x)−f (y)
= fg′ (z) , where z ∈ (a, a + δ). Hence | fg(x)−g(y)
(x)−f (y)
− L| < 2ǫ . Next observe that
g(x) 6= 0 (see exercise 8.12 below). Therefore taking limit as y → a+ and using condition 4,
we get | fg(x)
(x)
− L| ≤ 2ǫ < ǫ. Thus limx→a+ fg(x) (x)
= L.
The cases L = ±∞ are similar and are left as exercises. ✷

Exercise 8.12 Show that under the assumtions 1, 2 and 4 above, g(x) 6= 0 for all x ∈ (a, b).

Theorem 8.13 Let f and g be two functions from (a, b) to R such that

1. they are differentiable on (a, b),

2. g′ (x) 6= 0 for x ∈ (a, b),

f ′ (x)
3. limx→a+ g ′ (x) = L.

4. limx→a+ g(x) = ∞.
f (x)
Then limx→a+ g(x) = L.

Proof : Let us first consider the case when L ∈ R. Since g′ (x) 6= 0 for x ∈ (a, b), it follows
that g(x) 6= g(y) for x 6= y in (a, b). Also, by condition 4, there is
 ′a point c ∈ (a, b) such that
g(x) > 0 for all x ∈ (a, c). Now, first choose a δ > 0 such that  fg′ (x)
 (x)
− L < 2ǫ and g(x) > 0


for x ∈ (a, a + δ). Next, take any y ∈ (a, a + 2δ ). Then by generalized mean value theorem, we
have  
 f (y) − f (a + δ )  ǫ
 2 
 − L < ,
 g(y) − g(a + 2δ )  2
i. e.
ǫ f (y) − f (a + 2δ )
L− <
2 g(y) − g(a + 2δ )

46
 δ
Now choose a ξ > 0 such that ξ < 2 and g(x) > g(a + 2δ ) for x ∈ (a, a + ξ). Then we have, for
y ∈ (a, a + ξ),

ǫ ǫ g(a + 2δ ) f (y) − f (a + δ2 ) ǫ ǫ g(a + 2δ ) ǫ

L− − (L − ) < < L + − (L + ) <L+ .
2 2 g(y) g(y) 2 2 g(y) 2

Next, choose a ξ ′ > 0 such that ξ ′ ≤ ξ and the following two inequalities hold:
 

 ǫ  g(a + 2δ ) ǫ  f (a + δ )  ǫ
 2 
L −  < ,  < .
2 g(y) 4  g(y)  4

Then, for y ∈ (a, a + ξ ′ ), we will have L − ǫ < fg(y)

(y)
< L + ǫ.
Proofs for the cases L = ±∞ are left as exercises. ✷

Taylor’s theorem.

Theorem 8.14 Let f : R → R be a function, n be a positive integer and let α and β be two
real numbers. Assume that f (n) (x) exists for all x ∈ (α, β) and f (n−1) is continuous over [α, β].
Then for any two points a, b ∈ [α, β], there is a point c between a and b such that

(b − a)2 (b − a)n−1 (b − a)n

f (b) = f (a) + f ′ (a)(b − a) + f ′′ (a) + · · · + f (n−1) (a) + f (n) (c) .
2! (n − 1)! n!

Proof : Define two functions g and h as follows:

(b − x)n−1
g(x) = f (x) + f ′ (x)(b − x) + · · · + f (n−1) (x) , h(x) = (b − x)n
(n − 1)!

and apply the generalized mean value theorem. ✷

The following exercise outlines another proof of the above result.
2 n−1
Exercise 8.15 Write M = f (b)−f (a)−f ′ (a)(b−a)−f ′′ (a) (b−a)
2!
−· · ·−f (n−1) (a) (b−a)
(n−1)!
. Show that M =
n n−1
f (n) (c) (b−a)
n!
for some c by completing the following steps. Let P (t) = f (a)+f ′ (a)(t−a)+· · ·+f (n−1) (a) (t−a)
(n−1)!
M n
and g(t) = f (t) − P (t) − (b−a) n (t − a) . Show that

1. g(a) = 0 = g(b) and g (k) (a) = 0 for 1 ≤ k ≤ n − 1.

2. there is some c between a and b such that g (n) (c) = 0.
3. g (n) (t) = f (n) (t) − n!M
(b−a)n
.

47
 Week 10

8.3 Problems
1. Check whether the following functions are differentiable:

|x sin x|, |x cos x|, |x|3 .

2. Assume a1 , . . . , an be reals such that a1 + · · · + an = 0. Show that the equation a1 +

2a2 x + 3a3 x2 + · · · + nan xn−1 = 0 has a real root in (0, 1).

3. Let f : R → R be differentiable, and limx→∞ f ′ (x) = 0. Define an = f (n + 1) − f (n).

Show that the sequence {an } converges to 0.

4. Let f : R → R be a continuous function such that f ′ (x) exists for all x 6= 0 and
limx→0 f ′ (x) = 1. Show that f is differentiable at 0 and f ′ (0) = 1.

5. Let f : (0, ∞) → R be differentiable, and |f ′ (x)| < 1 for x ∈ (0, ∞). Define an = f ( n1 ).
Show that the sequence {an } converges.

6. Let f : [0, ∞) → R be a continuous function such that it is differentiable on (0, ∞),

f (0) = 0 and f ′ is increasing on (0, ∞). Show that the function x 7→ f (x)
x is increasing
over (0, ∞).

7. Define 
exp(− x12 ) if x 6= 0,
f (x) =
0 if x = 0.
Show that

(a) f is continuous,
(b) f ′ (x) exists for all x, and f ′ (0) = 0,
(c) for any n ∈ Z+ , 
(n) Pn ( x1 ) exp(− x12 ) if x 6= 0,
f (x) =
0 if x = 0,
where Pn is some polynomial.

48
 8. Let 
x2 sin( x1 ) if x 6= 0,
f (x) =
0 if x = 0.
′
Show that f is differentiable everywhere but f is not continuous at 0.

9. Prove that

(a) | sin x − sin y| ≤ |x − y| for all x, y ∈ R.

(b) if 0 < x < y and n ∈ Z+ , then nxn−1 (y − x) < y n − xn < ny n−1 (y − x).

9 Integration
9.1 Riemann integrals
Let [a, b] be a closed bounded interval. Recall that a finite subset P of [a, b] containing both
the endpoints a and b is called a partition of [a, b]. Suppose P = {x0 , x1 , . . . , xn } and a =
x0 < x1 < · · · < xn = b (we would often express this by saying that ‘let P : x0 < x1 < · · · < xn
be a partition of [a, b]’). Then the intervals [xi , xi+1 ] are called subintervals coming from
the partition P . Let P and Q be two partitions of the same interval [a, b]. Then P is said
to be finer than Q if Q ⊆ P . If P is finer than Q, then any subinterval from Q can be
written as a finite union of subintervals from P . Let P be a partition of [a, b]. The number
max{xi+1 − xi : 0 ≤ i ≤ n − 1} is called the length of the partition P and is denoted by |P |.
Let f : [a, b] → R be a bounded function and let P : x0 < x1 < · · · < xn be a partition of
[a, b]. Define two numbers U (f, [a, b], P ) and L(f, [a, b], P ) as follows:
n−1
X 
U (f, [a, b], P ) = sup{f (x) : x ∈ [xi , xi+1 ]} (xi+1 − xi ),
i=0

n−1
X 
L(f, [a, b], P ) = inf{f (x) : x ∈ [xi , xi+1 ]} (xi+1 − xi ).
i=0

When there is no ambiguity about the interval [a, b] in question, we will just write U (f, P ) and
L(f, P ) respectively for the above quantities. It is clear from the definitions that L(f, P ) ≤
U (f, P ) for any partition P of [a, b].

Lemma 9.1 Let P1 and P2 be two partitions of [a, b], and assume P1 is finer than P2 . Then

U (f, P1 ) ≤ U (f, P2 ), L(f, P1 ) ≥ L(f, P2 ).

Lemma 9.2 Let P1 and P2 be any two partitions of [a, b]. Then

L(f, P1 ) ≤ U (f, P2 ).

49
Proof : Let P = P1 ∪ P2 . Then P is finer than both P1 and P2 . Hence L(f, P1 ) ≤ L(f, P ) ≤
U (f, P ) ≤ U (f, P2 ). ✷
Next, let us define the upper integral of f over the interval [a, b] to be inf{U (f, [a, b], P ) :
P is a partition of [a, b]}. We will denote this number by U (f, [a, b]) or, if there is no confusion
about the interval [a, b], by just U (f ). Similarly define the lower integral of f over [a, b] by
sup{L(f, [a, b], P ) : P is a partition of [a, b]} and denote it by L(f, a, b]) or by simply L(f ).

Exercise 9.3 Show that L(f ) ≤ U (f ).

Exercise 9.4 Define a function f : [0, 1] → R as follows


1 if x is rational,
f (x) =
0 if x is irrational.
Compute U (f, [0, 1]) and L(f, [0, 1]).

Exercise 9.5 Let f : [0, 1] → R be the function given by f (x) = x. Compute U (f, [0, 1]) and L(f, [0, 1]).

We say that a function f is Riemann integrable over the interval [a, b] if U (f, [a, b]) =
Rb
L(f, a, b]). The common value is denoted by a f (x)dx and is referred to as the (Riemann)
integral of f over [a, b].

Exercise 9.6 Let f : [a, b] → R be a bounded function with f (x) ≥ 0 for all x. Show that both U (f ) and
L(f ) are nonnegative.
Use this to show that if f and g are two integrable functions over [a, b] such that f (x) ≤ g(x) for all x ∈ [a, b],
Rb Rb
then a f (x)dx ≤ a g(x)dx.

Rb
Proposition 9.7 Let f : [a, b] → R be bounded. The integral a f (x)dx exists if and only if
for every ǫ > 0, there is a partition P of [a, b] such that U (f, P ) − L(f, P ) < ǫ.

Proof : First assume that f is integrable, i. e. U (f ) = L(f ). Take an ǫ > 0. Then there exist
partitions P1 and P2 such that U (f, P1 ) < U (f ) + 2ǫ and L(f, P2 ) > L(f ) − 2ǫ . Let P = P1 ∪ P2 .
Then P is finer than both P1 and P2 . Hence U (f, P ) ≤ U (f, P1 ) and L(f, P ) ≥ L(f, P2 ).
Combining this with our earlier observation, we get U (f, P ) − L(f, P ) ≤ U (f, P1 ) − L(f, P2 ) <
U (f ) − L(f ) + ǫ = ǫ.
Now to prove the converse, assume that given any ǫ > 0, one can choose a partition P with
U (f, P ) − L(f, P ) < ǫ. Since for any partition P , U (f ) ≤ U (f, P ) and L(f ) ≥ L(f, P ), we have
U (f ) − L(f ) ≤ U (f, P ) − L(f, P ). Hence for any ǫ > 0, we have U (f ) − L(f ) ≤ ǫ. So we must
have U (f ) − L(f ) ≤ 0, which implies that U (f ) = L(f ), i. e. f is integrable. ✷

Exercise 9.8 Let f : [a, b] → R be a bounded function and let P : x0 = a < x1 < · · · < xn = b be a partition
of [a, b]. Show that
n−1
X
U (f, P ) − L(f, P ) = sup{f (x) − f (y) : x, y ∈ [xi , xi+1 ]}(xi+1 − xi )
i=0
n−1
X
= sup{|f (x) − f (y)| : x, y ∈ [xi , xi+1 ]}(xi+1 − xi ).
i=0

50
Proposition 9.9 Let f and g be two functions on [a, b]. Assume that both are integrable. Let
c ∈ R. Then the functions cf and f ± g are also integrable over [a, b].

Proof : The case c = 0 is trivial because U (cf ) = L(cf ) = 0 in that case. So assume c 6= 0. Let
P be a partition of [a, b]. Then by the exercise above,
n−1
X
U (cf, P )−L(cf, P ) = sup{|cf (x)−cf (y)| : x, y ∈ [xi , xi+1 ]}(xi+1 −xi ) = |c|(U (f, P )−L(f, P )).
i=0

So given any ǫ > 0, choose a partition P such that U (f, P ) − L(f, P ) < cǫ . Then one has
U (cf, P ) − L(cf, P ) < ǫ.
Next, observe that for any partition P , U (f + g, P ) ≤ U (f, P ) + U (g, P ) and L(f + g, P ) ≥
L(f, P ) + L(g, P ). Hence U (f + g, P ) − L(f + g, P ) ≤ U (f, P ) − L(f, P ) + U (g, P ) − L(g, P ).
Now given ǫ > 0, choose partitions P1 and P2 such that U (f, P1 ) − L(f, P1 ) < 2ǫ and U (g, P2 ) −
L(g, P2 ) < 2ǫ . Let P = P1 ∪ P2 . Then

U (f + g, P ) − L(f + g, P ) ≤ U (f, P ) − L(f, P ) + U (g, P ) − L(g, P )

≤ U (f, P1 ) − L(f, P1 ) + U (g, P2 ) − L(g, P2 )
< ǫ.

Proposition 9.10 Let f : [a, b] → R be integrable. Then the functions |f | and f 2 are inte-
grable.

Proof : Let P : x0 = a < x1 < · · · < xn = b be a partition of [a, b]. Then

n−1
X
U (|f |, P ) − L(|f |, P ) = sup{|f (x)| − |f (y)| : x, y ∈ [xi , xi+1 ]}(xi+1 − xi ).
i=0

Since |s| − |t| ≤ |s − t| for any s, t ∈ R, one gets

n−1
X
U (|f |, P ) − L(|f |, P ) ≤ sup{|f (x) − f (y)| : x, y ∈ [xi , xi+1 ]}(xi+1 − xi )
i=0
n−1
X
= sup{f (x) − f (y) : x, y ∈ [xi , xi+1 ]}(xi+1 − xi )
i=0
= U (f, P ) − L(f, P ).

Hence integrability of f implies that of |f |.

Next,
n−1
X
U (f 2 , P ) − L(f 2 , P ) = sup{(f (x))2 − (f (y))2 : x, y ∈ [xi , xi+1 ]}(xi+1 − xi ).
i=0

51
Now

(f (x))2 − (f (y))2 = (|f (x)| + |f (y)|)(|f (x)| − |f (y)|)

≤ (|f (x)| + |f (y)|)|f (x) − f (y)|.

Let M be a positive real such that |f (x)| ≤ M for x ∈ [a, b]. Then we have

(f (x))2 − (f (y))2 ≤ (|f (x)| + |f (y)|)|f (x) − f (y)| ≤ 2M |f (x) − f (y)|.

Hence
n−1
X
U (f 2 , P ) − L(f 2 , P ) ≤ 2M sup{|f (x) − f (y)| : x, y ∈ [xi , xi+1 ]}(xi+1 − xi )
i=0
= 2M (U (f, P ) − L(f, P )).

ǫ
Thus, given ǫ > 0, choose a partition P such that U (f, P ) − L(f, P ) < 2M . Then it follows
that U (f 2 , P ) − L(f 2 , P ) < ǫ. ✷

Exercise 9.11 Show that if f and g are two integrable functions on [a, b], then the product f g is also
integrable.

Theorem 9.12 Let f : [a, b] → R be monotone. Then f is integrable over [a, b].

Proof : Let us first look at the case when f is increasing. Then for any interval of the form
[s, t] ⊆ [a, b], we have

sup{f (x) : x ∈ [s, t]} = f (t), inf{f (x) : x ∈ [s, t]} = f (s).

Hence for any partition P : x0 = a < x1 < · · · < xn = b of [a, b],

n−1
X
U (f, P ) − L(f, P ) = (f (xi+1 ) − f (xi ))(xi+1 − xi ).
i=0

Therefore if we take xi to be a + i b−a

n , so that the differences xi+1 − xi are all equal to
1
n, then

n−1
1X f (b) − f (a)
U (f, P ) − L(f, P ) = (f (xi+1 ) − f (xi )) = .
n n
i=0

Now given ǫ > 0, choose n large enough so that f (b)−f

n
(a)
< ǫ. Then U (f, P ) − L(f, P ) < ǫ.
If f is decreasing and integrable, then −f is increasing and integrable. So the result in this
case follows from the earlier case. ✷

Theorem 9.13 Let f : [a, b] → R be continuous. Then f is integrable over [a, b].

52
Proof : Take any ǫ > 0. Since f is continuous over [a, b], there is a partition P : x0 = a < x1 <
· · · < xn = b such that for every i,
ǫ
sup{|f (x) − f (y)| : x, y ∈ [xi , xi+1 ]} < .
b−a
But then
n−1
X 
U (f, P ) − L(f, P ) = sup{|f (x) − f (y)| : x, y ∈ [xi , xi+1 ]} (xi+1 − xi )
i=0
n−1
ǫ X
< (xi+1 − xi )
b−a
i=0
= ǫ.

Exercise 9.14 Let f : [a, b] → R be a bounded function. Let c ∈ (a, b), and suppose P1 and P2 are two
partitions of [a, c] and [c, b] respectively. Let P = P1 ∪ P2 . Show that P is a partition of [a, b], and

U (f, P ) = U (f, P1 ) + U (f, P2 ), L(f, P ) = L(f, P1 ) + L(f, P2 ).

Theorem 9.15 Let f : [a, b] → R be bounded. Assume that f has only finitely many disconti-
nuities on [a, b]. Then f is integrable.

Proof : Take an ǫ > 0. We want to produce a partition P of [a, b] such that U (f, P )−L(f, P ) < ǫ.
Let d1 , d2 , . . . , dk be the discontinuity points of f , indexed in such a way that d1 < d2 < · · · < dk .
Let M = sup{?? : x ∈ [a, b]}. Let ξ > 0 be a positive number such that the intervals
[a, d1 + ξ], [d2 − ξ, d2 + ξ], . . . , [dk − ξ, b] are all disjoint. Then for any i, 1 ≤ i ≤ k, we have
sup{f (x) − f (y) : x, y ∈ [di − ξ, di + ξ] ∩ [a, b]} ≤ 2M . Now, first take ξ = 8Mǫ k . Observe that
f is integrable over the intervals

[a, d1 − ξ], [d1 + ξ, d2 − ξ], [d2 + ξ, d3 − ξ], . . . , [dk−1 + ξ, dk − ξ], [dk + ξ, b]

(omit the first interval if d1 −ξ ≤ a and omit the last if dk +ξ ≥ b). So These intervals will admit
ǫ
partitions P0 , P1 , . . . , Pk respectively such that for each i, one has U (f, Pi ) − L(f, Pi ) < 2k+2 .
Finally, let P = P0 ∪ . . . ∪ Pk ∪ {a, b}. Then
k 
X 
U (f, P ) − L(f, P ) = U (f, Pi ) − L(f, Pi )
i=0
k
X
+ sup{f (x) − f (y) : x, y ∈ [di − ξ, di + ξ] ∩ [a, b]}(di + ξ − di − ξ)
i=1
k
ǫ X
≤ (k + 1) + 2M 2ξ
2k + 2
i=1
ǫ ǫ
< + 4M k
2 8M k
= ǫ.

53
 Week 11

Exercise 9.16 Let f be integrable over [a, b]. Let g : [a, b] → R be another function such that f (x) = g(x)
Rb Rb
for all but one point in [a, b]. Show that g is integrable over [a, b] and a
g= a
f.

Theorem 9.17 (Lebesgue) Let f : [a, b] → R be a bounded function. Then f is Riemann

integrable over [a, b] if and only if the set of discontinuity points of f in [a, b] is of measure
zero.

Proposition 9.18 Let f be integrable over [a, b]. Then

Z b  Z b
 
 f (x)dx ≤ |f (x)|dx.
a a

Proof : We have already seen that if f is integrable over [a, b], then so is |f |. Next, observe that
Rb
for any partition P , we have a f ≤ U (f, P ) ≤ U (|f |, P ). Hence taking infimum over P , one
gets the required inequality. ✷

Proposition 9.19 Let f be integrable over [a, b] and [c, d] be a subinterval of [a, b]. Then f is
integrable over [c, d].

Proof : Take ǫ > 0. Choose a partition P of [a, b] such that U (f, P ) − L(f, P ) < ǫ. Let
P ′ := P ∪ {c, d} and let P1 , P2 and P3 be the restrictions of P ′ to [a, c], [c, d] and [d, b]
respectively. Then

U (f, P ′ ) = U (f, P1 ) + U (f, P2 ) + U (f, P3 ), L(f, P ′ ) = L(f, P1 ) + L(f, P2 ) + L(f, P3 ).

Therefore
U (f, P2 ) − L(f, P2 ) ≤ U (f, P ′ ) − L(f, P ′ ),
and since P ′ is finer than P , U (f, P ′ ) − L(f, P ′ ) ≤ U (f, P ) − L(f, P ). Hence

U (f, P ′ ) − L(f, P ′ ) < ǫ.

Proposition 9.20 Let f be a function on [a, b] and let c ∈ (a, b). Then f is integrable over
[a, b] if and only if f is integrable over both [a, c] and [c, b]. Moreover, in such a case, one has
Z b Z c Z b
f= f+ f.
a a c

54
Proof : If f is integrable over [a, b], then by the previous result it is integrable over both [a, c]
and [c, b]. To prove the converse, assume that f is integrable over [a, c] and [c, b]. Take an
ǫ > 0. Choose partitions P1 of [a, c] and P2 of [c, b] such that
ǫ ǫ
U (f, P1 ) − L(f, P1 ) < , U (f, P2 ) − L(f, P2 ) < .
2 2
Now take P to be the partition P1 ∪ P2 of [a, b]. Then

U (f, P ) − L(f, P ) = U (f, P1 ) − L(f, P1 ) + U (f, P2 ) − L(f, P2 ) < ǫ,

so that f is integrable over [a, b].

Now assume that f is integrable over [a, b] (and hence in [a, c] and on [c, b]). Take any
partition P1 of [a, c] and P2 of [c, b]. Then P := P1 ∪ P2 is a partition of [a, b]. Now U (f, P1 ) +
Rb Rc Rb Rb
U (f, P2 ) = U (f, P ) ≥ a f . Hence taking infimum ver P1 and P2 , we get a f + c f ≥ a f .
Rb
Again, L(f, P1 ) + L(f, P2 ) = L(f, P ) ≤ a f . This time taking supremum over P1 and P2 , one
Rc Rb Rb
gets a f + c f ≤ a f . ✷

Proposition 9.21 Let f : [a, b] → R be integrable. Then given any ǫ > 0, there is a δ > 0
such that if P is any partition with |P | < δ, and ci ∈ [xi , xi+1 ], then
n−1
X Z b
| f (ci )(xi+1 − xi ) − f (x)dx| < ǫ.
i=0 a

Rb
Proof : Write I = a f (x)dx. Take an ǫ > 0.Then there is a partition P1 : a = y0 < y1 · · · <
ym = b such that U (f, P1 ) < I + 2ǫ . Let n be the number of elements in P1 , and M be
sup{|f (x)| : x ∈ [a, b]}. Now let δ1 = 2Mǫ n and let P : a = x0 < x1 < · · · < xn = b be a
partition with |P | < δ1 . We will show that U (f, P ) < I + ǫ. Break the sum in the expression
for U (f, P ) into two parts as follows:
n−1
X  
U (f, P ) = sup f (x) xi+1 − xi
i=0 x∈[xi ,xi+1 ]
X   X  
= sup f (x) xi+1 − xi + sup f (x) xi+1 − xi
i:P1 ∩(xi ,xi+1 )=φ x∈[xi ,xi+1 ] i:P1 ∩(xi ,xi+1 )6=φ x∈[xi ,xi+1 ]
m−1
X X  
= sup f (x) xi+1 − xi
j=0 i:[xi ,xi+1 ]⊆[yj ,yj+1 ] x∈[xi ,xi+1 ]
X  
+ sup f (x) xi+1 − xi .
i:P1 ∩(xi ,xi+1 )6=φ x∈[xi ,xi+1 ]

Denote the first term by A and the second term by B. Then

m−1  
X ǫ
A≤ sup f (x) yj+1 − yj = U (f, P1 ) < I + ,
x∈[yj ,yj+1 ] 2
j=0

55
and
X ǫ
B≤ M (xi+1 − xi ) ≤ M · |P | · n < .
2
i:P1 ∩(xi ,xi+1 )]6=φ

Thus U (f, P ) < I + ǫ.

Applying this now to the function −f , one gets a δ2 > 0 such that whenever P is a
partition with |P | < δ2 , then U (−f, P ) < −I + ǫ. This, along with the observation that
U (−f, P ) = −L(f, P ), imply that L(f, P ) < I − ǫ. Now take δ = min{δ1 , δ2 }. Then for
partitions with |P | < δ, we have

I − ǫ < L(f, P ) ≤ U (f, P ) < I + ǫ.

P
Since L(f, P ) ≤ i f (ci )(xi+1 − xi ) ≤ U (f, P ), the result follows. ✷

Exercise 9.22 Let f : [a, b] → R be integrable. Show that

Z b n−1
b−a X b−a
f = lim f (a + k ).
a n→∞ n k=0 n

Theorem 9.23 (First mean value theorem) Let f : [a, b] → R be continuous. Then there
Rb
is a c ∈ [a, b] such that a f (x)dx = f (c)(b − a).

Proof : Let m = inf{f (x) : x ∈ [a, b]} and M = sup{f (x) : x ∈ [a, b]}. Since f is continuous on
[a, b], there exist points s and t in [a, b] such that f (s) = m and f (t) = M . Also observe that
Rb
since m ≤ f (x) ≤ M for all x ∈ [a, b], we have m(b − a) ≤ a f ≤ M (b − a), i. e.
Z b
1
m≤ f ≤ M.
b−a a

Therefore by intermediate value theorem, there is some point c between s and t (and hence in
1
Rb
the interval [a, b]) such that f (c) = b−a a f. ✷

Theorem 9.24 (First fundamental theorem) Let f : [a, b] → R be integrable. Define

Rx
F (x) = a f (t)dt for x ∈ [a, b]. If f is continuous at a point c ∈ [a, b], then F is differen-
tiable at c and F ′ (c) = f (c).

Proof : Take any ǫ > 0. Using continuity of f at c, choose a δ > 0 such that |t − c| < δ implies
|f (t) − f (c)| < 2ǫ . Then for 0 < |h| < δ, we have
Z
 F (c + h) − F (c)


 1  c+h 

 − f (c) =  (f (t) − f (c))dt
h |h| c
ǫ
≤ < ǫ.
2
✷

56
Corollary 9.25 (Change of variable) Let f and g be two functions. Assume that g is dif-
ferentiable on some interval containing [a, b] and g′ is continuous on [a, b]. Assume also that f
is continuous on g([a, b]).Then
Z b Z g(b)
′
f (g(t))g (t)dt = f (t)dt.
a g(a)

Proof : Define three functions F , G and H as follows:

Z x Z x
F (x) = f (t)dt, G(x) = f (g(t))g′ (t)dt, H(x) = F ◦ g(x).
g(a) a

Then H ′ (x) − G′ (x) = F ′ (g(x))g′ (x) − f (g(x))g′ (x) = f (g(x))g′ (x) − f (g(x))g′ (x) = 0 for all
x ∈ [a, b]. Therefore H(x) − G(x) is constant over [a, b]. Since H(a) = F (g(a)) = 0 and
G(a) = 0, we have G(x) = H(x) for all x ∈ [a, b]. Hence G(b) = H(b), which is what we needed
to prove. ✷

Theorem 9.26 (Second fundamental theorem) Let f be differentiable on some interval

Rb
containing [a, b]. Assume f ′ is integrable over [a, b]. Then a f ′ (x)dx = f (b) − f (a).

Proof : Take any partition P : a = x0 < x1 < · · · < xn = b of [a, b]. For each i, 0 ≤ i ≤ n − 1,
by the mean value theorem for derivatives, there is a ci ∈ [xi , xi+1 ] such that f (xi+1 ) − f (xi ) =
f ′ (ci )(xi+1 − xi ). Hence
n−1
X n−1
X
f ′ (ci )(xi+1 − xi ) = f (xi+1 ) − f (xi ) = f (b) − f (a).
i=0 i=0
Pn−1
But we also have L(f ′ , P ) ≤ i=0 f ′ (ci )(xi+1 − xi ) ≤ U (f ′ , P ). Thus for any partition P of
[a, b], we have L(f ′ , P ) ≤ f (b) − f (a) ≤ U (f ′ , P ). Since f ′ is integrable over [a, b], we must
Rb
have a f ′ = f (b) − f (a). ✷

Corollary 9.27 (integration by parts) Let f and g be two functions, both differentiable on
some interval containing [a, b] and assume that f ′ g and f g′ are both integrable over [a, b]. Then
Z b Z b
′
f (x)g (x)dx + f ′ (x)g(x)dx = f (b)g(b) − f (a)g(a).
a a

Proof : Apply the second fundamental theorem to the function φ(x) := f (x)g(x). ✷

Theorem 9.28 (Weighted mean value theorem) Let f, g : [a, b] → R be continuous. As-
sume that g does not change sign over [a, b]. Then there is a c ∈ [a, b] such that
Z b Z b
f (x)g(x)dx = f (c) g(x)dx.
a a

57
Proof : Since g does not change sign, either g(x) ≥ 0 for all x ∈ [a, b] or g(x) ≤ 0 for all
x ∈ [a, b]. Let us first assume that g(x) ≥ 0 for all x ∈ [a, b]. Let m = inf{f (x) : x ∈ [a, b]} and
M = sup{f (x) : x ∈ [a, b]}. Since f is continuous, there are points s and t in [a, b] such that
f (s) = m and f (t) = M . Now for all x ∈ [a, b], m ≤ f (x) ≤ M . Hence mg(x) ≤ f (x)g(x) ≤
M g(x) for x ∈ [a, b]. Integrating over [a, b], we get
Z b Z b Z b
m g≤ fg ≤ M g.
a a a
Rb Rb
Now if a g = 0, then a f g is also 0 and hence for any choice of c ∈ [a, b], the required equality
Rb
holds. So assume a g > 0. Then
Rb
fg
m ≤ Ra b ≤ M.
a g
Rb
fg
Since f is continuous, there is a point c between s and t (hence in [a, b]) such that f (c) = Rab .
a g
Proof for the other case when g(x) ≤ 0 for all x ∈ [a, b] is similar and is left as an exercise.
✷

Theorem 9.29 (second mean value theorem) Let g be continuous and f be diferentiable
over some interval containing [a, b]. Assume that f ′ does not change sign and is integrable on
[a, b]. Then there is some c ∈ [a, b] such that
Z b Z c Z b
f (x)g(x)dx = f (a) g(x)dx + f (b) g(x)dx.
a a c
Rx
Proof : Define G(x) = a g(t)dt for x ∈ [a, b]. Since g is continuous, G′ (x) = g(x) for all
Rb
x ∈ [a, b]. Hence the integral on the left hand side can be written as a f (x)G′ (x)dx. Since f ′
does not change sign, f is monotone and hence integrable. Apply integration by parts formula
to get Z Z
b b
f (x)G′ (x)dx = f (b)G(b) − f (a)G(a) − G(x)f ′ (x)dx.
a a
Next apply the weighted mean value theorem to the integral on the right hand side above to
get a point c ∈ [a, b] such that
Z b Z b
′
G(x)f (x)dx = G(c) f ′ (x)dx.
a a
Rb
By the second fundamental theorem, a f ′ (x)dx = f (b) − f (a). Hence combining these obser-
vations, one gets
Z b
f (x)G′ (x)dx = f (b)G(b) − f (a)G(a) − G(c)(f (b) − f (a))
a
   
= f (a) G(c) − G(a) + f (b) G(b) − G(c)
Z c Z b
= f (a) g(x)dx + f (b) g(x)dx.
a c
✷

58
9.2 Improper Riemann integrals
Let a ∈ R. A function f is said to be improper Riemann integrable over [a, ∞) if for any
Rb
b ≥ a, f is integrable over [a, b], and if the limit limb→∞ a f exists and is finite. The integral
of f over [a, ∞) in such a case is defined to be the above limit. One also uses the phrase ‘the
R∞
integral a f converges’ to mean that f is improper Riemann integrable over [a, ∞).
Similarly a function f is improper Riemann integrable over an interval (−∞, b] if it is
Rb
integrable over all intervals of the form [a, b] for a ≤ b and if the limit lima→−∞ a f exists and
Rb
is finite. The limit is then denoted by −∞ f and is called the integral of f over (−∞, b].

Exercise 9.30 Let a ∈ R and assume that f is improper Riemann integrable over both (−∞, a] and [a, ∞).
Show that f is improper Riemann integrable over (−∞, b] and [b, ∞) for any b ∈ R and
Z a Z ∞ Z b Z ∞
f+ f= f+ f.
−∞ a −∞ b

If there is an a ∈ R such that f is improper Riemann integrable over both (−∞, a] and [a, ∞),
R∞
then one says that f is improper Riemann integrable over (−∞, ∞) (or −∞ f converges) and
R∞ Ra R∞
one defines the integral −∞ f to be the sum −∞ f + a f .
The type of improper integrals defined above are called improper integrals of the first
kind.

Exercise 9.31 Let f : [a, ∞) → R. Show˛ that˛

R∞
a
f converges if and only if for every ǫ > 0, there is an
˛R c ˛
M ∈ R such that whenever b, c > M , one has ˛ b f ˛ < ǫ.

Proposition 9.32 Let f : [a, ∞) → R be nonnegative and integrable over [a, b] for any b ≥ a.
Rb R∞
Assume that there is an M > 0 such that a f ≤ M for all b ≥ a. Then a f converges.

Proposition 9.33 (Comparison test) Let f, g : [a, ∞) → R be two functions integrable over
R∞
any closed bounded subinterval and assume that 0 ≤ f (x) ≤ g(x) for all x. If a g converges,
R∞ R∞ R∞
then a f also converges and if a f diverges, then a g also diverges.
R∞ R∞
Corollary 9.34 If a |f | converges, then a f also converges.

Exercise 9.35 Let f : [a, b] → R be bounded. Assume that f is integrable on [c, b] for any c ∈ (a, b]. Show
that f is integrable on [a, b] and
Z b Z b
f = lim f.
a c→a+ c

Let f : R → R be a function. We will call a point d a point of infinite discontinuity

for f if there are points a < d < b such that sup{|f (x)| : x ∈ [a, b]} = ∞, but for any ǫ > 0,
sup{|f (x)| : x ∈ [a, b]\(d − ǫ, d + ǫ)} < ∞.

Exercise 9.36 Suppose d is a point of infinite discontinuity of a function f . Show that there is a sequence
{xn } converging to d such that the sequence {|f (xn )|} diverges to ∞.

59
 Let f : [a, b] → R be a function such that a is a point of infinite discontinuity of f but f
Rb
is integrable on [c, b] for any c ∈ (a, b]. If the limit limc→a+ c f exists and is finite, then one
Rb
says that f is improper Riemann integrable over [a, b] and the integral a f is defined to be the
above limit. Similarly if b is a point of infinite discontinuity of f in [a, b], f is integrable over
Rc
[a, c] for all c ∈ [a, b) and if the limit limc→b− a f exists and is finite, then one says that f is
Rb
improper Riemann integrable over [a, b] and the integral a f is defined to be this limit.

60
 Week 12–13

10 Sequences and series of functions

Let E be a subset of R. A sequence of functions on E is a map from N to the set of functions
from E to R. If n is mapped to the function fn for each n ∈ N, then we just say that we have a
sequence of functions {fn }. One says that a sequence of functions {fn } converges pointwise
to a function f if for each x ∈ E, one has limn→∞ fn (x) = f (x).

Examples.

1. Let {cn } be a sequence of reals. Define

fn (x) = cn , x ∈ R.

If {cn } converges to c, then {fn } converges pointwise to the constant function x 7→ c.

2. Let fn (x) = xn for x ∈ [0, 1]. Then {fn } converges pointwise to the function f on [0, 1]
given by 
0 if 0 ≤ x < 1,
f (x) =
1 if x = 1.
3. Let
x2n
fn (x) = , x ∈ R.
1 + x2n
The sequence fn converges pointwise to the function f given by

 0 if −1 < x < 1,
f (x) = 12 if x = ±1,

1 if |x| > 1.

4. Let
1
fn (x) = |x|1+ n , x ∈ (−1, 1).
The sequence fn converges pointwise to the function x 7→ |x| on (−1, 1).

5. Let 
0 if 0 ≤ x < 1 − n1 ,
fn (x) = n2 (x − 1 + 1 ) if 1 − 1 ≤ x < 1,
 n n
0 if x = 1.
The sequence fn converges pointwise to the constant function x 7→ 0.

61
 6. Let Q∩[0, 1] = {r1 , r2 , . . .}. Define fn to be the indicator function of the set {r1 , r2 , . . . , rn }.
The sequence fn converges pointwise to the function f on [0, 1] given by

1 if x ∈ Q,
f (x) =
0 otherwise.

7. Let
n
X xk
fn (x) = , x ∈ R.
k!
k=0

The sequence fn converges pointwise to the function x 7→ exp x.

8. Let
n
X x2k
fn (x) = (−1)k , x ∈ R.
(2k)!
k=0

The sequence fn converges pointwise to the function x 7→ cos x.

9. Let
n
X x2k+1
fn (x) = (−1)k , x ∈ R.
(2k + 1)!
k=0

The sequence fn converges pointwise to the function x 7→ sin x.

10. Let
sin nx
fn (x) = √ , x ∈ R.
n
The sequence fn converges pointwise to the constant function x 7→ 0.

Suppose we have a sequence of functions {fn }, defined on some commom domain E ⊆ R,

that converges to a function f pointwise. We will now try to see what properties of the functions
fn carry over to the limit function f . More specifically, the questions we will try to answer are:

1. if each fn is continuous, is f also continuous?

2. if each fn is differentiable, is f also differentiable? and whether or not the sequence {fn′ }
converge?

3. if each fn is integrable over some fixed interval [a, b], is f also integrable over [a, b]? If
Rb Rb
yes, can one relate the number a f with the numbers a fn ?

A quick look at the examples above will tell us that the answer to all these questions
are negative (look at examples 2 and 3 for question 1, examples 4 and 10 for question 2 and
examples 5 and 6 for question 3). Therefore what we do next is change the notion of convergence
a little bit and try to answer the above questions for this new convergence.
A sequence of functions fn defined on a set E is said to converge to a function f
uniformly if the sequence {dn } given by dn := supx∈E |fn (x) − f (x)| converges to 0.

62
Exercise 10.1 Show that a sequence of functions fn defined on a set E converges uniformly to a function f if
and only if given any ǫ > 0, there exists a natural number N such that whenever n ≥ N , one has |fn (x)−f (x)| < ǫ
for all x ∈ E.

Proposition 10.2 Let fn be a sequence of functions defined on a set E. Assume that fn

converges uniformly to a function f . Then fn converges to f pointwise.

A sequence of functions {fn } defined on a set E is said to be uniformly Cauchy if

for any given ǫ > 0, there is a natural number N such that whenever m, n ≥ N , one has
supx∈E |fn (x) − fm (x)| < ǫ.
Exercise 10.3 Let fn be a sequence of functions that converge uniformly to a function f on a set E. Show
that {fn } is uniformly Cauchy.

Next, we will show that just like in the case of a sequence of real numbers, the converse of the
above statement is also true.

Exercise 10.4 Let {fn } be uniformly Cauchy over some set E. Let c ∈ E. Show that the sequence {fn (c)}
is Cauchy.

It follows from the above exercise that if {fn } is uniformly Cauchy over E, then for any
c ∈ E, the sequence {fn (c)} converges. Define a function f on E by the prescription f (x) =
limn→∞ fn (x). Then the sequence fn converges to the function f pointwise. The next result
tells us that the sequence {fn } actually converges uniformly.

Proposition 10.5 Let {fn } be uniformly Cauchy. Then fn converges to some function f
uniformly.

Proof : Define f by
f (x) = lim fn (x), x ∈ E.
n→∞
We have seen already that {fn } converges to f pointwise. Next, we will show that {fn }
converges to this f uniformly. Take an ǫ > 0. Choose an N such that supx∈E |fn (x)−fm (x)| < 2ǫ
whenever m, n ≥ N . Take an n ≥ N . Then for any x ∈ E and m ≥ N , we have
ǫ
|fn (x) − fm (x)| < .
2
Taking limit as m → ∞ in the above inequality, one gets
ǫ
|fn (x) − f (x)| ≤ .
2
ǫ
Since this is true for all x ∈ E, we have supx∈E |fn (x) − f (x)| ≤ 2 < ǫ. ✷
Now let us go back to the first of the three questions that we wanted to answer.

Proposition 10.6 Let fn be a sequence of functions defined on an interval [a, b]. Assume that
the sequence converges uniformly to a function f and each fn is continuous at a point c ∈ [a, b].
Then f is also continuous at c.

63
Proof : We have to show that given any ǫ > 0, there exists a δ > 0 such that |x − c| < δ (and
x ∈ [a, b]) implies |f (x) − f (c)| < ǫ. Now for any n,

|f (x) − f (c)| = |f (x) − fn (x) + fn (x) − fn (c) + fn (c) − f (c)|

≤ |f (x) − fn (x)| + |fn (x) − fn (c)| + |fn (c) − f (c)|.

Since the sequence {fn } converges to f uniformly, we can choose an N such that supx |fn (x) −
f (x)| < 3ǫ for n ≥ N . Now use the earlier inequality for n = N to get
ǫ
|f (x) − f (c)| ≤ 2 + |fN (x) − fN (c)|.
3
Use continuity of fN at c to get a δ > 0 such that |x − c| < δ (and x ∈ [a, b]) implies
|fN (x) − fN (c)| < 3ǫ . ✷
Essentially the same proof works for the following slightly more general statement.

Proposition 10.7 Let fn be a sequence of functions defined on an interval [a, b]. Assume that
fn converges to f uniformly. Let c ∈ [a, b]. Asume that the limits an := limx→c fn (x) exist and
the sequence {an } converges to a real a. Then limx→c f (x) = a.

Proof : Use the inequality

|f (x) − a| = |f (x) − fn (x) + fn (x) − an + an − a|

≤ |f (x) − fn (x)| + |fn (x) − an | + |an − a|.

✷
Next we turn to integrability.

Proposition 10.8 Let fn be a sequence of functions that converge uniformly to a function f

on an interval [a, b]. Assume that each fn is integrable on [a, b]. Then f is also integrable on
[a, b] and
Z b Z b
f = lim fn .
a n→∞ a

Proof : First let us show that f is integrable over [a, b]. Take an ǫ > 0. Choose an N such that
ǫ
supx |fn (x)−f (x)| < 4(b−a) for n ≥ N . Next, use integrability of fN to get a partition P of [a, b]
such that U (fN , P ) − L(fN , P ) < 2ǫ . Since for all x ∈ [a, b], we have |fN (x) − f (x)| < 4(b−a)
ǫ
, it
ǫ ǫ
follows that U (f, P ) ≤ U (fN , P ) + 4 and L(f, P ) ≥ L(fN , P ) − 4 . Hence
ǫ
U (f, P ) − L(f, P ) ≤ U (fN , P ) − L(fN , P ) + < ǫ.
2
Thus f is integrable over [a, b].
Let N be as above. Then for n ≥ N , we have
ǫ ǫ
f (x) − < fn (x) < f (x) + .
4(b − a) 4(b − a)

64
Hence by integrating, we get
Z b Z b Z b
ǫ ǫ
f− ≤ fn ≤ f+ ,
a 4 a a 4
R R b 
 b ǫ
i. e.  a f − a fn  ≤ 4 < ǫ. ✷

Exercise 10.9 Let fn be a sequence of functions that converge uniformly to a function f on the interval
[0, 1]. Assume that fn is integrable on [0, 1 − n1 ] for each n, and f is bounded on [0, 1]. Show that f is integrable
on [0, 1] and
Z 1 Z 1− 1
n
f = lim fn .
0 n→∞ 0

Differentiability continues to be ill-behaved, even with uniform convergence, as the following

exercises illustrate.
sin
√nx ,
Exercise 10.10 Let fn (x) = n
x ∈ R. Show that {fn } converges uniformly to the zero function, each
fn is differentiable, but the sequence {fn′ } does not converge even pointwise.
1
Exercise 10.11 Let fn (x) = |x|1+ n , x ∈ (−1, 1). Show that the sequence {fn } converges uniformly to the
function f defined by f (x) := |x|. Show that each fn is differentiable throughout (−1, 1), but f is not.

Proposition 10.12 Let fn be a sequence of functions that converge uniformly to a function f

on an interval (a, b). Assume that each is differentiable on (a, b) and the sequence fn′ converges
uniformly to some function g. Then f is differentiable on (a, b) and f ′ = g.

Proof : Let c ∈ (a, b). We want to show that f is differentiable at c and f ′ (c) = g(c). Define a
sequence of functions hn on (a, b) as follows:
(
fn (x)−fn (c)
x−c if x 6= c,
hn (x) =
′
fn (c) if x = c.

Since fn converges to f and fn′ converges to g, it follows that the sequence hn converges
pointwise to the function h given by
(
f (x)−f (c)
x−c if x 6= c,
h(x) =
g(c) if x = c.

Next we will show that the sequence {hn } in uniformly Cauchy, from which it will then follow
that hn converges to h uniformly. Also observe that each hn is continuous at c. Therefore it will
then follow that the limit function h is also continuous at c. But this means limx→c f (x)−f
x−c
(c)
=
g(c), which is precisely what we want.
Take an ǫ > 0. Look at the difference |hn (x) − hm (x)|. For x = c, we have

|hn (c) − hm (c)| = |fn′ (c) − fm

′
(c)| ≤ sup |fn′ (x) − fm
′
(x)|.
x

For x 6= c,
 f (x) − f (x) − f (c) + f (c) 
 n m n m 
|hn (x) − hm (x)| =  
x−c

65
Now applying mean value theorem (to the function fn −fm), we get some d between x and c such
that fn (x)−fm (x)−f
x−c
n (c)+fm (c)
= fn′ (d) − fm
′ (d). Therefore |h (x) − h (x)| ≤ sup |f ′ (x) − f ′ (x)|.
n m x n m
′ ′
Thus combining the two cases, we get supx |hn (x) − hm (x)| ≤ supx |fn (x) − fm (x)|. Since the
sequence {fn′ } converges uniformly to g, it is uniformly Cauchy. Hence there is an N such that
for m, n ≥ N , supx |fn′ (x) − fm ′ (x)| < ǫ. Hence for m, n ≥ N , we get sup |h (x) − h (x)| < ǫ.
x n m
✷
The interval (a, b) plays no special role in the above result. It can be replaced by any other
set, possibly unbounded. When restricted to a bounded set, however, one could weaken the
assumptions by making use of the mean value theorem, which means one gets a slightly stronger
result.

Corollary 10.13 Let fn be a sequence of functions defined on an interval (a, b). Assume that
each is differentiable on (a, b) and the sequence fn′ converges uniformly to some function g.
Assume also that there is a point c ∈ (a, b) such that the sequence {fn (c)} converges. Then
there is a function f on (a, b) such that fn converges to f uniformly on (a, b), f is differentiable
on (a, b) and f ′ = g.

Proof : If the sequence {fn } converges uniformly to some function f , then by the previous
result, we are through. So it is enough to prove that the sequence {fn } is uniformly Cauchy.
Let x ∈ (a, b). Then

|fn (x) − fm (x)| = |fn (x) − fm (x) − fn (c) + fm (c) + fn (c) − fm (c)|
≤ |fn (x) − fm (x) − fn (c) + fm (c)| + |fn (c) − fm (c)|

Applying mean value theorem to the function fn − fm , we get fn (x) − fm (x) − fn (c) + fm (c) =
fn′ (d) − fm
′ (d) for some d between x and c (and hence in (a, b)). Therefore |f (x) − f (x)| ≤
n m
′ ′
supy |fn (y) − fm (y)| + |fn (c) − fm (c)|. Thus

sup |fn (x) − fm (x)| ≤ sup |fn′ (y) − fm

′
(y)| + |fn (c) − fm (c)|.
x y

Since the sequence {fn (c)} converges, it is Cauchy. So given ǫ > 0, there is an N1 such that
|fn (c) − fm (c)| < 2ǫ for m, n ≥ N1 . Similarly, since the sequence {fn } is uniformly convergent,
and hence uniformly Cauchy, there is an N2 such that supy |fn′ (y) − fm ′ (y)| < ǫ for m, n ≥ N .
2 2
Let N = max{N1 , N2 }. Then for m, n ≥ N , we have supx |fn (x) − fm (x)| < ǫ. ✷

10.1 Double sequences

A double sequence of real numbers is a map from N × N to R.

Proposition 10.14 Assume that {amn }n converges to bm uniformly in m and limm→∞ bm = b.

Then limm,n→∞ amn = b.

66
Proposition 10.15 Assume that limm,n→∞ amn = b and limn→∞ amn = bm for each m. Then
limm→∞ bm = b.

Existence of limm,n→∞ amn does not imply the existence of either limm→∞ amn or limn→∞ amn ,
as the following example illustrates.

67
 Week 14-15

11 Multivariable calculus
Functions from Rm to Rn .
differences and similarities between R and Rm for m > 1:

68
 Week ???

12 Misc

n
Let x > 0. Then lim 1 + nx exists and is finite. Call this number ex . We will now show that

n
the limit lim 1 − nx also exists and is equal to e1x .
 n
 x n  x n x2
1+ 1− = 1− 2
n n n
n
X n x2r
= 1+ (−1)r 2r .
r n
r=1

Hence
 n   2r
x n  x n X n x
| 1+ 1− − 1| ≤
n n r=1
r n2r
n  
1 X n x2r
≤
n r nr
r=1
n
1 X x2r
≤
n r=1
r!
1
≤ exp(x2 ).
n


x n

n
Therefore lim 1 + n 1 − nx = 1. This, combined with the observation that lim (1+1x )n =

n
1 x n
ex , gives us lim 1 − n = e1x .

n
Thus the limit lim 1 + nx exists and is finite for all real values of x. Let us denote the
limit by ex for every x, so that e−x = e1x for all x.
Next we show that ex ey = ex+y whenever x ≥ 0 and y ≥ 0. Observe that
 
 x n  y n x + y xy n
1+ 1+ = 1+ + 2
n n n n
 n X n  
x+y n (xy)2r
= 1+ + .
n r n2r
r=1

Therefore
 
 x n  y n x+y n
0 ≤ 1+ 1+ − 1+
n n n

69
 n
1 X (xy)2r
≤
n r=1 r!
1
≤ exp(xy)
n
Hence  n 
x+y x n  y n
lim 1+ = lim 1+ lim 1 + ,
n→∞ n n→∞ n n→∞ n
i. e. ex+y = ex ey .

Exercise 12.1 Let x, y > 0 and x − y > 0. Show that ex e−y = ex−y (Use the equality ex−y ey = ex ) and
e−x ey = ey−x
Now prove that ex ey = ex+y for all x and y.

12.1 Proof of irrationality of π

70