Imperial College 1st Year Physics UG, 2017-18

——————–
Maths Analysis
Lecture notes
——————–
Toby Wiseman

1
Toby Wiseman; Huxley 507, email: t.wiseman@imperial.ac.uk

Office hours: 11-12 Tuesday and Thursday

Tutorials: There will be 6 tutorials for the course. Three will be this term
in weeks 8, 9, 10. There following three will be next term.

Example problems: During the course a number of example sheets will

be handed out. Solutions will be given out one you have had a chance to
attempt the problems.

Books
This course is not based directly on any one book. Recommended textbooks
for the course are;

• K.E.Hirst, Numbers, Sequences and Series (London, Edward Arnold,

1995).

• G.Smith, Introductory Mathematics: Algebra and Analysis (Springer,

1998).

• M.Liebeck, A Concise Introduction to Pure Mathematics (Chapman

and Hall, CRC, 2000).

• K.G. Binmore, Mathematical Analysis. A Straightforward Approach

(Cambridge University Press, 1982).

The following books are less technical and even non-technical, but may be
enjoyable bed-time reading:

• G.Hardy, A Mathematician’s Apology ( Cambridge University Press,

1967).

• R.Courant and H.Robbins, What is Mathematics? An Elementary Ap-

proach to Ideas and Methods (Oxford University Press, 1941). (See also
an updated 1996 version with I.Stewart as an additional author).

• I.Stewart, Concepts of Modern Mathematics (Dover, 1995).

2
Plan for the course
• Sets and maps (or Basics):

Sets, notation, methods of proof, Russell’s paradox, maps (functions)

Physics: clear thought, logic, computing.

• Numbers (or the Big and the Small!):

Real numbers as infinite decimals, completeness of the reals, cardinal-

ity and countability, Cantor’s proof that the reals are not countable

Physics: infinitessimals, infinity and beyond!

• Infinite sequences (or An introduction to analysis):

Convergence using ✏ N . Monotone and bounded sequences, sub-

sequences. Bolzano-Weierstrass. Cauchy sequences as convergent se-
quences.

Physics: numerical algorithms

• Infinite series:

Convergence of a series, comparison test, Cauchy test, other tests,

power series, Riemann reordering

Physics: perturbation theory, physics of lattices

• Functions (and an introduction to ✏ analysis):

Limits and continuity, di↵erentiable functions, Taylor’s theorem and

series, analytic functions.

Physics: Calculus! Approximation and understanding perturbation

theory.

3
Some notation
In case you haven’t come across this notation....

s.t. such that

i↵ if and only if
9 there exists
8 for all
=) implies

4
1 Sets and Maps
1.1 Definitions and Notation
Definition (Set)
A set is a collection of distinct elements (or members).

Comments:
• The ordering of elements is irrelevant

• The number of elements of the same type is irrelevant; only the types
of distinct elements are important
Use comma separated notation with curly braces; eg.

{a, g, h, u} , {5, 6} , {r, 4, } , {1, 2, 3, . . .}

where . . . means ‘carry on in the obvious way’.

Equality: Two sets are equal if they contain the same elements.

Example:

{A, B} = {B, A} = {A, B, A} =

6 {1, 2}

Note that a set may contain elements which are themselves sets,

{A, {B, C}} = {{C, B}, A} =

6 {B, A, C}

We may name sets,

S1 = {r, 4, } , S2 = {r, 4} , S3 = { , r, 4}

and then,

S1 = S3 6= S2

Notation
We write ‘x is an element of the set X’ as,

x2X

5
Conversely if x is not contained in the set X we write,

x2
/X

Example: using S1 and S2 above;

r 2 S1 , 2 S1 , 2
/ S2

For contrast consider a sequence;

Definition (Sequence)
A sequence is an ordered collection of elements, denoted with brackets, eg.
(a, g, h, u, g, h) or (2, 2, 4, 4, 6, 6, 8, 8, . . .) where repetition is allowed. Then
(a, g, g, h) 6= (g, g, h, a) and (a, g, g, h) 6= (a, g, h).

We shall later encounter sequences, but now let us return to sets.

Definition (Subset)
A set X is a subset of a set Y if every element in X is contained in Y .

Notation
For a subset X of Y we write;

X✓Y

If X is a subset of Y , but we know it is not equal to Y (so X 6= Y ) then we

write

X⇢Y

Then there exists y 2 Y s.t. y 2

/ X.

If W is not a subset of Y then we write;

W 6✓ Y

Comment:

• Any set is a subset of itself.

6
Example: using the S1,2,3 defined above;

S2 ⇢ S1 , S1 6✓ S2 , S1 ✓ S 1 , S2 ✓ S2

Definition (Empty set)

The empty set, the set which contains no elements, is denoted ; or {}.

Comment:

• The empty set is always a subset of any set; ; ✓ X for any set X.

Notation (Set builder)

We may use set builder notation to build sets from all elements that obey a
logical condition. We denote this as,

X = {x| condition on x}

where x denotes/labels elements which satisfies the specified condition.

Reads: X is the set of all elements x such that the condition on x holds.

Example: given the set of natural numbers N, we can take the positive even
numbers;

N = {0, 1, 2, 3, . . .}
X = {x|x 2 N and x even} = {0, 2, 4, 6, . . .}

Example:

A = {a, b, b, c, d, e, f, h, z, z}
B = {d, f, e, h, h, g, i}
C = {x|x 2 A and x 2 / B} = {a, b, c, z}

Example:

D = {d|d = 2n and n 2 N} = {20 , 21 , 22 , 23 , . . .}

which can also be written more compactly as,

D = {2n |n 2 N}

7
A more compact notation is,

X = {x 2 Y | condition on x}

reading X is the set of all elements of Y which satisfy the given condition.

Example: the above can be written,

X = {x 2 N|x even}
C = {x 2 A|x 2
/ B}

Definition (Finite set)

A finite set is a set with finitely many (distinct) elements.

Definition (Infinite set)

An infinite set is a set which is not finite.

Examples of finite sets;

A = {1, 2, 3}
B = {x 2 N|x < 100}

Some important examples of infinite sets;

Definition (Natural numbers)

N = {0, 1, 2, 3, 4, . . .}

Definition (Positive natural numbers)

N+ = {1, 2, 3, 4, . . .}

Definition (Prime number)

A prime number (or ‘prime’) is a natural number greater than one which
has exactly two divisors; 1 and itself. Denote the set P ⇢ N+ ;

P = {2, 3, 5, 7, 11, . . .}

8
Fundamental Theorem of arithmetic: Every natural number greater
than one has a unique prime factorisation.

Definition (Integers)

Z = {. . . , 4, 3, 2, 1, 0, 1, 2, 3, 4, . . .}

Definition (Rational numbers)

a
Q = { |a 2 Z , b 2 N+ }
b
Real numbers
R is the set of all numbers on the continuous real number line. We will post-
pone a careful definition of them until later.

Definition (Irrational numbers)

The set of real numbers which are not rational; R \ Q = {x 2 R|x 2
/ Q}.

Note: An irrational number cannot be expressed as a

b
for any a 2 Z, b 2 N+ .
p
Example: In the first problem set you will prove n is irrational for any
prime number n. Other famous irrational numbers are ⇡, e.

Clearly,

N+ ⇢ N ⇢ Z ⇢ Q ⇢ R

There are some important subsets of R we will discuss later;

Definition (Interval): An interval A is a subset of R, A ⇢ R, such that if

x, y 2 A, and x  z  y then z 2 A.

Intervals are defined by their ’end’ positions, and may include their end points
of not. We use the notations;

9
Definition:

(a, b) = {x 2 R|a < x < b} ”Open interval”

[a, b] = {x 2 R|a  x  b} ”Closed interval”
(a, b] = {x 2 R|a < x  b} ”Half open interval”
[a, b) = {x 2 R|a  x < b}

Also we define;

(a, 1) = {x 2 R|a < x}

[a, 1) = {x 2 R|a  x}
( 1, a) = {x 2 R|x < a}
( 1, a] = {x 2 R|x  a}

10
1.2 Basics of proof
A proof is a clear and convincing argument. There are various methods of
proof we will employ in this course.

Your should lay out a proof with the proposition (or claim), and then the
argument. You should never use the result you are trying to prove in the
proof itself!

The proof should end with QED (Quod Erat Demonstrandum - which was to
be proved) or a square ⇤.

Examples of methods of proof

Proof by contradiction: You assume that what is to be proved is false,

and derive a logical contradiction from this assumption.

You must state clearly in the proof that you are assuming the proposition is
false for the purpose of contradiction.

Proposition: Each natural number n > 1 is a prime or product of primes.

Proof. Assume for contradiction that the proposition is false. Then there
exists n 2 N+ , and n > 1 which is the smallest number such that n is not a
prime nor a product of primes.

/ P (or it wouldn’t be a counter example).

Now n 2

But this implies n = x · y for some x, y 2 N+ and x, y < n (so x 6= 1, y 6= 1).

x, y are prime or products of primes (as n was the smallest counterexample).

But then we have a contradiction, as n = x · y is a product of primes.

Hence our assumption is false. Thus the proposition is true.

Note that all the above statements after the assumption for contradiction are
false.

11
Proof by induction: Suppose we have a proposition P (n) about a positive
number n 2 N+ and we can show;

• P (1) is true.

• Either one of the following;

– if we choose any n 2 N+ and assume P (n) is true then P (n + 1)

is also true.
– if we choose any n 2 N+ and assume P (r) is true for all r  n,
then P (n + 1) is also true.

Then it follows that P (n) is true for all n 2 N+ .

Note: We were not required to prove P (r + 1). Only to prove it under the
assumption that either P (r) or P (n) for all n  r were true.

Alternatively we can replace showing P (1) with showing P (k) for some k 2 N,
and then induction shows P (n) is true for all n k.
n(n+1)
Proposition: For n 2 N+ then 1 + 2 + 3 + . . . + n = 2
.
Proof. (By induction)

• The proposition is true for n = 1. The l.h.s. = 1 and the r.h.s.

= 1⇥2
2
= 1.

• Assume the proposition holds for some n 2 N+ . Then,

n(n + 1)
(1 + 2 + 3 + . . . + n) + (n + 1) = + (n + 1)
2
n2 + 3n + 2 (n + 1)(n + 2)
= =
2 2
Hence the proposition holds also for (n + 1).

By induction the proposition is true for all n 2 N+ .

12
Let us reprove using induction the previous proposition we proved using con-
tradition.

Proposition: (Again!) Each natural number n > 1 is a product of primes.

Proof. (By induction):

• The proposition is true for n = 2 as 2 is prime.

• Suppose for induction that the proposition holds for all r  n. Then
consider the proposition for n + 1.

If n + 1 2 P then the proposition holds.

Otherwise n + 1 2/ P. Then n + 1 = x · y for some x, y 2 N and x 6= 1,

y 6= 1. Then x, y < n + 1. But then x, y are products of primes by
assumption. Thus we see n+1 is a product of primes so the proposition
holds for n + 1.

By induction the proposition is true for all n 2.

13
Some general comments:

• Try not to write =) everywhere. Try to write the proof out in a clear
and well written way - just as you would want to read it.

• Obviously do not use the result you are trying to prove.

• When proving an algebraic relation such as x3 1 = (x 1)(x2 + x + 1),

prove that the l.h.s. is equal to the r.h.s. Do not land up proving 0 = 0!

• To contradict a proposition a single counterexample is sufficient.

For example - (False) Claim: the di↵erence between two irrational
numbers is always irrational.
p p
This is false as 2 2 = 0.

• There are necessary conditions and sufficient conditions. A condition

may be;

– Necessary and sufficient - eg. a necessary and sufficient con-

dition for a number to be even is its last digit must be even.
– Necessary but not sufficient - eg. a necessary but not sufficient
condition for a number to be divisible by 4 is its last digit must
be even.
– Sufficient but not necessary - eg. a sufficient but not necessary
condition for a number to be divisible by 2 is its last digit is a zero.

If we have a proposition with a ’necessary and sufficient’ or ’if and only

if’ condition, then must be carefully to prove the two logical statements.

14
1.3 Properties of sets
Definition (Cardinality)
The Cardinality of a finite set is the number of elements it contains; denoted
|X| for the cardinality of a set X.

Definition (Power set)

The power set of a set X is the set of all subsets of X and is denoted 2X ;

2X = {A|A ✓ X}

Reads: 2X is the set of all elements A such that A is a subset of X.

Example:

A = {1, 2, 3}
2A = {;, {1}, {2}, {3}, {1, 2}, {1, 3}, {2, 3}, {1, 2, 3}}

Theorem 1.1 (Cardinality of the power set). The power set of a finite set
X has cardinality |2X | = 2|X| , hence the notation.

Proof. Choose some ordering of the elements in the finite set X = {x1 , x2 , x3 , . . . , xn }
where n = |X|. Then a subset is specified by the n-digit binary number

p1 p2 . . . pn , pi 2 {0, 1}

where pi = 0 means xi is not in the subset, and pi = 1 means xi is included

in the subset.
Thus the number of subsets is equal to the number of n-digit binary num-
bers, i.e., 2n .

15
Definition (Union, Intersection, Set Di↵erence)
Let X and Y be sets. Then;
• The union of X and Y , denoted X [ Y , is;
X [ Y = {z|z 2 X or z 2 Y }

• The intersection of X and Y , denoted X \ Y , is;

X \ Y = {z|z 2 X and z 2 Y }

• The set di↵erence of X from Y , denoted X \ Y , is;

X \ Y = {z|z 2 X and z 2
/ Y}

We say two sets X and Y are disjoint i↵ X \ Y = ;.

Some properties:
• Union and intersection are associative: for any sets A, B, C then;
A [ (B [ C) = (A [ B) [ C
so we simply write this as A [ B [ C. Likewise for intersection.
• For any set A then; A [ ; = A and A \ ; = ;
Examples: for the sets A = {1, 2, 3} as before, B = {2, 6} and C =
{0, 1, 2, 3, 4} then,
A [ B = {1, 2, 3, 6}
A \ B = {2}
A \ B = {1, 3}
B \ A = {6}
A\C =;
and,
A[;=A
A\;=;
A\;=A

16
Definition (Cartesian product)
The Cartesian product of two sets X and Y , denoted X ⇥ Y , is the set
of ordered pairs (or sequences of length two) of elements, the first coming
from X and second from Y .

X ⇥ Y = {(x, y)|x 2 X , y 2 Y }

Comments:

• X ⇥ Y 6= Y ⇥ X unless X = Y .

• For finite sets |X ⇥ Y | = |X| ⇥ |Y |.

Examples: for the sets A, B and C above,

A ⇥ B = {(1, 2), (1, 6), (2, 2), (2, 6), (3, 2), (3, 6)}
B ⇥ A = {(2, 1), (6, 1), (2, 2), (6, 2), (2, 3), (6, 3)}

Note that A 6= B.

This generalizes in the natural way for the product of finitely many multiple
sets,

X1 ⇥ X2 ⇥ . . . ⇥ Xn = {(x1 , x2 , . . . , xn )|x1 2 X1 , x2 2 X2 , . . . , xn 2 Xn }

where the elements (x1 , x2 , . . . xn ) are sequences. Likewise for an infinity of

sets, so each Xn is a set for all n 2 N+ we may define;

X1 ⇥ X2 ⇥ . . . = {(x1 , x2 , . . . )|xn 2 Xn 8n 2 N+ }

Taking the Cartesian product of a set X with itself n-times we use the no-
tation; X n = X ⇥ X ⇥ . . . ⇥ X (for n-factors on the r.h.s.)

Examples:

• Two dimensional coordinates (x, y) where x, y 2 R are denoted R2 =

R ⇥ R.

• n-coordinates are elements of Rn = R ⇥ R ⇥ . . . ⇥ R

17
1.4 Russell’s Paradox
A set is a collection of elements.

• Can these elements be anything we like?

• Can the collection be arbitrary?

An example of a ’wild’ set is the set of all sets:

S = {x|x is a set}

Problems in the definitions of sets arise when we take the members of sets
to be themselves sets in an unrestricted manner.

Russell’s Paradox: Define the set of all sets that do not have themselves
as an element:

R = {x|x is a set , x 2
/ x}

Now R is a set whose elements are sets. Is it an element of itself?

• Suppose the answer were yes, so R 2 R. This would imply R 2

/ R by
construction of R.

• If the answer were no, so R 2

/ R, then by construction this would imply
R 2 R.

Thus either answer, yes or not, leads to contradiction.

Russell’s paradox was removed by introducing set axioms that don’t allow
the sets like S and R - for example the Zermelo-Fraenkel axioms.

These include a ‘foundation’ axiom; All non-empty sets X have a member

Y such that X and Y are disjoint sets. This rules out sets containing them-
selves, and sets like {{{{. . .}}}}.

18
1.5 Maps (functions)
Definition (Map/Function)
A map from a set D to a set T is an assignment of each element of D to
some element of T .
• D is the domain
• T is the target or codomain
• If we label the map - say f - we use the notations;
f: D!T
or if we are more explicit about specifying the map,
f: D!T
x ! f (x)
where x 2 D and f (x) 2 T .
• f (x) is the image of x under the map f
• x is the pre-image of f (x) under the map f

Example: D = {a, 6, 7} to T = {0, 12 , ;, {3}}. Consider a map g : D ! T

defined as,
g(a) = 0
g(6) = ;
g(7) = ;

19
Example:

h: R!R
x ! h(x) = x + 1

Two examples which are not maps;

Definition (Identity map)

For any domain we may define the identity map, notated idD , which ‘does
nothing’;

idD : D ! D
x!x

Definition (Image)
The image (or range) of a map f : D ! T is the set of elements of T that
those of D are mapped to.

Image(f ) = {y 2 T | 9 x 2 D s.t. f (x) = y}

This is sometime written f (D) = Image(f ).

20
Example: in the above maps then Image(g) = {0, ;} and Image(h) = R.

Examples:
p
1. f : R ! R, with f (x) = x, is not a map as x = 1 doesn’t have an
image in R.
However,
h : [0, 1) ! R
p
x ! x
is a well defined map. Note that Image(h) = [0, 1)
2. f : R ! R, with f (x) = 1/x, is not a map as f (0) isn’t well defined.
However,
j : R \ {0} ! R
1
x !
x
is a well defined map. Note that Image(j) = R \ {0}.
3. f : R ! R, with f (x) defined so that f (x) = tan(x), is not a map.
The map is not well defined at x = ± ⇡2 , ± 3⇡
2
, . . ..
However,
⇡ ⇡
k:( ,+ ) ! R
2 2
x ! tan x
is a well defined map.

4. f : R ! R, with f (x) defined so that sin (f (x)) = x, is not a map.

Firstly there are many values z such that sin z = 0 so there is no
unique specification of a map. Secondly consider |x| > 1 which doesn’t
get mapped anywhere.
However,
⇡ ⇡
l : [ 1, +1] ! [ ,+ ]
2 2
x ! arcsin x
⇡
is a well defined map. Note that Image(l) = [ 2
, + ⇡2 ].

21
Definition (Composition)
Given maps f : X ! Y and g : Y ! Z then we define the composition of
f with g as,

g·f :X !Z
x ! g · f (x) = g(f (x))

Note: g · f (x) means act first with f , then act with g on the result.

Proposition 1.1. Let f : X ! Y, g : Y ! Z and h : Z ! W be maps.

Then,

h · (g · f ) (x) = (h · g) · f (x)

This means we unambiguously write the composition of these 3 maps as

h · g · f.

Proof. Consider LHS:

h · (g · f ) (x) = h (g · f (x)) = h (g (f (x)))

Consider RHS:

(h · g) · f (x) = (h · g) (f (x)) = h (g (f (x)))

Hence the LHS = RHS.

22
Definition (Surjective)
A map f : D ! T is surjective (or onto) if for every y 2 T there exists an
x 2 D s.t. f (x) = y.

No target element is ’missed out’

Note: for a surjection f : D ! T then Image(f ) = T .

Definition (Injective)
A map f : D ! T is injective (or one-to-one) if x 6= y implies f (x) 6= f (y)
for all x, y 2 D.

Alternative definition (Injective)

A map f : D ! T is injective (or one-to-one) if f (x) = f (y) implies x = y
for all x, y 2 D.

Every domain element is uniquely paired with one in the target.

23
Definition (Bijective)
A map f : D ! T is bijective if it is both injective and surjective.

Each domain element is uniquely paired with a target element and vice versa.

Definition (Inverse)
1
Given a bijective map f : D ! T , then the inverse map, denoted f , is
defined by,
1
f :T !D
y ! x(y)

such that y = f (x). Such a map exists because f is surjective, and is unique
because f is an injection. Then,

f 1 · f = idD
f · f 1 = idT

Proposition 1.2. Let f : A ! B be a map between finite sets A and B.

Then,

1. If f is a surjection then |A| |B|

2. If f is an injection then |A|  |B|

3. If f is a bijection then |A| = |B|

Proof. Let A = {a1 , a2 , . . . , an } so n = |A|. Let,

R = Image(f ) = {b 2 B| 9 ai 2 A s.t. f (ai ) = b}

Hence we may write, R = {f (a1 ), f (a2 ), . . . , f (an )} ✓ B although generally

some f (ai ) will be equal, and hence the list for R may have repeated elements.
Hence we have |R|  |A| = n and |R|  |B|. Now,

24
 1. If f is a surjection then R = B and so |B|  n. Therefore |B|  |A|.

2. If f is an injection then f (ai ) 6= f (aj ) for i 6= j, and so |R| = |A| = n.

Then, |A|  |B|.

3. If f is a bijection then both |B|  |A| and |A|  |B| must be true
which implies |A| = |B|.

Corollary 1.1. The Pigeonhole Principle

If you have N pigeons and M pigeonholes and N > M , some pigeon hole
must have more than one pigeon in!

[ If f : A ! B for finite sets A, B s.t. |A| > |B| then f is not an injection.]

25
2 Numbers and counting
2.1 Real numbers
Rational numbers can be expressed as infinite decimals e.g.
1
• 2
= 0.5 = 0.50̇
1
• 9
= 0.11111 · · · = 0.1̇
1
• 7
= 0.142857142857 . . . = 0.1̇4̇2̇8̇5̇7̇
where the dots denote recurring digits.

Infinite decimals provide a practical definition of the real numbers. Let us try
the following definition - in fact we shall see it needs a slightly modification
shortly;

Trial definition (Real numbers)

A real number, r, is an infinite decimal:
r = n.d1 d2 d3 . . . (1)
where n 2 Z is the integer part, and di 2 {0, 1, 2, . . . , 9} for i = 1, 2, 3, . . .,
which is equal to the infinite sum,
d1 d2 d3
r =n+ + 2 + 3 + ... (2)
10 10 10
We return to infinite sums later.

Example: ⇡ = 3.14159265 . . ., so that,

1 4 1 5
⇡ =3+ + + + + ... (3)
10 100 1000 10000
Comment: Our notation isn’t quite the one we are used to; note that,
2 4 4
( 1).2440̇ = 1+ + + (4)
10 100 1000
2 4 4
in our notation, not, 1.244 = 1+ 10
+ 100
+ 1000
as we usually write.

Questions

26
 1. Can the same real number have two di↵erent decimal expressions?

2. Given a decimal can we tell if the real number is rational?

Answer to question 1: Yes! An example is,

0.99999 . . . = 1.00000 . . . (5)

To prove this we require the following Lemma...

Lemma 2.1. Consider a real x 2 R such that |x| < 1, then,

1
1 + x + x2 + x3 + . . . = (6)
1 x
This is a geometric series, and we prove this lemma later. Now we may prove
the above statement.

Proposition 2.1. 0.99999 . . . = 1.00000 . . .

Proof.
9 9 9
0.99999 . . . = + 2 + 3 + ...
10 ✓ 10 10 ◆
9 1 1
= 1+ + + ...
10 10 102
✓ ◆
9 1
= 1 =1 (7)
10 1 10

Any infinite decimal that ends in recurring 9’s has the same property. There
exists a number ending in recurring 0’s that represents the same real number.
For example;

5.86299999 . . . = 5.86300000 . . . (8)

In fact this is the only type of ambiguity. We may amend our definition of a
real number to not allow decimals ending in recurring 9’s.

27
Definition (Real numbers)
A real number, r, is the infinite decimal,
r = n.d1 d2 d3 . . . (9)
where n 2 Z is the integer part, and di 2 {0, 1, 2, . . . , 9} for i = 1, 2, 3, . . .,
which is equal to the infinite sum,
d1 d2 d3
r =n+ + 2 + 3 + ... (10)
10 10 10
and we further require the condition,
6 9 m 2 N+ s.t. di = 9 8 i m (11)

Now;
• Every real number uniquely corresponds to an infinite decimal
• Two di↵erence decimals are di↵erent real numbers
Answer to question 2:

Proposition 2.2. A real is rational if and only if it ends in recurring digits.

Proof. We must prove both the if and the only if.

Firstly we prove that ending in recurring digits implies the number is rational.
Suppose a real ends in recurring digits;
r = n.d1 d2 . . . dn d˙n+1 . . . d˙n+N (12)
Then,
d1 d2 . . . dn dn+1 . . . dn+N dn+1 . . . dn+N dn+1 . . . dn+N
r =n+ n
+ n+N
+ n+2N
+ + ...
10 10 ✓ 10 10◆n+3N
d1 d2 . . . dn dn+1 . . . dn+N 1 1
=n+ n
+ n+N
1 + N + 2N + . . .
10 10 10 10
✓ ◆
d1 d2 . . . dn dn+1 . . . dn+N 1
=n+ +
10 n 10 n+N 1 101N
✓ ◆
d1 d2 . . . dn dn+1 . . . dn+N 1
=n+ + (13)
10n 10n 10N 1

28
where we use the notation a1 a2 . . . am is the natural number with decimal
digits ai , so,

a1 a2 . . . am = am + 10am 1 + . . . + 10m 2 a2 + 10m 1 a1 (14)

Then we see r is the sum of 3 rational numbers and hence is rational.

Secondly we prove that a rational number ends in recurring digits. Suppose

r is rational which is positive and less than one, so 0  r < 1. Then we may
write r = p/q for some p 2 N, q 2 N+ such that p < q. We use long division
to get the decimal expansion;

r = 0.d1 d2 d3 . . . (15)

where,
10p
d1 = b c , remainder r1 = 10p d1 q
q
10r1
d2 = b c , remainder r2 = 10r1 d2 q
q
10r2
d3 = b c , remainder r3 = 10r2 d3 q
q
..
.
10rn 1
dn = b c, remainder rn = 10rn 1 dn q
q
..
. (16)

and then the remainders ri are elements of {0, 1, 2, . . . , q 1}, so there are
q possible remainders. Note that at step n, this procedure only depends on
the remainder from the previous step (n 1).

Notation: dxe is the smallest integer which is greater or equal to x 2 R and

bxc is the largest integer which is less or equal to x. ie. b5.5436c = 5.

Now the pigeon hole principle during q + 1 steps of this process we must have
that at least two of the remainders are equal. Suppose that the first repeated
remainder is rn and repeats at rn+k . Then the digits dn+1 , dn+2 , . . . , dn+k will

29
repeat indefinitely, so,
p
= 0.d1 d2 . . . dn d˙n+1 d˙n+2 . . . d˙n+k (17)
q
This proves a rational 0  r < 1 is a recurring decimal. We may write a
general rational as,
p
r=N+ (18)
q

where N 2 Z and p, q as as above (p 2 N, q 2 N+ with p < q). The decimal

expansion of this after the decimal point is the same as for p/q and therefore
must be recurring.

30
2.2 Completeness of R
Definition (upper/lower bounds): Let A ⇢ R and A 6= ;.

• x 2 R is an upper bound of A if x a for all a 2 A.

• x 2 R is a lower bound of A if x  a for all a 2 A.

If A ⇢ R has an upper/lower bound we say it is bounded from above/below.

If both an upper and lower bound exist we say it is bounded.

Definition (max/min): Let A ⇢ R and A 6= ;.

• a 2 A is the maximum of A if a b for all b 2 A.

• a 2 A is the minimum of A if a  b for all b 2 A.

Notation: we denote these max(A) and min(A).

Definition (sup/inf ): Let A ⇢ R and A 6= ;.

• x 2 R is the supremum of A if x is an upper bound of A such that

for any other upper bound y of A then x  y.

• x 2 R is the infimum of A if x is a lower bound of A such that for any

other lower bound y of A then x y.

Notation: we denote these sup(A) and inf(A).

Note that we also use the terminology,

• Supremum = least upper bound or l.u.b.

• Infimum = greatest lower bound or g.l.b.

Example: Let X = { 3, 1, 5}. Then,

• 6, 15, 1000 are examples of upper bounds. -1010 is an example of a

lower bound.

• X is bounded.

• The maximum of X is 5, the minimum us 3.

31
 • The supremum of X is 5 and the infimum is 3.

Example: Let Y = {x 2 R|0  x < 1}. Then,

• 6, 15, 1000 are examples of upper bounds. -1010 is an example of a

lower bound.

• X is bounded.

• The minimum of X is 0, but X has no maximum.

• The supremum of X is 1 and the infimum is 0.

Example: Let Z = {x 2 R|x > 5}. Then,

• -1010 is an example of a lower bound.

• Z is bounded from below, but is not bounded from above.

• Z as no maximum or minimum.

• The infimum of Z is 5, there is no supremum.

Theorem 2.1. Completeness of R

• Every non-empty set A ⇢ R which is bounded below has an infimum

(g.l.b).

• Every non-empty set A ⇢ R which is bounded above has a supremum

(l.u.b).

Aside: The rationals, Q, are not complete. Consider X = {y 2 Q|y 2 < 2}.
Then X is bounded from above by rational numbers, eg. 32 is an upper bound,
but has no supremum (l.u.b.) in Q.

Completeness of R can be thought of as saying there are no gaps in the real

number line. We see that the rationals Q has gaps at the irrationals.

32
Proof. (of existence of infimum):
Since A is bounded from below there exists some integer n 2 Z such that it
is the greatest integer which is a lower bound of A (ie. n is a lower bound,
but n + 1 is not).

Consider the rationals,

n.0
n.1
n.2
..
.
n.9 (19)

Choose the rational from these, n.d1 , such that it is the greatest one that is
a lower bound.

Now consider,

n.d1 0
n.d1 1
n.d1 2
..
.
n.d1 9 (20)

Choose the rational, n.d1 d2 , such that it is the greatest of these and still is a
lower bound.

Continue in this way to construct the number, n.d1 d2 d3 . . . which is the infi-
mum (g.l.b) of A.

Note: we never construct recurring 9’s using this procedure.

The proof of existence of a supremum proceeds in a similar way.

33
Note: At every step in this algorithm we are making decisions about rational
numbers or integers to construct a new lower bound. It is only upon iterating
that we construct an infinite sequence of lower bounds whose limit is the
infimum, which may not be a rational number but will be an infinite decimal,
and hence a real number by our definition of real numbers - we will discuss
limits of sequences in much more detail later...

Theorem 2.2. For any two real numbers a, b 2 R we may find both a rational
and an irrational number that lie between them.

The proof of this is left as an exercise using infinite decimals in Tutorial

problem set 2. However it may be proved directly from completeness of R
(see Liebeck’s book) and hence should be thought of as a consequence of
completeness.

34
2.3 Countability and Cardinality
Definition (Cardinality equality)
Let A, B be sets (possibly infinite). If there exists a bijection f : A ! B
then we define,

|A| = |B|

Comments:

• This is compatible with our previous results for finite sets.

1
• Note that if a bijection f : A ! B exists, then f : B ! A also exists.

• If |A| = |B| and |B| = |C| then |A| = |C|. This follows from the fact
that if f : A ! B is a bijection, and g : B ! C is a bijection, then
g · f : A ! C is also a bijection.

Theorem 2.3. (Cantor-Bernstein-Schroeder)

Consider sets A, B. If there exists an injection from A ! B and an
injection from B ! A then there exists a bijection from A ! B.

This is not simple to prove and we will not prove it here.

Definition (Cardinality relation)

Let A, B be sets (possibly infinite). If there exists an injection map f :
A ! B then we define,

|A|  |B|

If there exists an injection from A to B, but no bijection, then we define,

|A| < |B|

Comments:

• Again this is compatible with our previous results for finite sets.

• In a similar way to equality above, we may show that if |A|  |B| and
|B|  |C| then |A|  |C|.

35
Crucially these definitions make sense due to the Cantor-Bernstein-Schroeder
theorem through the following proposition;
Proposition 2.3. If |A|  |B| and |B|  |A| then |A| = |B|.
Proof. If |A|  |B| then by definition there exists an injection A ! B. If
|B|  |A|, then an injection exists B ! A. By Cantor-Bernstein-Schroeder
then there is a bijection from A ! B and hence |A| = |B|.
Proposition 2.4. If A ✓ B then |A|  |B|.
Proof. For a subset A ! B then the identity map idA is an injection, showing
the result.
Remarks:
• For reasonable set axioms (which include the ‘axiom of choice’) then for
sets A, B the existence of an injection f : A ! B implies the existence
of a surjection g : B ! A.
• For sets A, B exactly one of the following is true;
|A| = |B|, |A| < |B| or |A| > |B|.
Proposition 2.5. |N+ | = |N|
Proof. Consider the map,
f: N+ ! N
x ! f (x) = x 1 (21)
This map is a bijection.
Proposition 2.6. |N+ | = |Z|
Proof. Consider the map,
f: N+ ! Z (22)
defined by;
f (2n + 1) = n n 2 {0, 1, 2, . . .}
f (2m) = m m 2 {1, 2, 3, . . .}
so,
{f (1), f (2), f (3), f (4), . . .} = {0, 1, 1, 2, 2, 3, 3 . . .} (23)
This map is a bijection.

36
Note: these two results also imply; |N| = |Z|.

Definition (Countability)
A set A is countable if |A| = |N+ | (so there exists a bijection from N+ ! A).

Comment: While one can certainly count the number of elements in a finite
set, it is not a countable set.

Proposition 2.7. A set X is countable if and only if it can be ”written out”

as an infinite list,

X = {x1 , x2 , x3 , . . .} (24)

where all the xi are distinct.

Proof. Firstly we prove that if X is countable then it can be written out as

a list. Secondly we prove that if it can be written as a list, it is countable.

Suppose X is countable. Then there exists a bijection f : N+ ! X. Then

we can list the elements as,

X = {f (1), f (2), f (3), f (4), . . .} (25)

Note that all the elements of X are listed exactly once.

Now suppose we may list the elements as above so,

X = {x1 , x2 , x3 , . . .} (26)

Then we can construct the map,

f : N+ ! X
n ! f (n) = xn (27)

This is surjective and since the xn are distinct also injective so f is a bijection.
Hence X is countable.

37
This result is very useful as it allows a ”proof by list”; showing a set can be
listed then proves countability. Note that it must be clear and unambiguous
how the list is defined.

Example: We have seen Z is countable above. A slick proof is;

Claim: Z is countable.
Proof. We may list Z as,

Z = {0, 1, 1, 2, 2, 3, 3 . . .} (28)

so it is countable.
Notation (@0 )
The standard symbol for |N+ | is called ”aleph-zero” or ”aleph-nought”;

@0 = |N+ |

This is a ‘transfinite Cardinal number’.

The Cardinal numbers give the sizes of sets. They are the natural numbers
together with ‘transfinite’ Cardinal numbers so;

0, 1, 2, 3, . . . ; @0 , @1 , @2 , . . . (29)

Proposition 2.8. Any infinite subset of N is countable.

Proof. Consider the infinite subset S ✓ N. Choose s1 the smallest element of
S. Now consider the second smallest, s2 which is the smallest in S \{s1 }. And
so one. Then we list these {s1 , s2 , . . .}. Hence the subset is countable.
Proposition 2.9. If A and B are countable sets, then A [ B is countable.
Proof. A and B are both countable so we may list them as {a1 , a2 , . . .} and
{b1 , b2 , . . .} respectively. The set A [ B can be listed by alternating these
lists as,

A [ B = {a1 , b1 , a2 , b2 , a3 , b3 , . . .} (30)

and omitting any element that appears in the list already (since for the list
it is important that the elements are distinct).

38
Alternative? Could we argue A [ B = {a1 , a2 , . . . , b1 , b2 , . . .}? Recall a
proof must be convincing! This is not convincing as listing the a’s never
finishes, so one never starts listing the b’s.

Proposition 2.10. If {A1 , A2 , A3 , . . .} is a countable set of countable sets,

then the union of all these, [1
i=1 Ai , is countable.

”A countable union of countable sets is countable”

Proof. Since for any k 2 N+ the set Ak is countable we may list it as,

Ak = {ak,1 , ak,2 , ak,3 , . . .} (31)

Then the elements of [1

i=1 Ai can be displayed as a 2-dimensional array;

a1,1 ! a1,2 a1,3 ! a1,4 a1,5 ! ...

# " # "
a2,1 a2,2 a2,3 a2,4 a2,5 ...
# " # "
a3,1 ! a3,2 ! a3,3 a3,4 a3,5 ... (32)
# "
a4,1 a4,2 a4,3 a4,4 a4,5 ...
# "
a5,1 ! a5,2 ! a5,3 ! a5,4 ! a5,5 ...

and the elements can be listed by following the above arrows, and omitting
at element if it is equal to a previous element that has been listed;

[1
i=1 Ai = {a1,1 , a1,2 , a2,2 , a2,1 , a3,1 , a3,2 , a3,3 , a2,3 , a1,3 , a1,4 . . .} (33)

Note: We could take a di↵erent pattern of arrows!

Proposition 2.11. If A and B are countable sets, then A ⇥ B is countable.

Proof. A and B are both countable so we may list them as {a1 , a2 , . . .} and
{b1 , b2 , . . .} respectively. The elements may be listed by following the arrows

39
below;

(a1 , b1 ) ! (a1 , b2 ) (a1 , b3 ) ! (a1 , b4 ) (a1 , b5 ) ...

. % . %
(a2 , b1 ) (a2 , b2 ) (a2 , b3 ) (a2 , b4 ) (a2 , b5 ) ...
# % . %
(a3 , b1 ) (a3 , b2 ) (a3 , b3 ) (a3 , b4 ) (a3 , b5 ) ...
(34)
. %
(a4 , b1 ) (a4 , b2 ) (a4 , b3 ) (a4 , b4 ) (a4 , b5 ) ...
# %
(a5 , b1 ) (a5 , b2 ) (a5 , b3 ) (a5 , b4 ) (a5 , b5 ) . . .

Proposition 2.12. If n 2 N+ and Ai , 1  i  n are countable sets, then

A1 ⇥ A2 ⇥ . . . ⇥ An is countable.

Proof. (By induction)

Step 1: The statements obviously holds for n = 1, and also holds for
n = 2 by our last proposition.
Step 2: Assume the statement holds for some n 2 N+ . By assumption,
A1 ⇥ A2 ⇥ . . . An is countable. Hence by the last proposition we know that,

(A1 ⇥ A2 ⇥ . . . An ) ⇥ An+1 (35)

is countable. Define the map;

f: (A1 ⇥ A2 ⇥ . . . An ) ⇥ An+1 ! A1 ⇥ A2 ⇥ . . . An ⇥ An+1

((a1 , a2 , . . . an ), an+1 ) ! (a1 , a2 , . . . an , an+1 ) (36)

Clearly f is bijective, and hence | (A1 ⇥ A2 ⇥ . . . An ) ⇥ An+1 | = |A1 ⇥ A2 ⇥

. . . An ⇥ An+1 |. Hence A1 ⇥ A2 ⇥ . . . ⇥ An+1 is countable.
Step 3: By induction the statement holds for all n.

Proposition 2.13. The set of rational numbers Q is countable.

Proof. Any element can be written as r = pq for some p 2 Z and q 2 N+ .

Since we have previously listed Z = {0, 1, 1, 2, 2, 3, 3, . . .} we can display

40
the rationals in a two dimensional array as;
0 1 1 2 2 ...
1 0! 1 1! 2 2 ! ...
# " # "
1 1
2 0 2 2
1 1 ...
# " # "
1 1 2 2 (37)
3 0! 3
! 3 3 3
...
# "
1 1 1 1
4 0 4 4 2 2
...
# "
1 1 2 2
5 0! 5
! 5
! 5
! 5
...
Following the arrows, and omitting any rational that we have already en-
countered, we may list the rationals. Here we obtain,
1 1 1 1 2 1 1 1 1 2 2 2
Q = {0, 1, , , , , 1, 2, , , , , , , , , 2, . . .} (38)
2 3 3 2 3 4 4 5 5 5 5 3

Question: Is there a set that is not countable?

Theorem 2.4. Cantor’s Diagonal Argument

The set of real numbers, R, is not countable.
Proof. (By contradiction) Assume for contradiction that R is countable.
Then we can list its elements R = {r1 , r2 , . . .}. Then we can write the
elements as infinite decimals and form a two dimensional array of the digits;
r1 = n1 · d11 d12 d13 d14 . . . (39)
r2 = n2 · d21 d22 d23 d24 . . . (40)
r3 = n3 · d31 d32 d33 d34 . . . (41)
.. ..
. . (42)
rk = nk · dk1 dk2 dk3 dk4 . . . dkk . . . (43)
Now consider the real number,
⇢
2 if dii = 1
r = 0 · c1 c2 c3 c4 . . . , ci = (44)
1 if dii 6= 1

41
Now r 6= ri for any i, since it di↵ers from ri at least in the i-th decimal place.
Thus r is not in the list of reals, leading to contradiction. Hence R is not
countable.
Definition: An uncountable set is an infinite set which is not countable.

Corollary 2.1. @0 < |R|.

Proof. Clearly there is an injection f : N+ ! R given simply by the identity

map idN+ . Thus |N+ |  |R|. However we have just shown that R is not
countable, which implies there is no bijection from N+ ! R. Hence we see
|N+ | < |R|.

Proposition 2.14. If X is an infinite set, then X has a countable subset.

Proof. Choose an element x1 2 X.

Then X \ {x1 } is an infinite set. Thus we can choose an element x2 2
(X \ {x1 }).

Again X \ {x1 , x2 } is an infinite set so we can choose an element x3 2

(X \ {x1 , x2 }).

Repeating this we obtain a subset Y ✓ X which is listed as Y = {x1 , x2 , . . .}

and hence is countable.

The following two propositions are equivalent (with X = A\Y so X [Y = A).

Proposition 2.15. a) If A is an infinite set, and Y is a countable subset

Y ✓ A such that A \ Y is infinite, then |A| = |A \ Y |.

Proposition 2.16. b) If X is an infinite set, and Y is a countable set such

that X \ Y = ; then |X [ Y | = |X|.

The proof of proposition b) is left as questions 5 and 6 in example sheet 3.

Example:
By the above proposition a) |R| = |R \ Q|.

42
Proposition 2.17. The interval {x 2 R| ⇡
2
< x < ⇡
2
} has the same
cardinality as R.
Proof. The map,
⇡ ⇡
f: {x 2 R| <x< }!R
2 2
x ! f (x) = tan x (45)
is a bijection.
Proposition 2.18. The open interval (a, b) has the same cardinality as R.
Proof. Consider the map,
⇡ ⇡
g: {x 2 R|a < x < b} ! {y 2 R| <y< }
✓ 2 2 ◆
⇡ (a x) + (b x)
x!y= (46)
2 a b
which is a bijection. Then by the previous the open interval has the cardi-
nality equal to |R|.
Proposition 2.19. |R ⇥ R| = |R|
One can find an explicit bijection (for example using infinite decimal ex-
pansions). However a simpler proof uses the Cantor-Bernstien-Schroeder
theorem.

Proof. Firstly consider the injection:

f: R!R⇥R
x ! (x, 0) (47)

Secondly consider the injection (NOT a bijection!),

g : R⇥R!R
(n.a1 a2 a3 . . . , m.b1 b2 b3 . . .) ! h(n, m).a1 b2 a2 b2 a3 b3 . . . (48)
where h is some bijection h : Z ⇥ Z ! Z.

Hence by Cantor-Bernstien-Schroeder there exists a bijection from R ! R⇥R

and hence, |R| = |R ⇥ R|.

43
Proposition 2.20. |Rn | = |R| for n 2 N+ .
Proof is left as an exercise - question 3 in example sheet 3.

We may now show there are sets with cardinality greater than |R|.
Theorem 2.5. If A is a non-empty set, then |2A | > |A|.
Proof. Firstly consider the injection,
f: A ! 2A
a ! {a} (49)
A
This implies |A|  |2 |.

Now suppose for contradiction that |A| = |2A |. Then there must exist a
bijection g : A ! 2A . Then for any a 2 A, g(a) is a subset of A. Now define
the set,
Z = {a 2 A|a 2
/ g(a)} (50)
Now since Z ⇢ A, then Z 2 2A . Since g is a bijection there must exist some
y 2 A such that g(y) = Z.

Our original assumption for contradiction leads to the existence of the set Z,
and we can ask the logical question: is y 2 Z or y 2/ Z. One must be true.

• Suppose y 2 Z; this cannot be true since then y 2 g(y), but Z only

contains elements a such that a 2
/ g(a).

• Suppose y 2/ Z; this cannot be true since then y 2

/ g(y), but Z contains
all elements of A, a, such that a 2
/ g(a).

Hence we reach a contradiction and so there exists no bijection g. Then

6 |2A | so that |A| < |2A |.
|A| =

As a result we see that,

|R| < 2R (51)
Taking power sets we may construct larger and larger cardinalities.

44
3 Sequences
Consider ⇡ = 3.141592635 . . .. Now consider the decimals truncated (not
rounded!) to 1, 2, 3, . . . decimal places:

3.1 , 3.14 , 3.141 , 3.1415 , 3.14159 , . . . (52)

This is a sequence of real numbers that approach ⇡ as we go along the

sequence. The numbers get arbitrarily close to ⇡ and we say the sequence
converges to ⇡.

Definition (real sequence):

A real sequence is a map,

x: N+ ! R (53)

The map may be listed as the infinite sequence (x(1), x(2), x(3), . . .).

Notation: We refer to a real sequence as (xn ), where we understand this as

(x1 , x2 , . . .) with xi = x(i).

Definition (bounded sequence):

A sequence (xn ) is bounded from above/below if the set {xn 2 R|n 2 N+ }
is bounded from above/below. A sequence is bounded if it is bounded from
above and below.

Definition (increasing sequence):

A sequence (xn ) is increasing if xn+1 xn for all n 2 N+ .

Definition (decreasing sequence):

A sequence (xn ) is decreasing if xn+1  xn for all n 2 N+ .

Note: A constant sequence, xn = c for some c 2 R for all n 2 N+ is both

increasing and decreasing.

Definition (monotone sequence):

A sequence (xn ) is monotone if it is either increasing or decreasing.

45
3.1 Convergent sequences
We now introduce the very important tool of ✏ N.

Definition (convergent sequence):

A sequence (xn ) converges to the limit x 2 R if for any ✏ > 0 there exists an
N 2 N+ such that |xn x| < ✏ for all n > N . We say (xn ) is a convergent
sequence.

When reading this think: ”... for any ✏ > 0 no matter how small there
exists an N 2 N+ . . .

Thus for any ✏ > 0 there exists an N such that all subsequent terms n > N
lie in the interval xn 2 (x ✏, x + ✏).

Notation: When (xn ) converges to the limit x we write;

xn ! x as n ! 1 (54)

or,

x = lim xn (55)
n!1

Definition (divergent sequence):

If a sequence is not convergent we say it is divergent.

Note that while a sequence may not converge (ie. diverges), it need not be
unbounded. An example is,

(+1, 1, +1, 1, +1, 1 . . .) (56)

Example: Let xn = 1
n
for n 2 N+ .

Proposition: xn ! 0 as n ! 1.

46
Rough work: Want:

|xn x| < ✏
1
<✏
n
1
<✏
n
1
n> (57)
✏
End of rough work

Proof. Let ✏ > 0. Choose N 2 N+ such that N > 1✏ . Then for n > N we
have n > 1✏ and so xn = n1 < ✏. Hence,

|xn 0| < ✏ for n > N (58)

Example: Let a 2 R and |a| < 1. Let xn = an .

Proposition: xn ! 0 as n ! 1.

Rough work: Want:

|xn x| < ✏
|an | < ✏
|a|n < ✏
n log |a| < log ✏
log ✏
n> (59)
log |a|

noting that log |a| < 0. (For ✏ > 1 any n 2 N+ will do!). End of rough work

log ✏
Proof. Let ✏ > 0. Choose N 2 N+ such that N > log |a|
. Then for n > N we
log ✏
have n > log |a|
.

47
This implies, n log |a| < log ✏ since log |a| < 0, and so,

|a|n < ✏ =) |an | < ✏ (60)

Proposition 3.1. The limit of a convergent sequence is unique.

Proof. (By contradiction)

Assume for contradiction that (xn ) converges to both x 2 R and y 2 R with
x < y.

Draw diagram!

Let ✏ = y 2 x . Since (xn ) converges to x, there must exist N1 2 N+ such that

for n > N1 we have |xn x| < y 2 x .

However since (xn ) converges to y, there must exist N2 2 N+ such that for
n > N2 we have |xn y| < y 2 x .

Then for n > max{N1 , N2 } we have both |xn x| < y 2 x and |xn y| < y x
2
which is a contradiction as these intervals are disjoint.

Useful inequalities: The triangle inequality;

|x + y|  |x| + |y| (61)

or alternatively (taking y ! y)

|x y|  |x| + |y| (62)

Proposition 3.2. Let xn ! x and yn ! y as n ! 1.

1. If zn = xn + yn then zn ! x + y as n ! 1.

2. If wn = xn · yn then wn ! x · y as n ! 1.
1 1
3. If y 6= 0 and yn 6= 0 then if un = yn
, un ! y
as n ! 1.

48
Proof. of proposition 1);

Let ✏ > 0. There exists N1 2 N+ such that for n > N1 then |xn x| < ✏
2
and there exists N2 2 N+ such that for n > N2 then |yn y| < 2✏ .

✏
Let N = max{N1 , N2 }. Then for n > N we have |xn x| < 2
and |yn y| < 2✏ .
Now,
✏ ✏
|zn (x + y)| = |(xn x) + (yn y)|  |xn x| + |yn y|  + =✏
2 2
(63)

Now let us prove proposition 2.

Rough work:

xn yn xy = (xn x)(yn y) + xyn + yxn 2xy

= (xn x)(yn y) + x(yn y) + y(xn x)
(64)

hence,

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

 |(xn x)(yn y)| + |x(yn y)| + |y(xn x)| (65)

but,

|(xn x)(yn y)| + |x(yn y)| + |y(xn x)| = |(xn x)| |(yn y)| + |x| |(yn y)| + |y| |(xn x)
(66)

Hence,

|xn yn xy|  |(xn x)| |(yn y)| + |x| |(yn y)| + |y| |(xn x)| (67)

End of rough work

49
Proof. of proposition 2);

Case 1: Let assume x 6= 0 and y 6= 0. Let ✏ > 0. There exists N1 2 N+ such

that for n > N1 then,
r
✏ ✏
|xn x| < and |xn x| < (68)
3 3|y|
There exists N2 2 N+ such that for n > N2 then,
r
✏ ✏
|yn y| < and |yn y| < (69)
3 3|x|
Let N = max{N1 , N2 }. Then for n > N we have,

|xn yn xy|  |(xn x)| |(yn y)| + |x| |(yn y)| + |y| |(xn x)|
✏ ✏ ✏
< + + =✏ (70)
3 3 3
Case 2: Suppose x = 0 but y 6= 0 or both. Let ✏ > 0. There exists N1 2 N+
such that for n > N1 then,

r
✏ ✏
|xn | < and |xn | < (71)
2 2|y|
There exists N2 2 N+ such that for n > N2 then,
r
✏
|yn y| < (72)
2
Let N = max{N1 , N2 }. Then for n > N we have,

|xn yn |  |xn | |(yn y)| + |y| |(xn x)|

✏ ✏
< + =✏ (73)
2 2
Case 3: Suppose x = 0 and y = 0. Let ✏ > 0. There exists N1 2 N+ such
that for n > N1 then,

p
|xn | < ✏ (74)

50
There exists N2 2 N+ such that for n > N2 then,
p
|yn y| < ✏ (75)

Let N = max{N1 , N2 }. Then for n > N we have,

|xn yn | = |xn | |yn | < ✏ (76)

Proof of proposition 3 is left as an exercise.

Theorem 3.1. (Squeezing theorem) Consider sequences (xn ), (yn ) and (zn )
where the first two converge to the same limit, xn ! Z and yn ! Z as n !
1. Suppose there exists N 2 N+ such that for all n > N then xn  yn  yn .
Then zn ! Z as n ! 1 too.

The proof is an exercise.

51
3.2 Bounds, monotonicity and convergence
We now consider the implications of completeness of the reals for sequences.

Lemma 3.1. An increasing sequence that is bounded from above is conver-

gent.

Proof. Consider (xn ) to be an increasing sequence which is bounded from

above. Then the set X = {xn 2 R|n 2 N+ } is bounded from above, and
hence by completeness, there exists a supremum (l.u.b.) for X. Call this
x. We now show that xn ! x.

Let ✏ > 0. Since x ✏ is less than x it is not an upper bound. This implies
there must exist some xN 2 X such that x ✏ < xN . Since (xn ) is increasing
then for n > N we have x ✏ < xN < xn . Thus,

✏ < xn x (77)

Now since x is an upper bound for X then xn  x < x + ✏. Hence,

xn x<✏ (78)

Together these imply,

|xn x| < ✏ (79)

Lemma 3.2. A decreasing sequence that is bounded from below is convergent.

The proof is similar to the above and is left as an exercise.

Corollary 3.1. A monotone bounded sequence is convergent.

Comment: Crucial in this corollary was the completeness of the reals.

We now define a subsequence - a sequence chosen from within a sequence,

keeping the same order.

52
Definition (Subsequence):
Let (xn ) be a sequence. Let n1 , n2 , n3 , . . . 2 N+ by a strictly increasing se-
quence of natural numbers, ie. n1 < n2 < n3 < . . .. Define x̄k = xnk for all
k 2 N+ . Then (x̄k ) is a subsequence of (xn )

Notation: We denote a subsequence of the sequence (xn ) as (xnk ), where

n1 < n2 < . . ..

Comments:

• Suppose (xn ) is a bounded sequence. Then any subsequence of it is

bounded.

• Suppose (xn ) converges to l. Then any subsequence also converges to

l.

Lemma 3.3. Any sequence has a monotone subsequence.

Proof. Let (xn ) be a sequence. For each N 2 N+ define the set TN = {xn |n >
N }. Since these sets TN ✓ R, then either all the sets TN have a maximum
or there exists some set TN with no maximum.

Case 1:
Suppose every set TN has a maximum, max(TN ) 2 R. Let xn1 be the first
occurrence of max(T1 ) in the sequence (xn ), and consider the set Tn1 . Now
Tn1 ✓ T1 so max(Tn1 )  max(T1 ). Let xn2 be the first occurrence of max(Tn1 )
in the sequence (xn ) such that n2 > n1 .

Now xn1 = max(T1 ) and xn2 = max(Tn1 ) so xn2  xn1 .

Now consider Tn2 ✓ Tn1 , so that max(Tn2 )  max(Tn1 ). Let xn3 be the first
occurrence in (xn ) of max(Tn2 ) such that n3 > n2 . Then xn3  xn2 .

Now consider Tn3 ✓ Tn2 , and continue in this way to produce a decreasing
sequence (xn1 , xn2 , xn3 , . . .).

Case 2:
Now consider the case TM has no maximum. Choose xn1 = xM +1 . Since TM
has no maximum, then there must exist elements of TM larger than xM +1 .

53
Then let xn2 be the first term in the sequence such that n2 > M + 1 and
x n 2 > xn 1 .

Consider Tn2 1 . Note the terms (xM +1 , xM +2 , . . . xn2 1 ) are all smaller than
xn2 . Hence Tn2 1 cannot have a maximum because otherwise that would
have been a maximum of TM which would be a contradiction. So xn2 is not
a maximum of Tn2 1 and hence it contains an element greater than xn2 . Let
xn3 be the first term in the sequence such that n3 > n2 and xn3 > xn2 .

Now consider Tn3 1 , and continue to construct an increasing subsequence

(xn1 , xn2 , xn3 , . . .).

Theorem 3.2. Bolzano-Weierstrass

Every bounded (real) sequence has a convergent subsequence.

Proof. Let (xn ) be a bounded sequence. By the previous Lemma it has a

monotone subsequence. This is bounded since the entire sequence is bounded,
and hence it is convergent (by previous Lemma).

54
Application: Root finding methods
An important application of these ideas is in understanding the behaviour
of algorithms to find roots of functions. Suppose we consider an interval
I = (a, b) and a function f : I ! R. Suppose we are interested in solving
the equation x = f (x), and we know this has a solution ⇠ 2 I. Now assume,

• The function is increasing so if x, y 2 I and x < y then f (x) < f (y).

• For x 2 (a, ⇠) then x < f (x) < f (⇠).

Then graphically we might expect the sequence (xn ) defined by xn+1 = f (xn )
converges to the root taking any starting value x1 2 (a, ⇠).

The convergence of the sequence is simple to show;

• For any xn 2 (a, ⇠] then xn+1 = f (xn ) > xn . Hence the sequence is
increasing.

• If xn 2 (a, ⇠) then xn+1 = f (xn ) < f (⇠) = ⇠. Thus xn < xn+1 < ⇠
so xn+1 2 (a, ⇠). By induction all xn 2 (a, ⇠) (since x1 is). Thus all
xn < ⇠.

The sequence is monotone increasing and bounded and therefore must con-
verge.

However we have not yet shown that this converges to ⇠ - just somewhere. In
order to finalize the proof we need to constraint the behaviour of the function
- we will return to this example when we have discussed continuous functions
later.

55
3.3 Cauchy sequences
Our definition of a convergent sequence is a bit clunky as it requires us to
know the value of the limit in order to prove it is convergent. Suppose we
want to know if a sequence converges, but don’t know what the limiting value
is?

We have seen that in the case of a monotone bounded sequence we were

able to prove convergence without knowing the limit. Suppose though we are
interested in a sequence that isn’t monotone.

Using Bolzano-Weierstrass we may find a very useful characterization for se-

quences that is equivalent to convergence, but that doesn’t refer explicitly to
a limit.

Definition (Cauchy Sequence):

A sequence (xn ) is a Cauchy sequence if for any ✏ > 0 there exists an
N 2 N+ such that |xn xm | < ✏ for all n, m > N .

Note: The important point is that a Cauchy sequence is a convergent se-

quence and vice versa as we now show. However we emphasize the definition
does not explicitly make reference to a limit. Thus if one doesn’t know the
limit of a sequence one may prove convergence by proving it is Cauchy.

Firstly we show a convergent sequence is Cauchy. This is simple to see. In

particular this is important as if one can prove a sequence is not Cauchy,
then it cannot be convergent. Clearly if a sequence is not convergent, and
therefore has no limit, it is more elegant to prove it is not Cauchy than to
prove is does not converge for any choice of limit.

Proposition 3.3. A convergent sequence is a Cauchy sequence.

Proof. Let ✏ > 0 and (xn ) be a convergent sequence, so xn ! x. Choose
N 2 N+ such that |xn x| < 2✏ for all n > N .

Then for n, m > N we have,

✏ ✏
|xn xm | = |(xn x) (xm x)|  |xn x| + |xm x| < + = ✏ (80)
2 2

56
Now we aim to show the converse - a Cauchy sequence is convergent. This
is important as without knowing a limit we may show a sequence is Cauchy.
To prove this we first show simply a Cauchy sequence is bounded. Then we
go on to show it converges using Bolzano-Weierstrass.

Lemma 3.4. A Cauchy sequence is bounded.

Proof. Let (xn ) be a Cauchy sequence. There exists N 2 N+ such that for
n, m > N then |xn xm | < 1.

Now,

|xn | = |xn xm + xm |  |xn xm | + |xm | (81)

Now choose m = N + 1. Then for n > N we have,

|xn | < 1 + |xN +1 | (82)

Therefore for all n 2 N+ we have,

|xn |  max {|x1 |, |x2 |, . . . , |xN |, |xN +1 | + 1} (83)

So we see that the set {xn |n 2 N+ } is bounded.

The converse of this result is also true for real sequences.

Theorem 3.3. A (real) Cauchy sequence is a convergent sequence.

Proof. Let (xn ) be a Cauchy sequence, and thus it is bounded. Let ✏ > 0.
By Bolzano-Weierstrass it has a convergent subsequence (xnk ). Let x 2 R be
the limit of this subsequence, so xnk ! x as k ! 1.

(xn ) is Cauchy: hence there exists N 2 N+ such that for n, m > N we have
|xn xm | < 2✏ .

xnk ! x as k ! 1 : hence we can find a K 2 N+ such that nK > N and

|xnK x| < 2✏ .

57
Now,
|xn x| = |(xn xnK ) (x xnK )| < |xn xnK | + |xnK x| (84)
and for n > N then,
|xn x| < ✏ (85)

As an example we will prove the following sequence is convergent without

having to determine its limit;

1
Proposition 3.4. The sequence x1 = a, x2 = b and xn+2 = 2
(xn+1 + xn )
for n 1 is convergent.
Whilst we won’t use this, the limit is xn ! 23 b 1
3
a.

Proof.
1 1
xm+2 xm+1 = (xm xm+1 ) =) |xm+2 xm+1 | = |xm+1 xm |
2 2
1
=) |xm+2 xm+1 | = m |b a| (86)
2
Then for n > m
|xn xm | = |xn xn 1 + xn 1 + . . . xm+1 + xm+1 xm |
 |xn xn 1 | + |xn 1 xn 2 | + . . . + |xm+2 xm+1 | + |xm+1 xm |
✓ ◆
1 1 1 2n1 m 1
 1 + + . . . + n m |xm+1 xm | = 1 ⇥ m 1 |b a|
2 2 1 2 2
✓ ◆
1 1 1
=4 m n
|b a|  m 2 |b a| (87)
2 2 2
Let ✏ > 0. Choose N such that,
1
|b a| < ✏ (88)
2N 2
Then for any n, m > N ,
1
|xn xm | 
|b a| < ✏ (89)
2N 2
Hence the sequence is Cauchy and thus is convergent.

58
The following is an important example:

Example: Let r = m.d1 d2 d3 d4 . . . be a real, and let (xn ) be the sequence of

rational numbers that are truncations of r to n decimal places:

x1 = m.d1
x2 = m.d1 d2
x3 = m.d1 d2 d3
.. ..
. .
xn = m.d1 d2 d3 . . . dn
.. ..
. . (90)

Proposition: (xn ) is a convergent sequence.

The proof is left as an exercise, but hopefully it is clear that (xn ) is a Cauchy
sequence, and hence is convergent.

Note: This shows that we may think about real numbers as a limit of a
Cauchy sequence of rational numbers, rather than as infinite decimals.

Proposition: xn ! r as n ! 1.

Again the proof is left as an exercise.

Note: This is an example of the power of a Cauchy sequence. Once one

knows a sequence is Cauchy, we know it converges, without having to refer
to its limit.

On the other hand, if we want to make statements about its limit then we
require the usual ✏-N technology.

59
Sequences in Q:

One may consider sequences over a subset of the reals such as the rationals
Q - for example the Cauchy sequence above.

One might imagine such sequences would behave in a similar way to real
sequences. However this is not the case. Consider the Cauchy sequence over
Q given by the truncation of ⇡ in decimals,

3 , 3.1 , 3.14 , 3.141 , . . . (91)

This sequence limits to ⇡, which is not rational. So due to lack of com-

pleteness of the rationals, we cannot say that a rational convergent sequence
converges within the rationals.

This fact can be used to give an alternative definition of the real numbers
using only rational numbers!

60
4 Series
The following are examples of series or series sums.
p 4 1 4
2=1+ + + + ...
10 100 1000
1
= 1 + x + x2 + x3 + . . . if |x| < 1
1 x
x2 x3
ex = 1 + x + + + ... (92)
2! 3!
Definition (Series):
A series P
(or series sum) is the infinite sum of a real sequence (an ). It is
denoted, 1 k=1 ak . We term the an the summands.

Note that we may define bn = an N +1 and then,

1
X 1
X
bk = ak (93)
k=N k=1

so we may write a sum ’starting’ at any number.

For example,
1
X
1 xn
= if |x| < 1 (94)
1 x n=0
n!

The value of a series is defined by our notion of convergence.

Definition (Partial sum): P

The n’th partial sum of a series 1k=1 ak is denoted Sn and is the sum of
the first n terms, ie.
n
X
Sn = ak = a1 + a2 + . . . + an (95)
k=1

p 4 1 4
Example: for the series 2=1+ 10
+ 100
+ 1000
+ then S3 = 1.41.

The partial sums form a sequence (Sn ).

61
4.1 Convergent series
Definition (value
P of a series):
If for a series 1k=1 k the sequence of partial sums, (Sn ), converges so S =
a
limn!1 Sn , then we say the series converges (or the series is convergent),
and its value is S, so,
1
X
S= ak (96)
k=1

If (Sn ) does not converge then the series is divergent.

Recall from the example sheets you proved (by induction);

n
X 1 xn+1
xk = (97)
k=0
1 x

Proposition 4.1. Geometric series

1
X 1
xk = if |x| < 1 (98)
k=0
1 x

Proof. Define the partial sums,

n
X 1 xn+1 1 1
Sn = xk = = xn+1 · (99)
k=0
1 x 1 x 1 x

by the previous exercise. Now if |x| < 1 then xn+1 ! 0 as n ! 1, and hence
xn+1
1 x
! 0 in this limit leaving,
1
Sn ! as n ! 1 (100)
1 x

P1 P1
Proposition 4.2. If x = k=1 xk and y = k=1 yk are convergent series,
then,
1
X
a·x+b·y = (a · xk + b · yk ) (101)
k=1

for any a, b 2 R.

62
 P P
Proof. Consider the partial sums Sn = nk=1 xk and Tn = nk=1 yk . Then
these converge, Sn ! x and Tn ! y. Now consider,
n
X
Un = (axk + byk ) (102)
k=1

so Un = aSn + bTn for any n 2 N+ . From our previous results on sequences,

Un ! ax + by as n ! 1.
Convergence of a series depends only on the tail of the sequence of partial
sums (not the first terms - or head). For example,

X1
29 1 222
2 1
5 + 19 2 +1+ 15 + ...... +3+
2 k=1
2k
102000000 terms (103)

is a convergent series.

P1 1
Proposition 4.3. The series k=1 k is divergent.

63
Proof. Firstly we see for n > 0;
2
X
n ✓ ◆ ✓ ◆
1 1 1 1 1 1 1 1
S 2n = =1+ + + + + + + + ...
k=1
k 2 3 4 5 6 7 8
✓ ◆
1 1 1
+ + + ... + p +
9p 10p 16
..
.
✓ ◆
1 1
+ + ... + n
2n 1 + 1 2
n 1
2 terms
✓ ◆ ✓ ◆
1 1 1 1 1 1 1
1+ + + + + + + + ...
2 4 4 8 8 8 8
✓ ◆
1 1 1
+ + + ... + p +
16p 16p 16
..
.
✓ ◆
1 1
+ + ... + n
2n 2
1 1 1 1 n
= 1 + + + + ... + = 1 +
2 2 2 2 2
n terms (104)

Thus the partial sums (S2n ) are unbounded from above as n ! 1. Thus the
sequence (Sn ) is also unbounded above and diverges as n ! 1.

P
Proposition 4.4. If a series 1 k=1 ak has positive terms, an 0 and (Sn )
is bounded above then the series converges.

Proof. If an 0 the (Sn ) is an increasing sequence. An increasing sequence

that is bounded above converges.

If we can show (Sn ) is increasing and bounded above then we learn the series
converges. However we don’t learn what value it converges to.

64
 P1 1
Proposition 4.5. The series k=1 kp is convergent for p > 1. (This is the
Riemann zeta function ⇣(p)).
Proof. Firstly we see for n > 0;
n 1
2X ✓ ◆ ✓ ◆
1 1 1 1 1 1 1
p
=1+ p
+ p + p
+ p + p + p + ...
k=1
k 2 3 4 5 6 7
✓ ◆
1 1 1
+ + + . . . + +
8p 9p 15p
..
.
✓ ◆
1 1
+ + ... + n
(2n 1 )p (2 1)p
2n 1 terms
✓ ◆ ✓ ◆
1 1 1 1 1 1
1+ + + + + + + ...
2p 2p 4p 4p 4p 4p
✓ ◆
1 1 1
+ + + ... + p +
8p 8p 8
..
.
✓ ◆
1 1
+ + ... + n 1 p
(2n 1 )p (2 )
2 4 8 2n 1
= 1 + p + p + p + . . . + p(n 1)
2 4 8 2
1 1 1 1
= 1 + p 1 + 2(p 1) + 3(p 1) + . . . + (n 1)(p 1)
2 2 2 2
1 n
1 2(p 1)
= (105)
1 2(p1 1)
1
Now since p > 1 then 2(p 1) < 1 so,
n 1
2X
1 1
 1 (106)
k=1
kp 1 2(p 1)

Now since 2n 1 n for any n 2 N+ then,

1
Sn  S2n 1  1 (107)
1 2(p 1)

65
Thus the sequence (Sn ) is bounded. Since an > 0 (Sn ) is increasing. Hence
it must converge.

Lemma
P 4.1. (Simple comparison test I) P1
Let 1 k=1 bk be a convergent series such that 0  bk . Then the series k=1 ak
converges if 0  ak  bk for all k 2 N .
+

Proof. Consider the partial sums,

n
X n
X
Sn = ak , Tn = bk (108)
k=1 k=1

Then both (Sn ) and (Tn ) are both increasing sequences and Sn  Tn for all
n 2 N+ .
P
Now since 1k=1 ak converges there exists T 2 R such that Tn ! T as n ! 1.
The sequence (Tn ) is bounded by T and hence so is (Sn ).

So (Sn ) is an increasing sequence that is bounded above, so it converges.

P1
Lemma 4.2. (Simple comparison test P II) Let k=1 bk be a divergent series
such that 0  bk . Then the series 1 a
k=1 k diverges if 0  bk  ak for all
k2N . +

The proof is left as an exercise.

66
4.2 Tests of convergence
We have seen above in the case where an 0 and (Sn ) is bounded then the
series converges. Suppose that the summands an are not positive - what can
we say more generally about convergence? In fact there are a number of
‘tests’ of convergence.

Since we don’t know whether a given series converges, and what the limit
of (Sn ) is, a powerful method to deal with series is observing convergence is
equivalent to (Sn ) being a Cauchy sequence. Recall Cauchy sequences are
useful as they are equivalent to convergent sequences, but do not explicitly
refer to the limit.

Proposition 4.6. (Cauchy convergence test)

A series is convergent if and only if for any ✏ > 0 there exists N such that,
n
X
ak < ✏ (109)
k=m

for all n m > N.

Proof. A series is convergent i↵ the partial sums (Sn ) are a convergent se-
quence. Hence it is convergent i↵ (Sn ) is a Cauchy sequence.

Let ✏ > 0. If (Sn ) is Cauchy there exists N such that for all n, m0 > N such
that |Sn Sm0 | < ✏. Take n > m0 , then,
n
X
Sn Sm 0 = ak (110)
k=m0 +1

So taking m = m0 + 1, then for all n m > N,

n
X
ak < ✏ (111)
k=m

Corollary 4.1. A necessary, but not sufficient, condition for convergence is;

|an | ! 0 as n!1 (112)

67
 P1
Note: If (ak ) does not converge to zero, the series k=1 ak does not converge.
Proof. That this is necessary for a convergent series follows directly from
proposition 4.6 above taking the case n = m.
P1 1
That it is not sufficient is shown by the example k=1 k , which as an ! 0
as n ! 1 but is divergent.

P1
Example: The series k=0 ( 1)k is divergent since an 6! 0 as n ! 1.

In the case an 0, we required boundedness of (Sn ) to show convergence.

However, in an ‘alternating’ series where the signs of an alternate, conver-
gence is automatic if the norms |an | decrease with n.

Lemma 4.3. (Alternating series test)

Let (bk ) be a decreasing sequence such that bk ! 0 as k ! 1. Then the
‘alternating’ series,
1
X
( 1)k 1 bk = b1 b2 + b3 b4 + . . . (113)
k=0

converges.
Note: bk 0
Proof. We observe 0  bn bn+1 for all n as (bn ) is decreasing.

Consider any n > m and the sum; I = bm+1 bm+2 + bm+3 . . . bn .

If the number of terms is even we may write;

I = bm+1 (bm+2 bm+3 ) ... (bn 2 bn 1 ) bn  bm+1 (114)

or write,

I = (bm+1 bm+2 ) + (bm+3 bm+4 ) + . . . + (bn 1 bn ) 0 (115)

On the other hand if the number of terms is odd;

I = bm+1 (bm+2 bm+3 ) ... (bn 1 bn )  bm+1 (116)

68
or write,
I = (bm+1 bm+2 ) + (bm+3 bm+4 ) + . . . + (bn 2 bn 1 ) + bn 0 (117)
Hence we see for any n > m;
0  bm+1 bm+2 + bm+3 . . . bn  bm+1 (118)
Let ✏ > 0. Since bn ! 0 as n ! 1 there exists N such that for all k > N
then |bk | = bk < ✏.

So for any n > m > N then we have,

|Sn Sm | = |(b1 b2 + b3 . . . bn ) (b1 b2 + b3 . . . bm )|
= |bm+1 bm+2 + bm+3 . . . bn |  bm+1 < ✏ (119)
Thus (Sn ) is Cauchy and converges.

Corollary 4.2. The sequence,

X1
( 1)k 1
(120)
k=0
k
converges.
Note: In fact this converges to ln 2.

Lemma
P1 4.4. (Comparison test) P
Let k=1 bk be a convergent series such that 0  bk . Then the series 1k=1 ak
converges if there exists N 2 N+ such that |ak |  bk for all k N .
Note: Before we only had this if ak > 0. Also this now only applies to the
tail of the series.

P1
Proof.
P1 Let ✏ > 0. Take S n and
P1 T n to be the partial sums for k=1 ak and
k=1 kb respectively. Since b
k=1 k converges, then (T n ) is Cauchy so there
exists M such that for any n > m > M ,
n
X n
X
bk = bk < ✏ (121)
k=m k=m

69
Let may choose M N . Now,

|Sn Sm | = |am+1 + am+2 + . . . + an |

 |am+1 | + |am+2 | + . . . + |an |
X n
 bm+1 + bm+2 + . . . + bn = bk < ✏ (122)
k=m
P1
Hence (Sn ) is Cauchy, so k=1 ak converges.

Definition
P1(absolute convergence):
A
P1series k=1 ak converges absolutely (or is absolutely convergent) if
k=1 |ak | converges.

P1 ( 1)k 1
Example: The sum k=1 k
converges but is not absolutely convergent.

Clearly this example illustrates a convergent series need not be absolutely

convergent. The converse is true however.

Lemma 4.5. An absolutely convergent series is convergent.

P
Proof. We assume 1 |ak | converges. Take bk = |a
k=1P Pk |, and apply the Com-
1 1
parison test, comparing k=1 ak to the convergent k=1 bk , bk 0.
P1
Since |ak | = bk this implies k=1 ak converges.

Lemma 4.6. (The root test)

P 1
Let 1 k=1 ak be a series and let xn = |an | for all n 2 N . Suppose that (xn )
n +

is a convergent sequence. Then if;

P
• xn ! x as n ! 1 and x > 1, then 1 k=1 ak is divergent.
P1
• xn ! x as n ! 1 and x < 1, then k=1 ak is convergent.

• xn ! 1 as n ! 1 then the test is inconclusive.

70
Proof. Let us consider the cases.

Case 1: Suppose xn ! x as n ! 1 and x > 1. Then choose ⇢ 2 R such that

1 < ⇢ < x and let ✏ = x ⇢.

Now (xn ) is a convergent sequence, so there exists N 2 N+ such that for

n > N then |xn x| < ✏.

Hence for all n > N then xn > ⇢, and so |an | > ⇢n > 1. Thus (an ) does not
tend to zero, and so the series cannot converge.

Case 2: Suppose xn ! x as n ! 1 and x < 1. Choose r 2 R such that

0 < x < r < 1 and let ✏ = r x.

Then there exists N 2 N+ such that for n > N then |xn x| < ✏ and hence
xn < r.

Thus for all n > N then |an | < rn .

P P1 n
Then by the comparison test, comparing 1 n
n=1 a to n=1 r (which con-
verges since 0 < r < 1) for the terms n > N we see convergence.

Case 3: Consider two series;

P1 1 1
1
• k=1 k diverges. Note k
k
! 1 as k ! 1.
1
P1 ( 1)k 1 ( 1)k 1 k 1
1
• k=1 k
converges. Note k
= k
k
! 1 as k ! 1.

Thus both convergence and divergence is possible for |ak |1/k ! 1.

Lemma 4.7. (The ratio test)

P
Let 1k=1 ak be a series and let yn =
an+1
, for all n 2 N+ . Then if,
an
P
• yn ! y as n ! 1 and y > 1 then 1k=1 ak diverges.
P
• yn ! y as n ! 1 and y < 1 then 1k=1 ak is convergent.

• yn ! 1 as n ! 1 then the test is inconclusive.

71
Proof. The proof is very similar to the root test.

Case 1: Suppose yn ! y as n ! 1 and y > 1. Then choose ⇢ 2 R such that

1 < ⇢ < y. Then there exists N 2 N+ such that for n > N then yn > ⇢.

Hence for all n > N then,

an+1
>⇢ (123)
an

so that, |an+1 | > |an |. Thus for n > N the sequence (|an |) is strictly increas-
ing, so |an+1 | > |aN +1 | for all n > N , and so the sequence (an ) cannot tend
to zero, and so the series cannot converge.

Case 2: Suppose yn ! y as n ! 1 and y < 1. Then choose r 2 R such that

y < r < 1. Then there exists N 2 N+ such that for n > N then yn < r.
aN +2
Then, yN +1 = aN +1
< r so that |aN +2 | < r|aN +1 |.

Similarly, |aN +3 | < r|aN +2 | < r2 |aN +1 |, and so on, so we see,

|ak | < rk (N +1)

|aN +1 | for k > N + 1 (124)
P1
Then by the strong comparison test, we see ak converges by comparison
P1 ⇣ k |aN +1 | ⌘ |aN +1 | Pk=1
1 k
to the convergent k=1 r rN +1 = rN +1 k=1 r for the terms k > N + 1.
(Note 0 < r < 1).

Case 3: Consider two series;

P1 1
• k=1 k diverges.
P1 ( 1)k 1
• k=1 k
converges.

Both have |an+1 /an | ! 1, so the test is inconclusive in this case.

72
4.3 Power Series
One of the most important examples of a series in physics is a power series.

Definition (Power series):

A power series is a series which is a function of a real variable x of the
form;
1
X
ck xk (125)
k=0

for ck 2 R for all k 2 N.

We state without proof the following theorem:

Theorem 4.1. Power series

For a power series there are 3 possibilities.
1. The series diverges for all x 2 R, x 6= 0.

2. The series converges for all x 2 R.

3. There exists r 2 R with r > 0 such that the series is absolutely conver-
gent for |x| < r, diverges for |x| > r and may or may not converge for
x = ±r.
In the latter case r is called the radius of convergence.
Examples:
P1
k=0 k!x diverges, by ratio test, for any x 2 R.
k
•
P1 1 k
• k=0 k! x converges, by ratio test, for any x 2 R.
P1 k
• k=0 x converges for |x| < 1, and diverges for |x| 1.
P1 ( 1)k 1 k
• k=1 k
x converges for |x| < 1, diverges for |x| > 1. It converges
for x = 1 and diverges for x = 1.
P ⇣ 1
⌘
Proposition 4.7. If for a power series 1 c
k=0 k x k
the sequence |c n | n con-
1
verges so |cn | n ! 1
r
for r 2 R, with r > 0, then r is the radius of convergence.

73
Proof. P
Consider applyingPthe root test to a power series. Consider the power
series 1 k
k=0 ck x = c0 +
1 n
n=1 an with an = cn x .

Then consider (yn ) with,

1 1 1
yn = |an | n = |cn xn | n = |cn | n |x| (126)
⇣ 1
⌘ 1
Suppose the sequence |cn | n converges to a finite real, so |cn | n ! 1r for
r 2 R, r > 0. Then,

|x|
yn ! y = as n ! 1 (127)
r
Then by the root test the series is convergent if y < 1, ie. |x| < r, and is
divergent if y > 1 ie. |x| > r. For y = 1, so |x| = r then the series may or
may not converge.

74
4.4 Riemann reordering
We have defined series as a sum of a sequence. Now a sequence has a definite
ordering of its elements. Does that matter for the series, which after all is
simply their sum?

Consider the example;

X1
( 1)k 1
1 1 1 1 1
=1 + + + ...
k=1
k 2 3 4 5 6
✓ ◆ ✓ ◆ ✓ ◆
1 1 1 1 1 1 1
= 1 + +
2 4 3 6 8 5 10
✓ ◆
1 1 1
+ ...
12 7 14
✓ ◆
1 1 1
+ ...
4r 2r + 1 2(2r + 1)
1 1 1 1 1 1 1
= + + + + ...
2 4 6 8 10 12 14
1 1
+ + ...
2(2r) 2(2r + 1)
✓ ◆
1 1 1 1 1 1
= 1 + + + ...
2 2 3 4 5 6
1
1 X ( 1)k 1
= (128)
2 k=1 k

We have previously shown this series converges, and hence is finite. Thus we
have reached a contradiction.

This example illustrates that it isn’t valid to reorder terms in a series.

Theorem 4.2. (RiemannP reordering)

Consider the series 1
k=1 ak .

• If it converges absolutely then any reordering of the series converges to

the same value.

75
 • If it converges, but does not converge absolutely, then for every real
P1⇢ 20 R there exists a reordering of the summands, (ak ), such
0
number
that k=1 ak = ⇢.

Proof. (Sketch of proof only)

If a series converges but not absolutely there must be an infinite sequence

of positive summands and an infinite sequence of negative summands. Both
these sequences must tend to zero.

Let (pn ) be the positive sequence and (qn ) be the negative one. Suppose
⇢ > 0 is the value we want the reordered series to converge to.

Take just enough of the first (pn )’s so their sum is just larger than ⇢.

(Note this is possible as the sum of the (pn ) must diverge, as the series does
not converge absolutely).

Now take just enough of the first (qn )’s so the sum is now just less than ⇢.

Now take enough of the next (pn )’s so the sum is again just larger than ⇢.

Again take enough of the next (qn )’s so the sum is just less than ⇢.

Continue in this manner. Since both sequences tend to zero this process
converges on ⇢.

76