Discrete Structures

Beifang Chen
 Contents

Chapter 1. Set Theory 5

1.1. Sets and Subsets 5
1.2. Set Operations 6
1.3. Functions 9
1.4. Injection, Surjection, and Bijection 11
1.5. Infinite Sets 13
1.6. Permutations 15

Chapter 2. Number Theory 19

2.1. Divisibility 19
2.2. Greatest Common Divisor 20
2.3. Least Common Multiple 23
2.4. Modulo Integers 24
2.5. RSA Cryptography System 26

Chapter 3. Propositional Logic 29

3.1. Statements 29
3.2. Connectives 29
3.3. Tautology 31
3.4. Methods of Proof 33
3.5. Mathematical Induction 35
3.6. Boolean Functions 35

Chapter 4. Combinatorics 39
4.1. Counting Principle 39
4.2. Permutations 39
4.3. Combination 42
4.4. Combination with Repetition 45
4.5. Combinatorial Proof 46
4.6. Pigeonhole Principle 50
4.7. Relation to Probability 51
4.8. Inclusion-Exclusion Principle 53
4.9. More Examples 57
4.10. Generalized Inclusion-Exclusion Formula 59

Chapter 5. Recurrence Relations 63

5.1. Infinite Sequences 63
5.2. Homogeneous Recurrence Relations 64
5.3. Higher Order Homogeneous Recurrence Relations 67
5.4. Non-homogeneous Equations 68
3
4 CONTENTS

5.5. Divide-and-Conquer Method 69

5.6. Searching and Sorting 75
5.7. Growth of Functions 75
Chapter 6. Binary Relations 77
6.1. Binary Relations 77
6.2. Representation of Relations 79
6.3. Composition of Relations 80
6.4. Special Relations 82
6.5. Equivalence Relations and Partitions 84
 CHAPTER 1

Set Theory

1.1. Sets and Subsets

A set is a collection of objects satisfying certain properties; the objects in the
collection are called elements (or objects or members). A set is considered to be
a whole entity and is different from its elements. Given a set A; we write x A
to say that x is an element of A or x belongs to A, and write x / A to say
that x is not an element of A or x doesnt belong to A. We usually denote sets by
uppercase letters such as A, B, C, . . . , X, Y, Z, and denote the elements of a set by
lowercase letters such as a, b, c, . . . , x, y, z, etc.
There are two ways to express a set. One way is to list all elements of the set;
the other way is to point out the attributes of the elements of the set. For example,
let A be the set of integers whose absolute values are less than or equal to 3. The
set A can be described in two ways:
A = {3, 2, 1, 0, 1, 2, 3} or A = {a | a is an integer, |a| 3}.
A set X whose elements satisfying Property P is denoted by
X = {x | x satisfies P } or X = {x : x satisfies P } .
In this note, most of time we use the first notation X = {x | x satisfies P }, and
occasionally use the second notation when there is confusion to use the symbol |.
There are two important things to be noticed about the concept of sets. The
first one is that any set, when it is considered as an object, can not be an element of
itself, but can be an element of another set. The second one is that for a particular
object, it is possible to decide in principle whether or not the object is an element
of a given set.
Let A and B be sets of real numbers satisfying the equations x2 1 = 0 and
4
x 1 = 0, respectively. In set notation,
A = {x | x R, x2 1 = 0} and B = {x | x R, x4 1 = 0}.
Apparently, the equation to define the elements of A and B are different. However,
the sets A and B consist of exactly the same elements, namely, 1 and 1. For this
reason we say that A and B are equal to each other, written A = B; it does not
matter whether or not the sets A and B were defined in different ways.
The sets we will constantly use in our course are the following sets:
Z: = the set of integers;
Q: = the set of rational numbers;
R: = the set of real numbers;
C: = the set of complex numbers;
P: = the set of positive integers;
N: = the set of nonnegative integers.
5
6 1. SET THEORY

Two sets A and B are called equal, written A = B, if every element of X is

an element of B and every element of B is also an element of A. As usual, we
write A 6= B to say that the sets A and B are not equal. In other words, there
is at least one element of A which is not an element of B, or, there is at least one
element of B which is not an element of A.
A set A is called a subset of a set B, written A B, if every element of A
is an element of B. Thus, for two sets A and B, A = B if and only if A B and
B A. A set is called finite if it has only finite number of elements; otherwise, it
is called infinite. For a finite set A, we denote by |A| the number of elements of
A; we call |A| the cardinality of A.
Consider the set A of real numbers satisfying the equation x2 + 1 = 0. We
will see that the set contains no elements at all; we call it empty. The set without
any element is called the empty set. There is one and only one empty set, and is
denoted by . The empty set is a subset of any set and || = 0.
Exercise 1. Let A = {1, 2, 3, 4, a, b, c, d}. Identify each of the following as true
or false.
(a) 2 A; (b) 3 6 A; (c) c A; (d) d 6 A; (e) 6 A; (f) e A;
(g) 8 6 A; (h) f 6 A; (i) A; (j) A A; (k) } A; (l) , A.
Exercise 2. List all subsets of a set A, where (a) A = {1}; (b) A = {1, 2}; (c)
A = {1, 2, 3}; (d) A = {1, 2, 3, 4}.
Exercise 3. Draw the Venn diagram that represents the following relation-
ships.
(1) A B, A C, B 6 C, and C 6 B.
(2) x A, x B, x 6 C, y B, y C, and y 6 A.
(3) A B, x 6 A, x B, A 6 C, y B, y C.

1.2. Set Operations

Let A and B be two sets. The intersection of A and B, written A B, is the
set of all elements common to the both sets A and B. In set notation,
AB = {x | x A and x B}.
The union of A and B, written AB, is the set consisting of the elements belonging
to either the set A or the set B, that is,
AB = {x | x A or x B}.
The symmetric difference AB of A and B is the set
AB = {x | x A or x B, but x 6 A B}.
The relative complement of A in B is the set consisting of the elements of B
that is not in A, that is,
B A = {x | x B, x 6 A}.
When we only consider subsets of a fixed set U , this fixed set U is sometimes
called a universal set. It should be noticed that a universal set is not universal;
it does not mean that it contains everything. For a universal set U and a subset
A U , the relative complement U A is just called the complement of A, written
A = U A.
 1.2. SET OPERATIONS 7

Since we always consider the elements in U , so, when x A, it is equivalent to

x 6 A. Similarly, x A is equivalent to x 6 A. To save writing space, we
sometimes use the symbol instead of writing is (are) equivalent to. For
instance, we may write x A is equivalent to x 6 A as x A x 6 A.
A convenient way to visualize sets in a universal set U is the Venn diagram.
We usually use a rectangle to represent the universal set U and circles or ovals to
represent its subsets as follows:
F igure
The intersection of a finite number of sets A1 , A2 , . . . , An is the set consisting
of elements common to all A1 , A2 , . . . , An , that is,
\n n o
Ai = A1 A2 An = x | x A1 , x A2 , . . . , x An .
i=1
Similarly, the union of A1 , A2 , . . . , An is the set, each of its element is contained
in some Ai , that is,
[n
Ai = A1 A2 An
i=1
n o
= x | there exists at least one Ai such that x Ai .
Let A1 , A2 , . . . be infinitely many sets. We define the intersection

\ n o
Ai = A1 A2 = x | x A1 , x A2 , . . .
i=1
and the union

[ n o
Ai = A1 A2 = x | there exists one i such that x Ai
i=1
Let Ai , i I, be a family of sets, indexed by a set I. We can also define the
intersection
\ n o
Ai = x | x Ai for all i I
iI
and the union
[ n o
Ai = x | x Ai for at least one i I .
iI

Theorem 1.1 (DeMorgans Law). Let A and B be subsets of a universal set

U . Then
(1) A = A,
(2) A B = A B,


(3) A B = A B.
Proof. (1) x A x 6 A x A.
(2) x A B x 6 A B x 6 A or x 6 B x A or x B

x A B.

(3) x A B x 6 A B x 6 A and x 6 B x A and x B
x A B.

The power set of a set A, written P(A), is the set of all subsets of A.
8 1. SET THEORY

Example 1.1. The power set of the set A = {a, b, c} is the set
n o
P(A) = , {a}, {b}, {c}, {a, b}, {a, c}, {b, c}, {a, b, c} .

The Cartesian product (or just product) of two sets A and B, written AB,
is the set consisting of all ordered pairs (a, b), where a A and b B, that is,
AB = {(a, b) | a A and b B}.
The product of a finite family of sets A1 , . . . , An is the set
Yn n o
Ai = A1 An = (a1 , . . . , an ) | a1 A1 , . . . , an An ,
i=1

the element (a1 , . . . , an ) is called an ordered n-tuple. The product of an infinite

family A1 , A2 , . . . of sets is the set
Y n o
Ai = A1 A2 = (a1 , a2 , . . .) | a1 A1 , a2 A2 , . . . .
i=1
Q
Each element of i=1 Ai can be considered as an infinite sequence. If A1 = A2 =
= A, we write
An = A A,
| {z }
n
A = A A .
| {z }

Example 1.2. For sets A = {0, 1}, B = {a, b, c}, the product A and B is the
set
A B = {(0, a), (0, b), (0, c), (1, a), (1, b), (1, c)};
and the product and A3 = A A A is the set
(0,1,0)???
A3 = {(0, 0, 0), (0, 0, 1), (0, 1, 1), (1, 0, 0), (1, 0, 1), (1, 1, 0), (1, 1, 1)}.
For the set R of real numbers, the product R2 is the 2-dimensional coordinate plane
and R3 is the 3-dimensional coordinate space.
A sequence of a nonempty set A is a list of finite or infinite number of objects
of A in order:
a1 , a2 , . . . , an (finite sequence)
a1 , a2 , . . . (infinite sequence)
where a1 , a2 , . . . A. The sequence is called finite in the former case and infinite
in the latter case. A word of length n over a nonempty set A is a string
a1 a2 an ,
where a1 , a2 , . . . , an A. The string
a1 a2
with a1 , a2 , . . . A is called a word of infinite length over A. There is a unique
word of length zero, called the empty word, and is denoted by . The sets of all
words of length n, of finite length, and of infinite length over A are denoted by
A(n) , A , and A() ,
 1.3. FUNCTIONS 9

respectively. Note that

[
A = A(n) .
n=0

Exercise 4. Let A be a set, and let Ai , i I, be a family of sets. Show that

[ \
Ai = Ai ;
iI iI
\ [
Ai = Ai ;
iI iI
[ [
A Ai = (A Ai );
iI iI
\ \
A Ai = (A Ai ).
iI iI

Exercise 5. Let A, B, C be finite sets. Use Venn diagram to show that

|A B C| = |A| + |B| + |C|
|A B| |A C| |B C| + |A B C|.

1.3. Functions
Definition 1.2. Let X and Y be nonempty sets. A function f from X to Y ,
written
f : X Y,
is a rule that associates each element x X with a unique element y Y , denoted
y = f (x).
When this rule f is given, the sets X, Y , and f (X) = {f (x) | x X} are called the
domain, the codomain, and the range of f , respectively; the element y(= f (x)) is
called the image (or value) of x, and x is called the inverse image (or preimage)
of y under f . Functions are also called mappings or just maps.
Note that for a function y = f (x) in calculus, the variable x is called the
independent variable and y is called the dependent variable.
Let f : X Y be a function from a set X to a set Y . It can be viewed as a
black-box device
x f f (x),
where the input x is in X and the output f (x) is to be in Y . For a subset A X,
the image of A under f is the set
f (A) := {y Y | there is a A such that y = f (a)};
and for a subset B Y , the inverse image of B under f is the set
f 1 (B) := {x X | there is b B such that b = f (x)}.
The graph of f is the set
(f ) = {(x, y) X Y | y = f (x)}.
The set of all functions from a set A to a set B is sometimes denoted by B A ; that
is,
B A := {f | f : A B}.
10 1. SET THEORY

Example 1.3. Some ordinary functions.

(1) The usual function y = x2 can be considered as a function f : R R,
defined by f (x) = x2 for x R. Its domain and codomain are the set R
of real numbers; its range is the set R0 of nonnegative real numbers.
(2) The usual function y = ex can be considered as a function f : R R+ ,
defined by f (x) = ex . Its domain is R; its codomain and range are the set
R+ of positive real numbers.
(3) y = log x is a function from R+ to R; its domain is R+ and codomain is
R.
(4) y = |x| is a function from R to the set R0 .
(5) y = sin x is a function from R to the closed interval [1, 1] of real numbers.
Example 1.4. Some functions to be appeared in future lectures.
(1) A finite sequence s1 , s2 , . . . , sn of a set A can be viewed as a function
s : {1, 2, . . . , n} A, defined by
s(k) = sk , k = 1, 2, . . . , n.
An infinite sequence s1 , s2 , . . . of A can be viewed as a function s : P A,
defined by s(k) = sk , k P.
(2) The factorial is a function f : N P defined by
f (n) = n! = 1 2 3 (n 1)n.
(3) The floor function is the function bc : R Z, defined by
bxc = the greatest integer less than or equal to x.
(4) The ceiling function is the function de : R Z, defined by
dxe = the smallest integer greater than or equal to x.
(5) For a universal set U , the characteristic function of a subset A U is
the function A : U {0, 1}, defined by

1 if x A
A (x) =
0 if x 6 A.
If U is finite and its elements are listed as {a1 , a2 , . . . , an } or simply iden-
tify U as the set {1, 2, . . . , n}. Then the subsets can be identified as se-
quences of 0 and 1 of length n. For instance, let U = {1, 2, 3, 4, 5, 6, 7, 8},
then the subset A = {2, 4, 5, 7, 8} corresponds to the sequence
0 1 0 1 1 0 1 1
(6) Let a be a positive integer. Then for any integer b there exist unique
integers q and r such that
b = qa + r, 0 r < a.
We have a function Quoa : Z Z, defined by
Quoa (b) = q, b Z;
and a function Rema : Z {0, 1, 2, . . . , a 1}, defined by
Rema (b) = r, b Z.
 1.4. INJECTION, SURJECTION, AND BIJECTION 11

Let f : X R and g : X R be functions. The addition of f and g is the

function f + g : X R defined by
(f + g)(x) = f (x) + g(x), x X;
the scalar multiplication of f by a constant c is the function cf : X R defined
by
(cf )(x) = cf (x), x X.
Exercise 6. Let A and B be the characteristic functions of subsets A and
B of a universal set X, respectively. Express the characteristic functions of A B,
A B, AB, and B A in terms of A and B , respectively. (Hint: Since
{0, 1} R, the functions A and B can be viewed as functions from A and B to
R, respectively. So the linear combination of A and B is meaningful.)

1.4. Injection, Surjection, and Bijection

Definition 1.3. Let X and Y be nonempty sets.
(1) A function f : X Y is called injective (or one-to-one) if distinct
elements of X are mapped to distinct elements in Y ; that is, for x1 , x2
X, if x1 6= x2 , then f (x1 ) 6= f (x2 ). An injective function is also called an
injection (or one-to-one mapping).
(2) A function f : X Y is called surjective (or onto) if every element
in Y is an image of some elements of X; that is, for any y Y , there
exist x X such that f (x) = y. A surjective function is also called a
surjection (or onto mapping).
(3) A function f : X Y is called bijective if it is both injective and
surjective. A bijective function is also called a bijection (or one-to-one
correspondence).
Example 1.5. (1) The function f : R R, f (x) = ex , is injective, but
not surjective.
(2) The function f : R R0 , f (x) = x2 , is surjective, but not injective.
(3) The function f : R R, f (x) = x3 , is bijective.
(4) The function f : R+ R, f (x) = log x, is bijective.
Definition 1.4. The composition of a function f : X Y and a function
g : Y Z is a function g f : X Z, defined by
(g f )(x) = g(f (x)), x X.
Whenever composition g f is concerned, we assume that the codomain of f
is the same as the domain of g.
Theorem 1.5. Let f : X Y , g : Y Z, h : Z W be functions. Then
h (g f ) = (h g) f,
as functions from X to W . We write h g f = h (g f ) = (h g) f .
Proof. For any x X, we have
(h (g f ))(x) = h((g f )(x))
= h((g(f (x)))
= (h g)(f (x))
= ((h g) f )(x).
12 1. SET THEORY

F ig.

The identity function of a set X is the function idX : X X such that
idX (x) = x for all x X.
Definition 1.6. A function f : X Y is called invertible if there exists a
function g : Y X such that for any x X and y Y ,
g(f (x)) = x and f (g(y)) = y.
In other words, g f = idX and f g = idY ; the function g is called an inverse of
f , and is denoted by f 1 .
Example 1.6. Some invertible and non-invertible functions.
(1) The function f : R R, f (x) = x3 , is invertible; its inverse is the function
g : R R, g(x) = 3 x.
(2) The function f : R R+ , f (x) = ex , is invertible; its inverse is the
function g : R+ R, g(x) = log x.
(3) The function f : R R, f (x) = x2 , is not invertible. However, the
2
function f1 : R0 R0 , f1 (x) = x , is invertible, and its inverse is the
function g1 : R0 R0 , g1 (x) = x. The function f2 : R0 R0 ,
f2 (x) = x2 , is invertible; its inverse is g2 : R0 R0 , g2 (x) = x.
(4) The function f : R [1, 1], f (x) = sin x, is not invertible. However,
f1 : [ 2 , 2 ] [1, 1], f1 (x) = sin x, is invertible, and has the inverse
g1 : [1, 1] [ 2 , 2 ], g1 (x) = arcsin x.
Theorem 1.7. A function f : X Y is invertible if and only if f is one-to-one
and onto.
Proof. : Since f is invertible, there is a function g : Y X such that
g f = idX and f g = idY . Suppose f is not one-to-one. Then there are x1 , x2 X
such that f (x1 ) = f (x2 ). Thus x1 = g(f (x1 )) = g(f (x2 )) = x2 , a contradiction. So
f is one-to-one. On the other hand, for any y Y , we have an element g(y) X
and f (g(y)) = y. This means that f is onto.
: Since f is one-to-one and onto, then for any y Y there is one and
only one element x X such that f (x) = y. We define a function g : Y X
by g(y) = x, where x is the unique element of X such that f (x) = y. Then
(g f )(x) = g(f (x)) = g(y) = x and (f g)(y) = f (g(y)) = f (x) = y. Thus
g f = idX and f g = idY .
Exercise 7. Show that if f : X Y is invertible, then the inverse function
of f is unique.
Exercise 8. Let f : X Y be a function. Let Ai , i I, be a family of
subsets of X. Show that
T T
(1) f SiI Ai iI f (Ai );
S
(2) f T iI Ai = f (Ai );
T
iI
(3) f SiI Ai = iI f 1 (Ai );
1
S
(4) f 1 iI Ai = iI f
1
(Ai ).
Exercise 9. Let f : X Y and g : Y Z be functions. Show that
(1) If f and g are one-to-one, then g f is one-to-one;
 1.5. INFINITE SETS 13

(2) If f and g are onto, then g f is onto;

(3) If g f is one-to-one, then f is one-to-one;
(4) If g f is onto, then g is onto.
Exercise 10. Let X be a finite set and let f : X X be a function. Then f
is bijective f is one-to-one f is onto.
Exercise 11. Let f : X X be an invertible function. Let f 0 = idX . For
any positive integer k, let
fk = f k1 f = f f f ,
| {z }
k

f k = f (k1) f 1 = f 1 f 1 f 1 .
| {z }
k

For each x X, the orbit of x under the map f is the set

Of (x) = {f k (x) | k Z}.
Show that for x1 , x2 X, if Of (x1 ) Of (x2 ) 6= , then Of (x1 ) = Of (x2 ).
Exercise 12. Verify that the function f : R {0, 1} R {0, 1}, defined by
1
f (x) = 1x , is a bijection; then list all elements of the set {f k | k Z}.

1.5. Infinite Sets

Definition 1.8. A set A is called equivalent to a set B, written A B, if
there is a one-to-one correspondence from A to B. The quantity to measure the
number of elements of a set A is the cardinality of A, denoted |A|. Two sets
have the same cardinality if they are equivalent, that is, they are in one-to-one
correspondent.
The cardinality of a finite set is just the number of elements of the set. The
empty set has the cardinality 0. For any set A,
|A(n) | = |An | and |A() | = |A |.
If A has exactly m elements, then there is a one-to-one correspondence from A to
the set [m] = {1, 2, . . . , m}.
Definition 1.9. A set A is called countable if it is finite or there is a one-
to-one correspondence from A to the set P of positive integers. Sets that are not
countable are called uncountable.
Proposition 1.10. Every infinite set contains an infinite countable subset.
Proof. Let A be an infinite set. Select an element from A, say a1 . Since A is
infinite, one can select an element from A other than a1 , say a2 . Similarly, one can
select an element a3 from A other than both a1 and a2 . Since the infinity of A, one
can continue this procedure by selecting a sequence of elements one after the other
to get an infinite countable subset {a1 , a2 , a3 , . . .}.

Theorem 1.11. If A and B are countable subsets, then A B is countable.

14 1. SET THEORY

Proof. It is obviously true when one of A and B is a finite set. Let A =

{a1 , a2 , . . .} and B = {b1 , b2 , . . .} be countable infinite sets. If A B = , then A
B = {a1 , b1 , a2 , b2 , . . .} is countable as demonstrated. If A B 6= , we just need to
delete the elements that appeared more than once in the sequence a1 , b1 , a2 , b2 , . . ..
Then the leftover is the set A B.
Theorem
S 1.12. Let Ai (i = 1, 2, ) be countable sets and Ai Aj = (i 6= j).
Then i=1 Ai is countable.
Proof.
SWe assume that Ai = {ai1 , ai2 , ai3 , } (i = 1, 2, . . .). Then the count-
ability of i=1 Ai can be demonstrated as
a11 a12 a13 a14
. % .
a21 a22 a23 a24
% . %
a31 a32 a23 a34
. % .
a41 a42 a23 a44
.. .. .. ..
. . . .

The condition of disjointness in Theorem 1.12 can be omitted.
Theorem 1.13. The interval [0, 1] of real numbers is uncountable.
Proof. Suppose the set [0, 1] is countable; that is, the numbers in [0, 1] can be
listed as an infinite sequence {i }
i=1 . Write all real numbers i in infinite decimal
forms, say in base 10, as follows:
1 = 0.a1 a2 a3 a4 a5 ,
2 = 0.b1 b2 b3 b4 b5 ,
3 = 0.c1 c2 c3 c4 c5 ,

Then we can construct a number x = 0.x1 x2 x3 , defined by

1 if a1 = 2
x1 =
2 if a1 6= 2,

1 if b2 = 2
x2 =
2 if b2 6= 2,

1 if c3 = 2
x3 =
2 if c3 6= 2,

The number x is an infinite decimal of 1s and 2s and is a real number between 0
and 1. Since x1 6= a1 , x2 6= b2 , x3 6= c3 , and so on, it follows that x 6= 1 , x 6= 2 ,
x 6= 3 , etc. Thus x is not in the list 1 , 2 , . . .; that is, x is not a real number
between 0 and 1, a contradiction.
Theorem 1.14 (Bernstein). For sets A and B, if there are subsets A1 A
and B1 B such that A1 B and A B1 , then A B.
 1.6. PERMUTATIONS 15

Exercise 13. For real numbers a, b R such that a < b, find a one-to-one
correspondence between the open interval (a, b) and the closed interval [a, b], where
(a, b) = {x R | a < x < b} and [a, b] = {x R | a x b}.
Exercise 14. Show that the union of countably many countable sets is still
countable.
S That is, if Ai are countable sets, i = 1, 2, . . ., not necessarily disjoint,
then i=1 Ai is countable.
Exercise 15. Let A be a countable set. Show that A is countable.
Exercise 16. Let A be a set with at least two elements. Show that A is in
one-to-one correspondent with a subset of A() .
Exercise 17. Let B = {0, 1} and let B = {a1 a2 | a1 , a2 , . . . B} be the
set of words with infinite length. Show that B is uncountable.

1.6. Permutations
A bijective function from a finite set A to itself is called a permutation of A.
If A = {a1 , a2 , . . . , an } is a set of n objects and f : A A is a permutation, then
f (a1 ), f (a2 ), . . . , f (an )
is the same collection of a1 , a2 , . . . , an , and may be in different order. Let
f (a1 ) = ai1 , f (a2 ) = ai2 , . . . , f (an ) = ain .
The permutation f is usually expressed by the array

a1 a2 . . . an
f=
ai1 ai2 . . . ain
and sometimes by the word
ai1 ai2 ain .
Let a be an element of A. The sequence
a, f (a), f 2 (a), f 3 (a), . . .
must eventually return to a and then repeat the pattern. Let k be the smallest
positive integer such that f k (a) = a. The sequence a, f (a), f 2 (a), . . . , f k1 (a) is
called a cycle of length k of the permutation f , and is denoted by

a f (a) f 2 (a) f k1 (a) .
Of course,

f (a) f 2 (a) f k (a) , f 2 (a) f 3 (a) f k+1 (a) , . . .
are also cycles of f . However, they are viewed as the same cycle. For instance, for
the set A = {1, 2, 3, 4, 5, 6, 7, 8}, the permutation

1 2 3 4 5 6 7 8
=
6 7 5 4 3 8 2 1
has four cycles (168), (27), (53), (4). The permutation can be written as (681)(27)(53)(4).
However, the way of writing in this fashion is not unique. We may make this
kind of writing unique by requiring that the leading element of each cycle to be the
largest element inside the cycle and requiring that all leading elements of cycles to
be increasing. Hence the permutation can be uniquely written in the cycle
= (4)(53)(72)(816).
16 1. SET THEORY

Theorem 1.15. Every permutation of {1, 2, . . . , n} can be written as disjoint

cycles in a unique way that the leading element of each cycle is the largest element
in the cycle and all leading elements of cycles are in increasing order.
Let n be fixed. For a permutation of {1, 2, . . . , n}, let can be written as
disjoint cycles. Note that each cycle can be viewed as a permutation of {1, 2, . . . , n}.
For instance, when n = 8 and = (4)(53)(72)(816), the cycle (53) can be viewed
as the permutation (1)(2)(4)(53)(6)(7)(8), the cycle (816) can be viewed as the
permutation (2)(3)(4)(5)(816)(7), and the cycle (4) can be viewed as the identity
permutation. We then have
(4)(53)(72)(816) = (53) (72) (816).
A permutation of {1, 2, . . . , n} is called a transposition if it has one cycle
of length 2 and all other cycles are of length 1. For instance, if = (ai aj ) is a
transposition, then ai 6= aj , (ai ) = aj , (aj ) = ai , and (a) = a for all a 6= ai ,
a 6= aj . Note that each cycle can be written as composition of transpositions. If
(a1 a2 . . . ak ) is a cycle, then
(a1 a2 ak ) = (a1 ak ) (a1 ak1 ) (a1 a3 ) (a1 a2 )
For instance, the cycle (8164) can be written as
(8164) = (84) (86) (81).
Corollary 1.16. Every permutation of {1, 2, . . . , n} can be written as compo-
sition of finite number of transpositions.
A permutation is called even if it can be written as composition of even number
of transpositions, and it is called odd if it can be written as composition of odd
number of transpositions.
Proposition 1.17. (1) The product of two even permutations is even.
(2) The product of two odd permutations is even.
(3) The product of an even permutation and an odd permutation is odd.
(4) There are n! n!
2 even permutations and 2 odd permutations of {1, 2, . . . , n}.

Let = a1 a2 an be a permutation of {1, 2, . . . , n}. A pair (ai , aj ) with i < j

is called an inversion of if ai > aj . The number of inversions of is denoted by
inv().
Exercise 18. Consider the permutation 387169425 of the set {1, 2, . . . , 9}, that
is,
1 2 3 4 5 6 7 8 9
.
3 8 7 1 6 9 4 2 5
(1) Find the permutation 1 .
(2) Write the permutation in disjoint cycles.
(3) Write the permutation as the composition of cycles.
(4) Write it as the composition of transpositions.
(5) Determine its parity.
(6) Find the total number of inversions of the permutation.
Exercise 19. Consider the permutation

1 2 ... n 1 n
= .
n n 1 ... 2 1
 1.6. PERMUTATIONS 17

(1) Write as a product of minimal number of transpositions;

(2) Determine the parity of ;
(3) Find inv().
Exercise 20. Let a1 a2 an be a permutation of {1, 2, . . . , n}.
(1) Find the parity of the permutation an an1 a2 a1 in terms of the parity
of a1 a2 an ;
(2) Find inv(an an1 a2 a1 ) in terms of inv(a1 a2 an ).
 CHAPTER 2

Number Theory

2.1. Divisibility
Leopold Kronecker said: God created integers, all else are the work of man.
We assume that the set of integers are well defined and we are familiar with the
properties of integers such as addition, subtraction, multiplication, and division. In
particular, we assume the following axiom for subsets of integers bounded below.
Axiom. For every nonempty subset of integers, if it is bounded below, then it
has a unique minimum integer.
It follows easily from the axiom that for every subset of integers, if it is bounded
above, then it has a unique maximum integer. Given two integers a and b with
a 6= 0, we say that a divides b, written a|b, if there exists an integer q such that
b = qa.
When this is true we say that a is a factor (or divisor) of b, and say that b is
a multiple of a. Obviously, any integer n has divisors, 1 and n, called the
trivial divisors of n. The divisors of n other than the trivial divisors are called
nontrivial divisors. Note that every integer is a divisor of 0. A positive integer
p (6= 1) is called a prime if its positive divisors are only the trivial divisors 1 and
p. A positive integer is called composite if it is not a prime. The first few primes
are listed as follows:
2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 57.
Proposition 2.1. Let a, b, c be nonzero integers.
(a) If a|b and b|a, then a = b.
(b) If a|b and b|c, then a|c.
(c) If a|b and a|c, then a|(bx + cy) for x, y Z.
Proof. (a) Let b = q1 a and a = q2 b for some integers q1 and q2 . Then
b = q1 q2 b.
Dividing both sides by b, we have q1 q2 = 1. It follows that q1 = q2 = 1. Thus
b = a.
(b) Let b = q1 a and c = q2 b for some integers q1 and q2 . Then c = q1 q2 a, that
is, a|c.
(c) Let b = q1 a and c = q2 a for q1 , q2 Z. Then for x, y Z,
bx + cy = (q1 x + q2 y)a,
that is, a|(bx + cy).
Theorem 2.2. There are infinitely many prime numbers.
19
20 2. NUMBER THEORY

Proof. Suppose there are finitely many primes, say, p1 , p2 , . . . , pk . Then the
integer
a = p1 p2 p k + 1
is not divisible by any of the primes p1 , p2 , . . . , pk because the remainders of a
dividing by p1 , p2 , . . . , pk respectively are always 1. This means that a has no
prime factors. By definition of primes, the integer a must be a prime, and this
prime is larger than all primes p1 , p2 , . . . , pk , a contradiction.
Theorem 2.3 (Division Algorithm). For any integers a and b, where a > 0,
there are unique integers q and r such that
b = qa + r, 0 r < a.
Proof. Consider the set S = {b ta 0 | t Z}. Obviously, S is nonempty
and is bounded below. Then S has the unique minimum element r, that is, there
is an unique integer q such that b qa = r. We claim that r < a. Suppose r a,
then b (q + 1)a = r a 0 shows that r a is an element of S. This is contrary
to that r is the minimum element of S.

2.2. Greatest Common Divisor

For integers a and b, not simultaneously 0, a common divisor of a and b is
an integer c such that c|a and c|b. Clearly, there are finitely many common divisors
for a and b; the very greatest one is called the greatest common divisor of a
and b, and is denoted by gcd(a, b). For convenience, we assume gcd(0, 0) = 0. Two
integers a and b are called coprime (or relatively prime) if gcd(a, b) = 1.
Theorem 2.4. Let d be the greatest common divisor of integers a and b, that
is, d = gcd(a, b). Then there exist integers x and y such that
d = ax + by.
Proof. It is obviously true when a = b = 0. Assume that a and b are not
simultaneously zero. We consider the set S = {au + bv | u, v Z} and the set
S+ = S Z+ . Note that S+ is bounded below and S+ 6= because a2 + b2 > 0.
Let s be the smallest integer in S+ and write
s = au0 + bv0
for some u0 , v0 Z. We claim that d = s.
Clearly, d divides every integer in S because d|a and d|b. In particular, d|s. We
then have d s. To show that s d, we claim that d divides every integer in S.
In fact, for any au + bv S with u, v Z, let
au + bv = qs + r, 0 r < s.
Then r = a(u qv0 ) + b(v qu0 ) S; so d|r. If r was positive, then r S+ and s
could not be the smallest integer in S+ . Thus r = 0; so s|(au + bv). In particular,
taking (u, v) = (0, 1) and (u, v) = (1, 0), we see that s is a common divisor of a and
b. Hence by definition of gcd, s d.
Theorem 2.5. For integers a, b, q, and r, if
b = qa + r,
then
gcd(a, b) = gcd(a, r).
 2.2. GREATEST COMMON DIVISOR 21

Proof. Let d1 = gcd(a, b) and let d2 = gcd(a, r). Obviously, d1 |a; d1 |r because
r = b qa and d1 |b. This means that d1 is a common divisor of a and r. Thus
d1 d2 . On the other hand, d2 |a; d2 |b because b = qa + r and d2 |r. This means
that d2 is a common divisor of a and b. Hence, d2 d1 . Therefore d1 = d2 .

The above proposition gives rise a simple constructive method to calculate gcd
by repeating the Division Algorithm. For example, gcd(297, 3627) can be calculated
as follows:
3627 = 12 297 + 63,
297 = 4 63 + 45,
63 = 1 45 + 18,
45 = 2 18 + 9,
18 = 2 9;

gcd(297, 3627) = gcd(63, 297)

= gcd(45, 63)
= gcd(18, 45)
= gcd(9, 18)
= 9.

The procedure to calculate gcd(297, 3627) applies to any pair of nonnegative in-
tegers. Let a be a positive integer and b a nonnegative integer. Repeating the
Division Algorithm will produce finite sequences of nonnegative integers qi and ri
such that
b = q 0 a + r0 , 0 r0 < a,
a = q 1 r0 + r1 , 0 r1 < r0 ,
r0 = q 2 r1 + r2 , 0 r2 < r1 ,
r1 = q 3 r2 + r3 , 0 r3 < r2 ,
..
.
rk2 = qk rk1 + rk , 0 rk < rk1 ,
rk1 = qk+1 rk + rk+1 , rk+1 = 0.

Notice that the sequence {ri } is strictly decreasing; it ends eventually to 0 at some
step, say, the remainder rk+1 becomes zero in the very first time, i.e., rk+1 = 0 and
ri 6= 0 for all 0 i k. Reverse the sequence {ri }ki=0 and make substitutions as
follows:
gcd(a, b) = rk ,
rk = rk2 qk rk1 ,
rk1 = rk3 qk1 rk2 ,
..
.
r1 = a q 1 r0 ,
r0 = b q0 a.

We see that gcd(a, b) can be expressed as an integral linear combination of a and

b. This procedure is known as the Euclidean Algorithm.
22 2. NUMBER THEORY

Example 2.1. The greatest common divisor of 297 and 3627, written as an
integral linear combination of 297 and 3627, can be obtained as follows:
gcd(297, 3627) = 45 2 18
= 45 2(63 45)
= 3 45 2 63
= 3(297 4 63) 2 63
= 3 297 14 63
= 3 297 14(3627 12 297)
= 171 297 14 3627.
Proposition 2.6. For integers a, b, c, if a|bc and gcd(a, b) = 1, then a|c.
Proof. By the Euclidean Algorithm, there are integers x and y such that
ax + by = 1. Then
c = 1 c = (ax + by)c = acx + bcy.
It is clear that a|c because a|ac and a|bc.
Theorem 2.7 (Unique Factorization). Every integer a 2 can be uniquely
factorized into the form
a = pe11 pe22 pemm ,
where p1 , p2 , . . . , pm are distinct primes, e1 , e2 , . . . , em are positive integers, and
p1 < p 2 < < p s .
Proof. If a has only the trivial divisors, then a itself is a prime, and it obvi-
ously has unique factorization. If a has some nontrivial divisors, then
a = bc
for some positive integers b and c other than 1 and a. Obviously, b < a and
c < a. By induction, the positive integers b and c have factorizations into primes.
Consequently, a has a factorization into primes.
Let a = q1f1 q2f2 afnn be another factorization, where q1 , q2 , . . . , qn are distinct
primes, f1 , f2 , . . . , fn are positive integers, and q1 < q2 < < qn . We claim that
m = n, pi = qi , ei = fi for all 1 i m.
Suppose p1 < q1 . Then p1 is distinct from the primes q1 , q2 , . . . , qn . It is
clear that gcd(p1 , qi ) = 1 and so gcd(p1 , qifi ) = 1 for all 1 i n. Note that
p1 |q1f1 q2f2 afnn . Since gcd(p1 , q1f1 ) = 1, by Proposition 2.6, we have p1 |q2f2 afnn .
Since gcd(p1 , q2f2 ) = 1, again by Proposition 2.6, we have p1 |q3f2 afnn . Repeating
the argument, we finally obtain that p1 |qnfn , which is contrary to gcd(p1 , qnfn ) = 1.
We thus conclude that p1 q1 . Similarly, p1 q1 . Hence p1 = q1 . Next we claim
that e1 = f1 .
Suppose e1 < f1 . Then
pe22 pemm = pf11 e1 q2f2 qnfn .
This implies that p1 |pe22 pemm . If m = 1, it would imply that p1 divides 1, which
is impossible because p1 is a prime. If m 2, note that gcd(p1 , pi ) = 1 and so
gcd(p1 , pei i ) = 1 for all 2 i m; by the same token of applying Proposition 2.6
repeatedly, we have p1 |pemm , which is contrary to gcd(p1 , pemm ) = 1. This means that
we must have e1 f1 . Similarly, e1 f1 . Hence e1 = f1 .
Now we have obtained pe22 pemm = q2f2 qnfn . If m < n, then by induction we
fm+1
have p1 = q1 , . . . , pm = qm and e1 = f1 , . . . , em = fm . Thus 1 = qm+1 qnfn ; this
 2.3. LEAST COMMON MULTIPLE 23

is impossible because qm+1 , . . . , qn are primes. So m n. Similarly, m n. Hence

we have m = n. By induction on m = n, we obtain that e2 = f2 , . . . , em = fm .
Proposition 2.8. A positive integer d is the gcd of integers a and b if and
only if
(a) d|a and d|b,
(b) if c|a and c|b, then c|d.
Proof. The conditions are obviously sufficient. For necessity, it is clear that
(a) is necessary. If d = gcd(a, b), then there exist integers x and y such that
ax + by = d. Thus for any common divisor c of a and b, c obviously divides the
linear combination ax + by; consequently, c|d.

2.3. Least Common Multiple

For two integers a and b, a positive integer m is called a common multiple of
a and b if a|m and b|m. The smallest integer among the common multiples of a and
b is called the least common multiple of a and b, and is denoted by lcm(a, b).
Proposition 2.9. For any nonnegative integers a and b,
ab = gcd(a, b) lcm(a, b).
Proof. Let a = penn and b = pf11 pf22 pfnn , where p1 < p2 < < pn ,
pe11 pe22
ei and fi are nonnegative integers, 1 i n. Then by the Unique Factorization
Theorem,
gcd(a, b) = pg11 pg22 pgnn ,
lcm(a, b) = ph1 1 ph2 2 phnn ,
where gi = min(ei , fi ) and hi = max(ei , fi ) for all 1 i n. It is clear that
ab = pg11 +h1 pg22 +h2 pngn +hn ,
and gi + hi = ei + fi for all 1 i n.
Proposition 2.10. An integer m is the least common multiple of integers a
and b if and only if
(1) a|m, b|m;
(2) if a|n and b|n, then m|n.
Proof. By the Unique Factorization Theorem.
Example 2.2. Find all integer solutions for the linear equation 25x+65y = 10.
Solution: By the Division Algorithm,
65 = 2 25 + 15,
25 = 15 + 10,
15 = 10 + 5.
Then by the Euclidean Algorithm,
gcd(25, 65) = 15 10
= 15 (25 15)
= 25 + 2 15
= 25 + 2 (65 2 25)
= 5 25 + 2 65.
24 2. NUMBER THEORY

Thus the integer solutions are given by

x = 2(5) + 13k = 10 + 13k
k Z.
y = 2 2 5k = 4 5k,
Theorem 2.11. Let a and b be integers, not simultaneously zero, and d =
gcd(a, b). Then the linear Diophantine equation
ax + by = c
has an integer solution if and only if d|c. Moreover, if (x, y) = (u, v) is a special
solution, then all solutions are given by
(
bk
x = u + gcd(a,b)
ak k Z.
y = v gcd(a,b) ,

Proof. It is clear that (x, y) = (u, v) + d1 (kb, ka) are integer solutions. We
only need to show that all integer solutions of ax + by = 0 are of the form
k
(x, y) = (b, a).
d
Since ax = by, we have a|by and b|ax. This means that ax and by are both
common multiples of a and b. Thus lcm(a, b) is a common divisor of ax and by,
that is, ax = k lcm(a, b) and by = k lcm(a, b) for some k Z. Thus
( k lcm(a,b) bk
x = a = gcd(a,b)
y = k lcm(a,b)
b
ak
= gcd(a,b) .

2.4. Modulo Integers

Given a positive integer n. We consider the equivalence relation n defined by
x n y if and only if y x 0 (mod n).
It is clear that there are n equivalence classes; the quotient set Z/ n is denoted
by Zn , that is, n o
Zn = [0], [1], [2], . . . , [n 1] ,
where [a] = {kn + a | k Z} for any integer a. A standard notation for Zn is Z/nZ.
We can define an addition and a multiplication on Zn in a natural way:
[a] + [b] = [a + b],
[a] [b] = [ab].
The element [0] is the zero in Zn and [1] is the identity in Zn . Sometimes, we
just write [a] as a. However, for [a] = [b], we write it as
a b (mod n).
The addition and multiplication are well defined for integers modulo n. In fact, let
a, a0 , b, b0 be integers such that [a] = [a0 ] and [b] = [b0 ], that is,
a a0 (mod n) and b b0 (mod n).
Then a0 a = kn and b0 b = ln for some k, l Z. Thus
(a0 + b0 ) (a + b) = (a0 a) + (b0 b) = (k + l)n,
 2.4. MODULO INTEGERS 25

a0 b0 ab = a0 b0 ab0 + ab0 ab = (a0 a)b0 + a(b0 b) = (kb0 + al)n.

This means that [a + b] = [a0 + b0 ] and [ab] = [a0 b0 ], that is,
a + b a0 + b0 (mod n) and ab a0 b0 (mod n).
Proposition 2.12. The addition and multiplication in Zn satisfy the following
properties:

(1) [a] + [b] + [c] = [a] + [b] + [c] ,
(2) [a] + [b] = [b] + [a],
(3) [a] + [0] = [a],
(4) [a]([b]
+ [c]) = [a][b]
+ [a][c],
(5) [a][b] [c] = [a] [b][c] ,
(6) [a][b] = [b][a],
(7) [a][1] = [a],
(8) [a][0] = [0].
For any [a] Zn , there is unique element [b] such that [a] + [b] = [0]; we write
[b] = [a], called the negative element of [a]. An element [a] Zn is called
invertible if there is an element [b] in Zn such that
[a][b] = [1];
the element [b] is called the inverse of [a], and is denoted by [a]1 . If [a] is invertible,
its inverse is unique. In fact, if [c] is also an inverse of [a], that is, [a][c] = [1], then
[b] = [1][b] = ([a][c])[b] = ([c][a])[b] = [c]([a][b]) = [c][1] = [c].
We denote by Zn the set of all invertible elements of Zn .
Theorem 2.13. An element [a] in Zn is invertible if and only if a is coprime
with n. Thus n o

Zn = [a] gcd(a, n) = 1 .

Proof. By definition of invertibility, [a] is invertible there is an element

[b] such that [a][b] = [1] ab 1 (mod n) ab = kn + 1 for some k Z
ab nk = 1 for some k Z gcd(a, n) = 1.
Corollary 2.14. If p is a prime, then every nonzero element in Zp is invert-
ible.
For any positive integer n, let (n) denote the number of integers a in [1, n]
such that gcd(a, n) = 1. The function : P P is known as the Euler function.
If p is a prime, then (p) = p 1.
Theorem 2.15 (Eulers Theorem). If [a] is invertible in Zn , that is, gcd(a, n) =
1, then
a(n) 1 (mod n).
Proof. Let m = (n). Then Zn has m elements. Let
Zn = {[a1 ], [a2 ], . . . , [am ]} and [u] = [a1 ][a2 ] [am ].
Since [a] is invertible, there is a one-to-one correspondence f[a] : Zn Zn defined
by
f[a] ([x]) = [a][x] = [ax].
26 2. NUMBER THEORY

The inverse of f[a] is the map f[a]1 = f[b] , where [a][b] = [1]. Then
[u] = [a1 ][a2 ] [am ] = [aa1 ][aa2 ] [aam ] = [am ][a1 ][a2 ] [am ]
Thus [am ] = [1]. This means that a(n) 1 (mod n).
Corollary 2.16 (Fermats Little Theorem). Let p be a prime. If p - a,
then
ap1 1 (mod p).
Proof. Apply the Euler Theorem and (p) = p 1.

2.5. RSA Cryptography System

Common sense tells us that if we know how a secret message was encoded
then we can easily decode it. For example, suppose we are given the message
EJT DSF U F N BU IF N BU JDT , and are told that it was encoded by replacing
each letter in the original message by the one that immediately follows it in the
alphabet. To decode the message, all we have to do is to replace each letter in the
message by the one that proceeds it, yielding DISCRET E M AT HEM AT ICS.
Since the equivalence of coding and decoding, the encryption process must be kept
secret for security reason.
In modern communication networks, the network security leads to two situa-
tions: The first was that only highly trusted individuals were allowed to access to
the encoding key. This means that the code could not be used by many people. The
second was just opposite, that many people were given the secret. So the decoding
process was known in principle. The problem with this is that the process is hardly
unknown in practice due to computational complexity.
Imagine a network where a number of individuals (corporations, banks, gov-
ernments) send messages to each other over public wires, so that eavesdropping
is possible. The problem is to ensure the privacy of such communications, as to
guarantee against fake messages (forgeries).
The usual solution to the problem is to send the messages in a cryptography
know only to the two parties involved (the person-in-the-street calls it a code,
but this word already has at least two other meanings for computer scientists, and
it therefore not used here). The difficulty with this solution is that it requires

pre-arrangement, and therefore communication is limited. Also, it requires n2
different cryptography, if n individuals are involved.
Theorem 2.17 (RSA). Let n = pq be a positive number, where p and q are
primes. Let e and d be positive integers such that gcd(e, (n)) = 1 and ed
1 (mod (n)). Let E : Zn Zn be defined by
E(x) = xe (mod n),
and let D : Zn Zn be defined by
D(y) = y d (mod n).
Then E and D are inverses of each other.
Proof. For any x Zn = {0, 1, 2, . . . , n 1},
(D E)(x) = (E D)(x) = xed (mod n).
Our aim is to show that xed x (mod n). We divide it into three cases.
 2.5. RSA CRYPTOGRAPHY SYSTEM 27

Case 1: x = 0. It is trivial that xed x (mod n).

Case 2: gcd(x, n) = 1. Since ed 1 (mod (n)), there exists an integer k such
that ed = k(n) + 1. Then
xed = xk(n)+1 = (x(n) )k x.
Since gcd(x, n) = 1, by the Euler Theorem, x(n) 1 (mod n). Obviously,
xed x (mod n).
Case 3: gcd(x, n) 6= 1. Since n = pq and p, q are primes, we have either x = kp
for some 1 k < q or x = lq for some 1 l < p. In the former case, we have
xed = (kp)k(n)+1 = (kp)k(p1)(q1)+1 = ((kp)q1 )k(p1) (kp)
because (n) = (p)(q) = (p 1)(q 1). Since q - kp, by the Fermat Little
Theorem, (kp)q1 1 (mod q). Thus
xed kp x (mod q).
Note that xed (kp)ed 0 x (mod p). Since gcd(p, q) = 1, it follows that
xed x (mod n).
A RSA public key cryptography system is a tuple (S, n, e, d, E, D), where
S = {1, 2, . . . , n 1}, n = pq, p and q are distinct primes, e and d are positive
integers such that gcd(e, (n)) = 1 and ed 1 (mod (n)), E and D are bijective
functions S S defined by E(x) = xe (mod n) and by D(x) = xd (mod n), re-
spectively. The integer e is called the encryption number and d the decryption
number; the function E is called the encryption map and D the decryption
map.
In a RSA public key cryptography system (S, n, e, d, E, D), the numbers n and e
are public, while the numbers p, q, d are secret. Mathematically speaking, knowing
the numbers n and e in (S, n, e, d, E, D), the numbers p, q, d are known in principle.
However, if the primes p and q are selected to large enough, say having 100 digits,
then it is impossible in practice to find out the numbers p, q, d. The difficulty to
find p, q, d is based on the difficulty to factor integers.
Example 2.3. Let p = 7, q = 11. Then n = pq = 77, (n) = (p)(q) =
6 10 = 60. If e = 7, then d = 43. If e = 11, then d = 11. If e = 13, then d = 37. If
e = 17, then d = 53.
Example 2.4. Let p = 11, q = 13. Then n = pq = 143, (n) = (p 1)(q 1) =
120. Then there are RSA systems e = 7, d = 43; e = 11, d = 11; and e = 13, d = 37.
For the RSA system e = 13, d = 37, we have
E(2) = 213 41 (mod 143)
(22 = 4, 24 = 16, 28 = 162 113, 213 = 28 24 2 113 16 2 41); and
D(41) = 4137 2 (mod 143)
(412 108, 414 1082 81, 418 812 17, 4116 172 3, 4132 9,
4137 = 4132 414 41 2).
 CHAPTER 3

Propositional Logic

3.1. Statements
By a mathematical statement (or just statement) we mean a declarative
sentence that is either true or false, but not both. The truth value (true or
false) for any statement can be determined and is not ambiguous in any sense. For
example, the following sentences are statements.
(1) Today is 1st of July 1997.
(2) The course number of Discrete Structure in HKUST is Math132.
(3) The equation x2 + y 2 = z 2 has no positive integer solutions.
(4) There are 7,523,804 people in Hong Kong.
However, many sentences in daily life languages are not mathematical statements.
For instance, the following sentences are not mathematical statements.
(1) How are you?
(2) Hong Kong is a big city.
(3) What a beautiful campus!
(4) This sentence is false.
For the last sentence above, if we say that the sentence is true, then it is false. If, on
the other hand, we claim that the sentence is false, then it is true. Such sentences
will not be considered as mathematical statements. Statements are usually denoted
by lowercase letters such as p, q, r, . . ., etc.

3.2. Connectives
Given several statements, we wish to set up rules by which we can decide the
truth of various combinations of the given statements. New statements can be
formed by using connectives not, and, and or.
The Negation of a statement p is the statement not p, denoted p. The
truth values of p are given by the table
p p
T F
F T
The conjunction of two statements p and q is the statement p and q, denoted
p q. Its truth values are given by the table
p q pq
T T T
T F F
F T F
F F F
29
30 3. PROPOSITIONAL LOGIC

The disjunction of statements p and q is the statement p or q, denoted pq.

Its truth table is
p q pq
T T T
T F T
F T T
F F F
The conditional implication from a statement p to a statement q is the state-
ment if p, then q. The statement p is called the hypothesis of this implication
and q the conclusion. This logical connector is symbolized by p q, and its truth
table is defined by
p q pq
T T T
T F F
F T T
F F T
Whenever p is false, the implication is irrelevant and the argument is valid for any
conclusion, thus it was assigned the true value T.

Example 3.1. Let

p : It is a week day.
q : I go to school.
Then the statement p q is the sentence
If it is a week day, then I go to school.
This example may help the read to understand why the truth table for p q
is given above. Let us say that, suppose it is really a week day, and I did go to
school; then the statement is logically right; so the statement p q receives a
true value T. Suppose it is really a week day and I did not go to school when it is
a week day; then there is something wrong; so the statement p q receives a false
value F. However, suppose it is not a week day (say weekend or holiday); then I
dont need go to school, so it is all right either I go to school or not go to school;
the statement p q always receive a true value T.
The Biconditional Implication of statements p and q is the statement (p
q) (q p). Its truth table is given by
p q pq
T T T
T F F
F T F
F F T
Let p and q be statements. The converse of the statement p q is the state-
ment q p. The inverse of p q is the statement p q. The contrapositive
form of p q is the statement q p.

Example 3.2. Let

p : I got SARS.
q : I stayed in hospital.
 3.3. TAUTOLOGY 31

The converse, inverse, and contrapositive forms of p q are given, respectively, as

follows:
qp: If I stayed in hospital, then I got SARS.
p q : If I didnt get SARS, then I didnt stay in hospital.
q p : If I didnt stay in hospital, then I didnt get SARS.
Sometimes we need to consider a family of statements P (x) indexed by a vari-
able x (such statement form indexed by a variable is called a predicate). We
need universal quantifier to express whether all of the statements are simultane-
ously true or one of them is false. We also need existential quantifier to express
whether at least one of the statements is true or all of them are false. The universal
quantification of a predicate P (x) is the statement for all values of x P (x) is true,
denoted x P (x). This means that the statement x P (x) has true value when
all P (x) have true value and x P (x) has false value when one of P (x) has false
value. For example, let P (x) denote x + 1 < 4, where x are real numbers. Then
x P (x) is a false statement because P (4) is not a true statement. The existential
quantification of a predicate P (x) is the statement there exists a value of x for
which P (x) is true, denoted x P (x). This means that x P (x) has true value
when there is at least one x such that P (x) has true value and x P (x) has false
value when all statements P (x) have false value. For example, let Q(x, y, z) denote
x2 + y 2 = z 2 . Then xyz Q(x, y, z) is a true statement, because Q(3, 4, 5) is a
true statement.
Note that here the index x in a predicate is not a propositional variable and
its values are sometimes specified by its domain X. So we may have sentences
x X, P (x) and x X, P (x). For instance, let x be the square root of
real numbers x and let P (x) denote the statement that x is irrational. Then the
statement

for all primes x the number x is irrational
can be expressed as primes x, P (x).

3.3. Tautology
A statement is called a tautology if it is always true for all possible values of its
propositional variables; a contradiction if it is always false; and a contingency if
it can be either true or false, depending on the truth values of its propositional vari-
ables. For instance, (p q) q is a tautology; (p q) p q is a contradiction;
and (p q) p is a contingency.
Two statements p and q are said to be logically equivalent or simply equiv-
alent, written p q or even p = q, if p q is a tautology; that is, p and q have
the same truth values.
Proposition 3.1. Let p, q, r be arbitrary statements. Then
(1) p q q p
(2) p q q q
(3) p (q r) (p q) r
(d) p (q r) (p q) r
(4) p (q r) (p q) (p r)
(5) p (q q) (p q) (p r)
(6) p p p
(7) p p p
32 3. PROPOSITIONAL LOGIC

(8) (p) p
(9) (p q) p q
(10) (p q) p q
Example 3.3. (p q) (p) q is a tautology.
p q pq p p q (p q) (p q)
T T T F T T
T F F F F T
F T T T T T
F F T T T T
Example 3.4. (p q) (q p) is a tautology.
p q pq q p q p (p q) (q p)
T T T F F T T
T F F T F F T
F T T F T T T
F F T T T T T
For instance, let us consider the statement
If I got SARS, then I stayed in hospital.
Let p denote I got SARS and let q denote I stayed in hospital. One feel that
statement If I got SARS, then I stayed in hospital is logically equivalent to the
statement
If I didnt stay in hospital, then I didnt get SARS.
It is also logically equivalent to
I didnt get SARS or I stayed in hospital.
Theorem 3.2. (1) (p q) (p) q
(2) (p q) (q p)
(3) (p q) (p q) (q p)
Theorem 3.3. (1) (x P (x)) x P (x)
(2) (x P (x)) x P (x)
(3) x (P (x) Q(x)) (x P (x)) (x Q(x))
(4) x (P (x) Q(x)) (x P (x)) (x Q(x))
(5) (x P (x)) (x Q(x)) x (P (x) Q(x)) is a tautology.
(6) x (P (x) Q(x)) (x P (x)) (x Q(x)) is a tautology.
(7) ((x P (x)) (x Q(x))) x (P (x) Q(x)) is a tautology.
(8) x (P (x) Q(x)) (x P (x)) (x Q(x))
Proof. (1)-(4) are trivial.
(5) If the statement (x P (x)) (x Q(x)) has T value, then (x P (x)) = T or
(x Q(x)) = T , say (x P (x)) = T . Obviously, (x P (x) Q(x)) = T . Note that
(x P (x)) (x Q(x)) and (x P (x) Q(x)) are not equivalent.
(6) It is an equivalent form of (5).
(7) (x P (x)) (x Q(x)) (x P (x))(x Q(x)) (x P (x))(x Q(x));
x (P (x) Q(x)) x (P (x) Q(x)). The tautology follows from (5).
 3.4. METHODS OF PROOF 33

(8) (x P (x) Q(x)) = (x P (x)Q(x)). It follows from (4) that (x P (x)

Q(x)) is equivalent to
(x P (x)) (x Q(x)) = (x P (x)) (x Q(x)) = (x (x)) (x Q(x)).

Definition 3.4. A subset of connectives is called adequate if every statement
can be represented by a statement form containing only connectives from that
subset.
Theorem 3.5. The subset {, , } is adequate. In this adequate subset, can
be replaced by either or ; and can be replaced by .

3.4. Methods of Proof

Let p and q be statements. If p q is a tautology, then we say that q follows
logically from p, and write p q. The statement p q is also called a theorem.
For statements p1 , p2 , , pn , if
(p1 p2 pn ) q,
that is, (p1 p2 pn ) q is a tautology, we say that q follows logically from
p1 , p2 , , pn , and write
p1
p2
..
.
pn
q
The statements p1 , p2 , , pn are called the hypothesis (or premises) and q the
conclusion. To prove the theorem p q, it means to show that the implication
p q is a tautology. Arguments based on tautology are called rules of inference.
The true of rules of inference is universal, and is independent of the context of the
truth values of the simple statements involved.

Modus Ponens, also called Rule of Detachment (method of affirming), is

the inference
p
pq
q
This means that the statement (p (p q)) q is a tautology.
p q p q p (p q) (p (p q)) q
T T T T T
T F F F T
F T T F T
F F T F T

Law of Syllogism, also known as Chain Rule, is the inference

pq
qr
pr
34 3. PROPOSITIONAL LOGIC

This means that the statement ((p q) (q r)) (p r) is a tautology. In

fact, the statement has false value if and only if (p q) (q r) has true value
and p r has false value. Then both p q and q r have true value, but p
must have true value and r must have false value. Thus q must have true value
by the true value of p q. Since q r has true value, then r has true value, a
contradiction.

Example 3.5. If two integers a and b are even, then their sum a + b is even.

Proof.
Statement Reason
1. a = 2a0 , b = 2b0 . Hypothesis and the definition of even
2. a + b = 2a0 + 2b0 . Step 1 and the meaning of = and +
3. a + b = 2(a0 + b0 ) = 2c. Factoring
4. a + b is even. Step 3 and the definition of even
Note that we have not yet proved that a + b is even; we have simply proved If a
and b are even, then a + b is even. The above informal argument can be made into
the following formal argument.
Symbol Statement Reason
1. p a and b are even. Hypothesis
2. p q If a and b are even, then Definition of even
a = 2a0 and b = 2b0 .
3. q r If a = 2a0 and b = 2b0 , Meaning of = and +
then a + b = 2a0 + 2b0 .
4. p r If a and b are even, then Steps 2 and 3, and
a + b = 2a0 + 2b0 the Law of the Syllogism
5. r s If a + b = 2a0 + 2b0 , then Factoring
a + b = 2(a0 + b0 ) = 2c.
6. p s If a and b are even, then Steps 4 and 5, and
a + b = 2c. the Law of the Syllogism
7. s t If a + b = 2c, then a + b Definition of even
is even.
8. p t If a and b are even, then Steps 6 and 7, and
a + b is even the Law of Syllogism
9. t a + b is even. Steps 1 and 8, and
the Rule of Detachment

Proof by Contradiction, also called Modus Tollens (method of denying),

is the inference
pq
q
p
This means that the statement ((p q) q) p is a tautology.
Example 3.6. If n2 is even, then n is even. (p q)
 3.6. BOOLEAN FUNCTIONS 35

Proof.
Statement Reason
1. n = 2m + 1 Definitiomn of odd
2. n2 = (2m + 1)2 = 4m2 + 4m + 1 Meaning of =, +; and
= 2(2m2 + 2m) + 1 = 2M + 1 using of algebra
3. n2 is odd Step 2 and definition of odd
The informal argument can be made into the following formal argument.
Symbol Statement Reason
1. q n is not even. Denying
2. q r If n is not even, then n = 2m + 1. Meaning of not even
3. r s If n = 2m + 1, then Using of algebra
n2 = 2(2m2 + 2m) + 1 = 2M + 1.
4. q s If n is not even, then n2 = 2M + 1. Law of Syllogism
5. st If n2 = 2M + 1, then n is odd. Definition of odd
6. q t If n is not even, then n2 = 2M + 1. Law of Syllogism
7. t n2 = 2M + 1. Rule of Detachment
8. p n2 is not even. Meaning of not even

3.5. Mathematical Induction

Mathematical Induction (MI) is the following inference about a family of
statements P (k), indexed by positive integers k
P (1)
k P (k) P (k + 1)
k P (k)
Mathematical Induction is a consequence of applying the Modus Ponens and the
Law of Syllogism again and again.
Example 3.7. For any positive integer n,
n
X n(n + 1)(2n + 1)
i2 = 12 + 22 + + n2 = .
i=1
6

3.6. Boolean Functions

The truth values T and F are sometimes denoted by 1 and 0, and write B =
{0, 1}. The nth product B n is called the n-dimensional binary space. Any function
f : B n B is called a Boolean function of n variables; the value of f for Boolean
variables (x1 , . . . , xn ) B n is denoted by f (x1 , . . . , xn ); the variables x1 , . . . , xn and
their negations x 1 , . . . , x
n are called literals. The variables x1 , . . . , xn care called
Boolean variables and can be viewed as variables of statements. A Boolean
formula is a sentence consisting of some literals and connectives and .
Theorem 3.6. Any Boolean function can be written as a Boolean formula with
literals and connectives of disjunction and conjunction.
36 3. PROPOSITIONAL LOGIC

Proof. We proceed by induction on the number of variables. For one variable

x, there are four Boolean functions: f1 (1) = f1 (0) = 1; f2 (1) = f2 (0) = 0; f3 (1) =
1, f3 (0) = 0; f4 (1) = 0, f4 (0) = 1. It is clear that
f1 (x) = x x
; f2 (x) = x x
; f3 (x) = x; f4 (x) = x
.
Assume it is true for n 1 variables; consider a Boolean function f of n variables
(x1 , . . . , xn ). Define two Boolean functions of n 1 variables as follows:
(3.1) g1 (x2 , . . . , xn ) = f (1, x2 , . . . , xn ),
(3.2) g0 (x2 , . . . , xn ) = f (0, x2 , . . . , xn ).
Then it is not hard to verify
f (x1 , . . . , xn ) = (x1 g1 (x2 , . . . , xn )) (
x1 g0 (x2 , . . . , xn )).
By induction hypothesis, g1 and g0 can be written as Boolean formulas of literals
and connectives of disjunction and conjunction; so does f .

Example Given the Boolean function

(x1 , x2 , x3 ) f (x1 , x2 , x3 )
(0,0,0) 1
(0,0,1) 1
(0,1,0) 0
(0,1,1) 0
(1,0,0) 0
(1,0,1) 1
(1,1,0) 0
(1,1,1) 1
It can be written as
f (x1 , x2 , x3 ) = (x1 g1 (x2 , x3 )) (
x g0 (x2 , x3 )),
where
g1 (x2 , x3 )= f (1, x2 , x3 )
= (x2 g11 (x3 )) (x2 g10 (x3 )),
g11 (x3 ) = f (1, 1, x3 ) = x3 ,
g10 (x3 ) = f (1, 0, x3 ) = x3
and
g0 (x2 , x3 ) = (x2 g01 (x3 )) (
x2 g00 (x3 ))
g01 (x3 ) = f (0, 1, x3 ) = x3 x
3 ,
g00 (x3 ) = f (1, 1, x3 ) = x3 x
3 .
Then
g1 (x2 , x3 ) = (x2 x3 ) (
x2 x3 ) = x3 ,
g0 (x2 , x3 ) = (x2 x3 x3 ) (
x2 (x3 x
3 )) = x
2 .
Hence
f (x1 , x2 , x3 ) = (x1 x3 ) (
x1 x
2 ).
 3.6. BOOLEAN FUNCTIONS 37

EXERCISES
(1) Consider the statement
If 1 = 4, then 1 = 2.
Proof. Since 1 + 3 = 4 and 1 = 4, we have 0 = 3. Dividing both sides of
0 = 3 by 3, we further have 0 = 1. Hence 1 = 0 + 1 = 1 + 1 = 2. Is the
proof a true argument? What can you conclude from the statement and
proof?
(2) Define the connectives and by
p q pq p q pq
T T F T T F
T F F and T F T
F T F F T T
F F T F F F
respectively. Find the truth tables for
(p q) r, (p q) (p r), (p q) (p r), (p q)p, (pq)(qr).
(3) Let
p : John is a student of Computer Science Department in HKUST.
q : John takes course Math132
Write the English sentences of the converse, inverse and contrapositive
forms for the statement p q. Write the English sentence for the state-
ments
p q, q p
(4) Show that the set {, , } is adequate. Are the sets {, , } and
{, , } adequate?
(5) Show that the statement
(x P (x)) (x Q(x)) x (P (x) Q(x))
is a tautology. Is the converse of the statement a tautology? If yes, prove
it. If no, find a counterexample.
(6) Show that if statements p and p q are tautologies then q is a tautology.
Give a daily life example of the argument.
(7) Express p q and pq in terms of p, q, and other connectives without
and .
(8) If p q and q r are tautologies, then p r is a tautology.
(9) If p q and q are tautologies, then p is a tautology.
 CHAPTER 4

Combinatorics

4.1. Counting Principle

For two tasks T1 and T2 to be performed in sequence, if the task T1 can be
performed in m ways, and for each of these m ways the task T2 can be performed
in n ways, then the task sequence T1 T2 can be performed in mn ways. Using the
set language notation, let X be the set of plans to have task T1 performed and Y
the set of plans to have the task T2 performed, then the product
X Y = {(x, y) | x X, y Y }
is the set of plans to have the task sequence T1 T2 to be performed, and
|X Y | = |X||Y |.
Example Suppose a lady has three hats, seven shirts, five skirts, and four pairs of
shoes. Assume all hats, shirts, skirts, and shoes are distinct. In how many ways
can the lady dress herself by selecting each from the hats, the shirts, the skirts, and
the shoes?
answer = 3 7 5 4 = 420.
Example Courses Calculus, Linear Algebra, and Discrete mathematics are taught
by twenty, fifteen, and ten different instructors respectively in a meg university. In
how many ways can a student take two of the three courses by selecting instructors?
answer = 20 15 + 20 10 + 15 10 = 650.
Let X and Y be arbitrary finite sets and let f : X Y be a function from
X to Y . If the inverse image f 1 (y) = {x X | f (x) = y} has equal number of
elements for all y Y , that is, |f 1 (y)| = k, then
|X|
|Y | = .
k

4.2. Permutations
Let A be a set of n objects. An arrangement of r elements from A in linear
order is called an r-permutation of n objects. The number of r-permutations of
n objects is denoted by P (n, r). In forming an r-permutation of n objects, the 1st
element can be selected in n choices, the 2nd element in n 1 choices, and so on.
Thus
P (n, r) = n(n 1) (n r + 1).
In particular, when r = n, an n-permutation of n objects is simply called a per-
mutation of n objects. The number of permutations of n objects is given by
n! := n(n 1)(n 2) 3 2 1,
39
40 4. COMBINATORICS

where n! is read n factorial.

Example Find the number of possible seating plans for four persons to be seated
at a round table.
Solution. Let A = {a, b, c, d} be the set of four persons. We denote by X the set
of all permutations of A and by Y the set of all round permutations of A. We
define a map f : X Y such that for any permutation x1 x2 x3 x4 , f (x1 x2 x3 x4 ) is
the round permutation by letting x1 sit next to x4 and keeping all other persons
in same order. It is clear that the map f is onto; and for each round permutation
there are 4 permutations sending to the same round permutation. For instance, the
four permutations

x1 x2 x3 x4 , x2 x3 x4 x1 , x3 x4 x1 x2 , x4 x3 x2 x1

are sent to a same round permutation. We then have

4! = |X| = 4|Y |.

Therefore
4!
|Y | = = 3! = 6.
4
Proposition 4.1. The number of round permutations of n objects is

n!
= (n 1)!.
n

Corollary 4.2. The number of necklaces with n ( 3) distinct marbles is

(n 1)!
.
2
Elements in a set are always considered distinct. When considering indistin-
guishable elements we need the concept of multisets. By a multiset we mean a
collection of objects such that some of them may be identically the same, called
indistinguishable. Let A be a multiset of n objects such that there are k distin-
guishable types of objects. If there are ni indistinguishable objects for the ith type,
where 1 i k, the multiset is called a multiset of type (n1 , n2 , . . . , nk ).

Example How many ways to arrange 6 color balls of the same size, of which two
are white, 3 are black, and 2 is red, in linear order?
Solution. Let the 6 balls be represented by w, w, b, b, b, r, r. Let us label
the balls of the same color by numbers to get 6 distinct balls w1 , w2 , b1 , b2 , b3 ,
r1 , r2 . Then each permutation of the 6 balls without labels produces 12 = 2!3!2!
permutations of the 6 distinct balls (6 color balls with labels). More precisely, let X
be the set of permutations of b1 , b2 , b3 , w1 , w2 , r1 , r2 and Y the set of permutations
of b, b, b, w, w, r, r. Then there is a map f : X Y , sending each permutation of
b1 , b2 , b3 , w1 , w2 , r1 , r2 to a permutation of b, b, b, w, w, r, r by erasing the labels. For
 4.2. PERMUTATIONS 41

instance,


r1 b1 w1 b2 b3 r2 w2 r2 b1 w1 b2 b3 r1 w2





r1 b1 w1 b3 b2 r2 w2 r2 b1 w1 b3 b2 r1 w2





r1 b2 w1 b1 b3 r2 w2 r2 b2 w1 b1 b3 r1 w2





r1 b2 w1 b3 b1 r2 w2 r2 b2 w1 b3 b1 r1 w2





r1 b3 w1 b1 b2 r2 w2 r2 b3 w1 b1 b2 r1 w2


r1 b3 w1 b2 b1 r2 w2 r2 b3 w1 b2 b1 r1 w2 f
2!3!2! 7 rbwbbrw.

r1 b1 w2 b2 b3 r2 w1 r2 b1 w2 b2 b3 r1 w1





r1 b1 w2 b3 b2 r2 w1 r2 b1 w2 b3 b2 r1 w1





r1 b2 w2 b1 b3 r2 w1 r2 b2 w2 b1 b3 r1 w1





r1 b2 w2 b3 b1 r2 w1 r2 b2 w2 b3 b1 r1 w1





r1 b3 w2 b1 b2 r2 w1 r2 b3 w2 b1 b2 r1 w1


r1 b3 w2 b2 b1 r2 w1 r2 b3 w2 b2 b1 r1 w1
Clearly, f is onto. For each permutation P of b, b, b, w, w, r, r, its inverse image
f 1 (P ) consists of 2!3!2! permutations of b1 , b2 , b3 , w1 , w2 , r1 , r2 . Thus |X| = 24|Y |,
that is,
|X| 7!
|Y | = = = 210.
2!3!2! 2!3!2!
Theorem 4.3. The number of permutations of n objects of type (n1 , n2 , . . . , nk )
is
n!
.
n1 !n2 ! nk !
Example How many ways to put five same calculus books, three same physics
books, and two same chemistry book in a bookshelf?
10!
answer = = 2520.
5!3!2!
Corollary 4.4. The number of sequences of 0 and 1 of length n with exact r
1s and (n r) 0s is given by
n!
.
r!(n r)
Example Counting the number of nondecreasing coordinate paths from the origin
(0,0) to the point (6,4).
Solution. Note that each such path can be viewed as a walk by moving to the right
and up. If we denote the moving of one unit to the right by R and the moving of
one unit up by U , then each such path can be viewed as a sequence of R and U of
length 10 with 6 Rs and 4 U s. Thus the answer is
10!
= 210.
6!4!
Proposition 4.5. The number of non-decreasing coordinate paths from (0, 0)
to (a, b), where a and b are both non-negative integers, is given by
(a + b)!
.
a!b!
Thinking Problem Find a formula for the number of round permutations of n
objects of type (n1 , n2 , . . . , nk ).
42 4. COMBINATORICS

Example How many possible seating plans can be made for r people to be seated
at a round table of n seats, leaving n r seats empty? (Assume n r 1)

Solution. Each seating plan produces n permutations of n objects of type (1, . . . , 1, n

| {z }
r
r). Then the answer is given by

n
1,...,1,nr n!
= .
n (n r)!n
4.3. Combination
A combination is a collection of objects (order is immaterial) from a given
source of objects. An r-combination of n objects is a collection of r objects from
a source of n objects, that is, an r-subset of an n-set. The number of r-combinations
n objects is denoted by n
,
r
read n choose r.

Example Find the number of 3-subsets of a 5-set A = {a, b, c, d, e}

Solution. First Method: Let X be the set of all permutations of the five elements
a, b, c, d, e and let Y be the set of all 3-subsets of A. Then there is a map f :
X Y , sending each permutation x1 x2 x3 x4 x5 to the 3-subset {x1 , x2 , x3 }, that
is, f (x1 x2 x3 x4 x5 ) = {x1 , x2 , x3 }. It is clear that f is onto, and the inverse image
of every 3-subset from Y has 3!2! permutations in X. For instance,


acebd



aecbd



caebd




ceabd





eacbd


ecabd f
12 = 3!2! 7 {a, c, e}

acedb



aecdb



caedb




ceadb





eacdb


ecadb
Thus |X| = 3!2!|Y | and the answer is

5 5!
= = 10.
3 3!2!
Second Method: Let X be the set of 3-permutations of A and Y the set of
3-subsets of A. Define f : X Y by f (x1 x2 x3 ) = {x1 , x2 , x3 } for each 3-
permutation x1 x2 x3 X. Obviously, there are 3! permutations of {x1 , x2 , x3 }
sent to {x1 , x2 , x3 }. We then have
|X| P (6, 3)
|Y | = = .
3! 3!
 4.3. COMBINATION 43

Theorem 4.6. The number of r-combinations of n objects is

n n! P (n, r)
= = .
r r!(n r)! r!

Theorem 4.7. (Binomial Theorem or Binomial Expansion)

n
X n
(x + y)n = xk y nk .
k
k=0

Proof.

(x + y)n = (x + y)(x + y) (x + y)
| {z }
n
X
= u1 u2 un (ui = x or y, 1 i n)
X number of sequences of x and y of
n
=
length n with k xs and (n k) ys
k=0
n
X n k nk
= x y .
k
k=0

A collection of k disjoint subsets of an n-set is called a combination of n ob-
jects of type (n1 , n2 , . . . , nk ) if the k subsets have the cardinalities n1 , n2 , . . . , nk
and n = n1 + n2 + + nk . A combination of n objects of type (n1 , n2 , . . . , nk ) can
be viewed as a placement of n objects into k boxes so that the 1st box contains n1
objects, the 2nd box contains n2 objects, and so on. The number of combinations
of n objects of type (n1 , n2 , . . . , nk ) is denoted by

n
,
n1 , n2 , . . . , nk

read n choose n1 , n2 , dot dot dot, and nk .

Example How many ways can six distinct objects be placed into three boxes so
that the 1st box contains two objects, the 2nd box contains three objects, and the
3rd box contains one object?
Solution. Given a set A = {a, b, c, d, e, f } of six objects. Let X be the set of
permutations of A, and let Y be the set of placements of elements of A into three
boxes so that the 1st box receives two elements, the 2nd box receives thee elements,
and the 3rd box receives one element. Then there is map f : X Y , sending each
permutation x1 x2 x3 x4 x5 x6 of A to the placement {x1 , x2 }{x3 , x4 , x5 }{x6 }. It is
clear that f is onto. For each placement {x1 , x2 }{x3 , x4 , x5 }{x6 }, its inverse image
44 4. COMBINATORICS

f 1 ({x1 , x2 }{x3 , x4 , x5 }{x6 }) has 2!3!1! permutations of A. For instance,

acbdf e ac abf e




acbf de
ac bf d e




acdbf e
ac dbf e




acdf be
ac df b e




acf bde
ac f bd e



acf dbe ac f db e f
2!3!1! 7 a, c b, d, f e {a, c}{b, d, f }{e}

cabdf e
ca bdf e




cabf de
ca bf d e




cadbf e
ca dbf e




cadf be
ca df b e




caf bde
ca f bd e

caf dbe ca f db e

Thus |X| = 2!3!1!|Y |. Therefore the answer is

|X| 6! 6
|Y | = = = = 60.
2!3!1! 2!3!1! 2, 3, 1
Theorem 4.8. The number of ways to place n distinct objects into k distinct
boxes so that the 1st box contains n1 objects, the 2nd box contains n2 objects, . . .,
the kth box contains nk objects is given by

n n!
= .
n1 , n2 , . . . , nk n1 !n2 ! nk !
Note: When considering placement of n objects to two boxes of type (r, n r), we
write
n
n n!
:= = .
r r, n r r!(n r)!
Theorem 4.9. (Multinomial Theorem and Multinomial Expansion)
X
n
n
(x1 + x2 + + xk ) = xn1 1 xn2 2 xnk k .
n +n ++n =n
n1 , n2 , . . . , nk
1 2 k
n1 0,n2 0,...,nk 0

Proof.

(x1 + x2 + + xk )n = (x1 + x2 + + xk ) (x1 + x2 + + xk )

| {z }
n
X
= u1 u2 un (where ui = x1 , x2 , . . . , xk , 1 i n)
X number of sequences of x1 , x2 , . . ., xk of
=
length n with n1 x1 s, n2 x2 s, . . ., nk xk s
X
n
= xn1 1 xn2 2 xnk k .
n +n ++n =n
n1 , n2 , . . . , nk
1 2 k
n1 0,n2 0,...,nk 0

 4.4. COMBINATION WITH REPETITION 45

4.4. Combination with Repetition

We now consider combinations with repetition allowed. The number of r-
combinations of n objects with repetition allowed is denoted by
DnE
.
r
Example How many ways to take seven objects with repetition allowed from a set
A = {a, b, c, d} of four objects?

Solution. Take a collection of seven objects from A with repetition allowed, say,
a, a, b, b, b, c, d, we insert bars (denoted by the symbol 1) between the as and bs,
the bs and the cs, the cs and ds; and denote the objects a, b, c, d by the same
symbol 0. Then each collection of seven objects from A is encoded into a sequence
of 0 and 1 of length 10 with seven 0s and three 1s. For instance,
aa 1 bbb 1 c 1 d 7 0010001010
aaa 1 bbbb 1 1 7 0001000011
1 bb 1 ccccc 1 7 1001000001
a 1 1 ccc 1 ddd 7 0110001000
Note that different collections of seven objects from A with repetition allowed are
encoded into different sequences, and every sequence of 0 and 1 of length 10 with
exact seven 0s and three 1s can be obtained in this way. Thus

4 4+71 10
= = .
7 7 7
Theorem 4.10. The number of r-combinations of n objects with repetition
allowed is DnE n + r 1
= .
r r
Example Eight students plan to have dinner together in a restaurant where the
menu shows 20 different dishes. Now each student decides to order one dish. How
many possible combinations of dishes can be ordered by the students for their din-
ner?

Solution.
20 20 + 8 1 27
Answer = = = .
8 8 8
Theorem 4.11. The number of nonnegative integer solutions for the equation
x1 + x2 + + xn = r
is given by
DnE n+r1
= .
r r

Example There are five types of color T-shirts on sale, black, blue, green, orange,
and white. John is going to buy ten T-shirts; he has to buy at least two blues and
two oranges, and at least one for all other colors. Find the number of ways that
John can select ten T-shirts.
46 4. COMBINATORICS

Solution. We label the colors black, blue, green, orange and white by numbers
1, 2, 3, 4, 5. Let xi be the number of T -shirts that John would select for the ith
color T -shirt. Then the problem is to find the number of integer solutions for the
equation
x1 + x2 + + x5 = 10,
where x1 1, x2 2, x3 1, x4 2, x5 1.

5 5+31 7
Answer = = = = 35.
3 3 3
Example In how many ways can a student order eight dumplings from three dif-
ferent kinds? In how many ways can a student eat eight dumplings from the three
kinds, assuming that there are infinitely many supply of dumplings for each kind
and that the dumplings of the same kind should be eaten consecutively one by one?

3 10
= = 45.
8 8
3
X
k
answer = k!
8k
k=1

4.5. Combinatorial Proof

A proof for a result by using bijection is sometimes called a combinatorial
proof of the result. Here are a few examples.

Example
n
n+1 n
= + .
r r1 r

Proof. Let An be a set {a1 , a2 , . . . , an } of n elements and let An+1 be the {a1 , . . . , an , an+1 }
of n + 1 elements. The r-subsets of An+1 can be divided into two kinds: the r-
subsets
n of An and the r-subsets of An+1 that are not subsets of An . There are
r r-subsets of the first kind. Each r-subset of the second kind must contain the
element an+1 . Thus the r-subsets of the second kind can be obtained by taking all
(r
1)-subsets
of An and adding the element an+1 to each of them; so there are
n
n+1 n n
r1 r-subsets of the second kind. Thus we have r = r + r1 .

1
1 2 1
1 3 3 1
1 4 6 4 1
1 5 10 10 5 1
Example 4.1.
n 2 n 2 n 2
2n
= + + +
n 0 1 n

Proof. Let us consider n-combinations of 2n balls of the same size, of which n balls
are white and the other n balls are black. Each such combination can be obtained
 4.5. COMBINATORIAL PROOF 47

by taking k balls from the n white balls and taking (n k) balls from the n black
balls, where k ranges from 0 to n. Then we have
n n n n n n
2n
= + + +
n 0 n 1 n1 n 0
n 2 n 2 n 2
= + + + .
0 1 n

Exercises
(1) A computer user name consists of three English letters followed by five
digits. How many different user names can be made?
(2) A set lunch is to include a soup, a main course, and a drink. Suppose a
customer can select from three soups, five main courses, and four drinks.
How many different set lunches can be selected.
(3) Give a procedure for determining the number of zeros at the end of n!.
Justify your procedure and make examples for 12! and 26!.
(4) Find the number of different permutations of the letters in HONGKONG.
(5) A bookshelf is to be used to display 10 math books. Suppose there are
8 kinds of calculus books, 6 kinds of linear algebra books, and 5 kinds of
discrete math books. It is required that books for the same subject should
be displayed together.
(a) Find the number of ways to display 10 distinct books so that there
are 5 calculus books, 3 linear algebra books, and 2 discrete math
books.
(b) Find the number of ways to display 10 books (not necessarily distinct)
so that there are 5 calculus books, 3 linear algebra books, and 2
discrete math books.
(6) There are n men and n women to form a circle or a line, n 2. Find the
number of patterns of
(a) circles could be formed so that each man is next to at least one
woman;
(b) straight lines could be formed so that each man is next to at least
one woman.
(7) Four identical six-sided dice are tossed simultaneously and numbers show-
ing on the top faces are recorded as a multiset of four elements. How many
different multisets are possible?
(8) Find the number of non-decreasing coordinate paths from the origin (0, 0, 0)
to the lattice point (a, b, c).
(9) In how many ways can a six-card hand be dealt from a deck of 52 cards.
(10) How many different eight-card hands with five red cards and three black
cards can be dealt from a deck of 52 cards?
(11) Fortune draws are arranged to select six ping pang balls simultaneously
from a box in which 20 are orange and 30 are white. A draw is lucky if
it consists of three orange and thee white balls. What is the chance of a
lucky draw?
(12) Determine the number of integer solutions for x1 + x2 + x3 + x4 38,
where
(a) xi 0, 1 i 5.
48 4. COMBINATORICS

(b) x1 0, x2 2, x3 2, 3 x4 8.
(13) Determne the number of nonnegative integer solutions to the pair of equa-
tions
x1 + x2 + x3 = 8, x1 + x2 + x3 + x4 + x5 + x6 = 18.
(14) Let M be a multiset of type (n1 , n2 , . . . , nk ) such that ni 1 for 1 i k.
If the numbers n1 , n2 , . . . , nk are all coprime with n = n1 + n2 + + nk ,
then the number of round permutations of M is

n
n1 ,n2 ,...,nk
.
n
The formula is actually valid when gcd(n1 , . . . , nk ) = 1, but we didnt
define the gcd yet for more than two integers. Find a counterexample if
the conditions are not satisfied.
(15) Find the number of non-decreasing lattice paths from the origin (0, 0)
to a non-negative lattice point (a, b), allowing only horizontal, vertical,
and diagonal unit moves; that is, allowing moves (x, y) (x + 1, y),
(x, y) (x, y + 1) and (x, y) (x + 1, y + 1).
(16) *Find the number of non-decreasing lattice paths from the origin (0, 0)
to a non-negative lattice point (a, b), allowing arbitrary straight moves
from one lattice point to another lattice point; that is, allowing moves
(x, y) (x + k, y + h), where k and h are non-negative integers such that
(k, h) 6= (0, 0).
Hint for Ex. 14: Let M be a multiset of type (n1 , . . . , nk ) and S(M ) the set of all
permutations of M . Define : S(M ) S(M ) by
(x1 x2 x3 xn ) = x2 x3 xn x1 .
A permutation w = x1 x2 xn is called primitive if the permutations
w, (w), 2 (w), . . . , n1 (w)
are distinct.
For a non-primitive permutation x1 x2 xn of M , there are integers a and b,
0 a < b < n, such that
a (x1 x2 xn ) = b (x1 x2 xn ).
Let l = b a. Then
l (x1 x2 xn ) = x1 x2 xn .
That is,
xl+1 xl+2 xn x1 x2 xl = x1 x2 xnl xnl+1 xnl+2 xn .
This is equivalent to saying that
xl+1 = x1 , xl+2 = x2 , . . . , xn = xnl ; x1 = xnl+1 , x2 = xnl+2 , . . . , xl = xn .
We claim that l|n. In fact, suppose l doesnt divide n, then n = ql+r with 0 < r < l.
Then
xl+1 = x1 , xl+2 = x2 , . . . , x2l = xl , x2l+1 = x1 , . . . , xn = xql+r = xr .
 4.5. COMBINATORIAL PROOF 49

Thus

xl+1 xl+2 xn x1 x2 xl = x1 x2 xl x1 x2 xl x1 x2 xr x1 x2 xl .
| {z } | {z }
| {z }
q1

On the other hand,

xl+1 xl+2 xn x1 x2 xl = x1 x2 xn = x1 x2 xl x1 x2 xl x1 x2 xr .
| {z } | {z }
| {z }
q

This means that

x1 x2 xr x1 x2 xl = x1 x2 xl x1 x2 xr .

Continuing this procedure, one can find a positive integer s such that n = sd and

x1 x2 xn = x1 x2 xs x1 x2 xs .
| {z } | {z }
| {z }
d

Let si be the number of elements of the ith type in the segment x1 x2 xs . Then
si d = ni . So d|ni , 1 i k. Hence d divides gcd(n1 , . . . , nk ). We have shown
that if gcd(n1 , . . . , nk ) = 1 then there is no positive integer l < n such that
` (x1 x2 xn ) = x1 x2 xn , that
is, every
permutation is primitive. Thus the
n
number of round permutations is n1 ,...,nk /n.
Now let m = gcd(n1 , . . . , nk ) and let Dm be the set of divisors of m. Then Dm
is a finite partially ordered set by its divisibility. For each d D, let

f (d) : = number of permutations of type (dn1 /m, . . . , dnk /m),

dn/m
=
dn1 /m, . . . , dnk /m
g(d) : = number of primitive permutations of type (dn1 /m, . . . , dnk /m).

Note that each permutation x1 x2 xdn/m of type (dn1 /m, . . . , dnk /m) can be
classified as a unique primitive permutation x1 x2 xd0 n/m for some d0 such that
d0 |d and
x1 x2 xdn/m = x1 x2 xd0 n/m x1 x2 xd0 n/m .
| {z } | {z }
| {z }
d/d0

Then
X
f (d) = g(d0 ).
d0 |d

By the Mobius inversion formula, we have

X X
d d
g(d) = (d0 )f 0
= 0
f (d0 ).
0
d 0
d
d |d d |d
50 4. COMBINATORICS

Thus the number of round permutations of type (n1 , . . . , nk ) is given by

X 1
RP (n1 , . . . , nk ) = g(d)
dn/m
d|m

1 XX d m
= 0
f (d0 )
n d d
d|m d0 |d

1 X X d m
= f (d0 ) 0

n 0 0
d d
d |m d|m,d |d

1 X X m/d0
= f (d0 ) (k)
n 0 k
d |m k|(m/d0 )
1 X m
= f (d0 ) 0
n 0 d
d |m
1 X m
= f (d)
n d
d|m

1X n/d
= (d).
n n1 /d, . . . , nk /d
d|m

Here is the Mobius function, defined for positive integers by

1 if d = 1,
(d) = (1)k if d is a product of k distinct primes,

0 if d has a repeated prime factor;
and is Eulers function, defined for positive integers by
(d) = number of integers coprime to d in [1, d].

min{a,b}
X a+bk
answer: .
a k, b k, k
k=0
For any such path with k diagonal moves (0 k min{a, b}), the number of
horizontal moves should be a k and the number of vertical moves should be b k.

4.6. Pigeonhole Principle

Theorem 4.12. (Pigeonhole Principle) If n objects are placed into m boxes and
n > m, then there is at least one box which contains two or more objects.
The pigeonhole principle is a common Chinese saying: When pigeons are put
into pigeonholes with more pigeons than pigeonholes, there are at least two pigeons
must be put in a same pigeonhole.

Example Among any five integers between 1 and 8 inclusive, there are at least two
of them adding up to 9.
Solution. We can divide the set {1, 2, , 8} into four disjoint subsets where each
has two elements adding up to 9: {1, 8}, {2, 7}, {3, 6}, and {4, 5}. When selecting
five numbers from these four subsets, at least two of the five selected numbers must
 4.7. RELATION TO PROBABILITY 51

come from a same subset of the four subsets. Thus their addition is 9.

Example Show that in any group of two or more persons there are at least two
having the same number of friends (It is assumed that if a person x is a friend of a
person y then y is also a friend of x).
Solution. Assume that there are n persons in the group. The number of friends of
a person x should be between 0 and n 1. If there is one person x who has n 1
friends, then everyone is a friend of x . So both 0 and n 1 can not be numbers
of friends of some people in the group. Thus the pigeonhole principle tells us that
there are at least two people who have the same number of friends.

Example Show that if a1 , a2 , . . . , ak are integers (not necessarily distinct), then

some of them can be added up to a multiple of k.
Solution. Consider the integers of the following k + 1 objects:
(4.1) 0, a1 , a1 + a2 , a1 + a2 + a3 , , a1 + a2 + + ak .
Note that integers modulo k can only be 0, 1, 2, . . . , k 1. By the Pigeonhole
Principle there are at least two integers in (4.1), say
a1 + + ai and a1 + + aj ,
whose remainders modulo k are the same. The number a1 + + ai could be the
0, the very first element in (4.1). Thus
ai+1 + ai+2 + + aj
is a multiple of k.

Example Given 10 distinct integers a1 , a2 , . . . , a10 such that 0 ai < 100, can we
find a subset of {a1 , . . . , a10 } such that the sum of numbers in the subset with sign
is zero?
Solution. Consider all possible partial sums of the selected numbers a1 , a2 , . . . , a10 .
The values of these sums should be between 0 and 1000. Note that the number of
subsets of 10 objects is 210 = 1024. By the Pigeonhole Principle there are at least
two subsets A and B of {a1 , a2 , . . . , a10 } such that the sum of the elements of A
and the sum of the elements of B are the same, that is,
X X
ai = aj .
ai A aj B

Now we move all elements from the right side to the left; the elements in both A
and B will be canceled. Thus sum of the elements of AB with positive sign for
the elements in A B and negative sign for the elements in B A is equal to 0.
Theorem 4.13. If n objects are placed in m boxes, then one of the boxes must
n n
contain at least d m e objects, where d m e denotes the smallest integer greater or equal
n
to m .

4.7. Relation to Probability

There are lot problems in our daily life about chance, the possibility or proba-
bility. When we flip a coin, we have two possible outcomes, Head and Tail. If the
coin is fair, the chance to have the outcome Head is one-half or 50%. When we
52 4. COMBINATORICS

roll a pair of dice we may have outcomes a collection of pairs of numbers between
1 and 6. The chance of the event of the outcomes that the sum of the pair is even
is one-half. For instance, we may be interested in computing the probability of the
event of the outcomes that the sum of the pair is 8.
Definition 4.14. Any collection of outcomes in a probabilistic experiment is
called an event. If each outcome is equally likely to be happened, we define
Total number of favorite outcomes
Probability of even A = P (A) = .
Total number of possible outcomes
Example What is the probability of selecting three distinct numbers from 1, 2, . . . , 11
so that two are less that 5, one is equal to 5, and four are larger than 5?

Solution. The total number of possible outcomes is 11 7 , and the total number of
4 1 6
favorite outcomes is 2 1 4 . Then the probability is
4 1 6
2 1 4
11 .
7
Example Find the probability that no two persons have the same birthday in a
party of 40 people.
40
Solution. The total number
365 of possible outcomes is 365 and the total number of
favorite outcomes is 40 40!. The probability is
365
40 40!
0.109.
36540
Example What is the probability of rolling a pair of dice so that the sum of
numbers on the top faces is 8?
Solution. Since there is no order between the two dice, there are twenty-one possible
outcomes
{i, j}, 1 i j 6
3
and three favorite outcomes {2, 6}, {3, 5}, {4, 4}. So the answer might be 21 = 17 .
One may color the two dice as black and white so that the two dice are ordered.
There are 36(= 6 6) possible outcomes and five favorite outcomes (2, 6), (3, 5),
5
(4, 4), (5, 3), (6, 2). The answer should be 36 . Which one is correct and why?

Example Find the probability of rolling four dice simultaneously so that the sum
of points is exactly 9.
Solution. The total number of possible outcomes is 64 . The total number of favorite
outcomes is the number of positive integer solutions of the equation
x1 + x2 + x3 + x4 = 9
which is equivalent to the number of non-negative integer solutions of the equation
y1 + y2 + y3 + y4 = 5.
Thus the answer is given by
4
5 7 1
= .
64 162 23
Exercises
 4.8. INCLUSION-EXCLUSION PRINCIPLE 53

(1) Show that there must be 90 ways to choose six numbers from 1 to 15 so
that all the choices have the same sum.
(2) Show that if five points are selected in a square whose sides have length
2, then there are at leat two points whose distance is at most 2.
(3) Prove that if any 14 numbers from 1 to 25 are chosen, then one of them
is a multiple of another.
(4) Twenty disks numbered 1 through 20 are placed face down on a table.
Disks are selected one at a time and turned over until 10 disks have been
chosen. If two of the disks add up to 21, the play loses. Is it possible to
win this game?
(5) Show that it is impossible to arrange the numbers 1, 2, . . . , 10 in a circle
so that every triple of consecutively placed numbers has a sum less than
15.

4.8. Inclusion-Exclusion Principle

Let U be a finite set. For two subsets A1 and A2 of U , we have
|A1 A2 | = |A1 | + |A2 | |A1 A2 |,
or equivalently,
|A1 A2 | = |U | |A1 | |A2 | + |A1 A2 |.
For three subsets A1 , A2 , A3 of U , we have
|A1 A2 A3 | = |A1 | + |A2 | + |A3 |
|A1 A2 | |A1 A3 | |A2 A3 |
+|A1 A2 A3 |,
or equivalently,
|A1 A2 A3 | = |U | |A1 | |A2 | |A3 |
+|A1 A2 | + |A1 A3 | + |A2 A3 |
|A1 A2 A3 |.
These formulas and similar kinds for more subsets are called Inclusion-Exclusion
Principle.

Example The pin numbers Hang Seng Bank card are nonnegative integers of six
digits. How many pin numbers can be made so that the triple 444 doesnt appear?
Solution. Let U be the set of all possible secrete codes. Then |U | = 106 . Let
A1 = set of codes of the form 444xxx,
A2 = set of codes of the form x444xx,
A3 = set of codes of the form xx444x,
A4 = set of codes of the form xxx444,
where x varies from 0 to 9. Then |A1 | = |A2 | = |A3 | = |A4 | = 103 ; |A1 A2 | =
|A2 A3 | = |A3 A4 | = 102 , |A1 A3 | = |A2 A4 | = 10, |A1 A4 | = 1; |A1 A2 A3 | =
|A2 A3 A4 | = 10, |A1 A2 A4 | = |A1 A3 A4 | = 1; |A1 A2 A3 A4 | = 1.
Thus
answer = 106 4 103 + (3 102 + 2 10 + 1) (2 10 + 2) + 1 = 996310.
54 4. COMBINATORICS

Example Find the number of positive integer solutions for the linear equation
x1 + x2 + x3 = 8.
Solution. Let Ai be the set of non-negative integer solutions of the above equation
such that xi = 0, 1 i 3. Then

3 3+81 10
|U | = = = = 45;
8 8 8

|A1 | = |A2 | = |A3 | = 9;

|A1 A2 | = |A1 A3 | = |A2 A3 | = 1;

|A1 A2 A3 | = 0.

By the Inclusion-Exclusion Principle,

|A1 A2 A3 | = |U | |A1 | |A2 | |A3 |

+|A1 A2 | + |A1 A3 | + |A2 A3 |
|A1 A2 A3 |
= 45 3 9 + 3 1 0 = 21.

Of course, one can easily get the same answer by setting xi 1 = yi , 1 i 3.

Then the answer is the number of non-negative integer solutions of the equation
y1 + y2 + y3 = 5. That is,

3 3+51 7
= = = 21.
5 5 5

Theorem 4.15. Let U be a finite set. Let A1 , A2 , , An be subsets of U .

Then

|A1 A2 An | = |A1 | + |A2 | + |An |

h
|A1 A2 | + |A1 A3 | + + |A1 An |
+ |A2 A3 | + |A2 A4 | + + |A2 An |
i
+ + |An1 An | +
h i
+ |A1 A2 A3 | + + |An2 An1 An |

+ (1)n1 |A1 A2 An |
Xn X
= (1)k1 |Ai1 Aik | ;
k=1 1i1 <<ik n
 4.8. INCLUSION-EXCLUSION PRINCIPLE 55

or equivalently,
X X
|A1 A2 An | = |U | |Ai | + |Ai Aj |
i i<j
X
|Ai Aj Ak |
i<j<k
+
+ (1)n |A1 A2 An |
n
X X
(4.2) = |U | + (1)k |Ai1 Aik | .
k=1 1i1 <<ik n

Proof. For each element x U , we show that x contributes the same count to
both sides of (4.2).
Case I: x 6 A1 A2 An . Note that the element x is counted once in
A1 A2 An = A1 A2 An ,
once in U , and 0 times in all
Ai1 Ai2 Aik , i1 < i2 < < ik .
Thus x is counted once on both sides of (4.2).
Case II: x A1 A2 An . We assume that x belongs r to rexactly
r rsubsets

of
rA1 , A 2 , . . . , An , say A t1
, A t2
, . . . , Atr
. Then x is counted 0 , 1 , 2 , 3r , . . .,
r times in
X X X
U, |Ati |, |Ati Atj |, |Ati Atj Atk |, . . . , |At1 At2 Atr |
i i<j i<j<k

respectively. Consequently, the contribution of x on the right side of (4.2) is

r r r r r
+ + + (1)r = [1 + (1)]r = 0.
0 1 2 3 r
Therefore x is counted zero times on both sides of (4.2).

The subsets A1 , . . . , An of U may be given by conditions or properties c1 , . . . , cn

on the elements of U respectively. Let N be the number of elements of U , N
the number of elements of U satisfying none of the conditions c1 , . . . , cn . Let
N (ci1 ci2 cik ) be the number of elements of U satisfying the conditions ci1 , ci2 , , cik ,
1 i1 < i2 < < ik n. Then the Inclusion-Exclusion Principle can be stated
as
h i
N = N N (c1 ) + N (c2 ) + + N (cn )
h
+ N (c1 c2 ) + N (c1 c3 ) + + N (c1 cn )
+N (c2 c3 ) + N (c2 c4 ) + + N (c2 cn )
i
+ + N (cn1 cn )
h i
+ N (c1 c2 c3 ) + + N (cn2 cn1 cn )
(4.3) + + (1)n N (c1 c2 cn ).
56 4. COMBINATORICS

Let Nk be the number of elements of U satisfying at least k conditions, 0 k n,

that is,
X
Nk = |Ai1 Ai2 Aik |.
i1 <i2 <<ik

Then the Inclusion-Exclusion Formula can be further stated as

(4.4) |A1 A2 An | = N0 N1 + N2 N3 + + (1)n Nn .

Example Find the number of possible HK telephone numbers having no 14?

Proof. Let U be the set of HK phone numbers. Obviously, |U | = 108 . Let Ai be
the set of phone numbers such that the ith digit is 1 and the (i + 1)th digit is 4,
1 i 7. Then
X 7
7
N1 = |Ai | = 106 ,
i=1
1
X
6
N2 = |Ai Aj | = 104 ,
2
1i<j7
X
5
N3 = |Ai Aj Ak | = 102 ,
3
1i<j<k7
X
4
N4 = |Ai Aj Ak Al | = 100 .
4
1i<j<k<l7

Thus the answer is

8 7 6 6 4 5 2 4
10 10 + 10 10 + 100 = 93149001.
1 2 3 4

There is an algebraic proof for the Inclusion-Exclusion formula. For subset A

of U , the characteristic function of A is defined by

1 if x A
1A (x) =
0 if x 6 A.
For functions f : U R and g : U R and for any real number a, we define
functions
f +g :U R by (f + g)(x) = f (x) + g(x),
f g :U R by (f g)(x) = f (x) g(x),
fg : U R by f g(x) = f (x)g(x),
af : U R by (af )(x) = af (x).
Proposition 4.16. For subsets A and B of U ,
(a) 1AB = 1A 1B ,
(b) 1A = 1U 1A , where A = U A,
(c) 1U f = f for any function f : U R,
(d) 1AB = 1A + 1B 1AB .
 4.9. MORE EXAMPLES 57

For a function f : U R, the weight of f is the number

X
|f | = f (x).
xU

It is clear that
|af + bg| = a|f | + b|g|.

Note that A1 An is the set of elements of U satisfying none of the conditions
c1 , . . . , cn . The set Ai1 Aik consists of the elements of U satisfying the
conditions ci1 , . . . , cik . On the one hand by Proposition 4.16, we have
1A1 An = 1A 1An
= (1U 1A1 ) (1U 1A1 )
X
= f1 f2 fn (each fi is either 1U or 1Ai )
X
= 1U 1U + 1U 1U (1Ai1 ) (1Aik )
| {z } | {z }
1i1 <<ik n
n 1kn nk
n
X X
= 1U + (1)k 1Ai1 1Aik
k=1 1i1 <<ik n
Xn X
= 1U + (1)k 1Ai1 Aik .
k=1 1i1 <<ik n

On the other hand,

1A1 An = 1A1 An = 1U 1A An .
Then
n
X X
1A An = (1)k1 1Ai1 Aik .
k=1 1i1 <<ik n

Thus
|A1 An | = |1A1 An |
Xn X
= (1)k1 |1Ai1 Aik |
k=1 1i1 <<ik n
Xn X
= (1)k1 |Ai1 Aik |.
k=1 1i1 <<ik n

4.9. More Examples

Example Let A and B be finite sets, |A| = m and |B| = n. Let C(m, n) be the
number of surjective functions from A to B. What is C(m, n)?
Solution. Let U be the set of all functions from A to B, and let B = {b1 , . . . , bn }.
We call a function f : A B to satisfy condition ci if bi is not contained in
f (A). Then N0 (= nm ) is the number of functions from A to B; N is the number
of surjective functions from A to B; and N (ci1 cik ) is the number of functions
f from A to B such that {bi1 , . . . , bik } is not contained in f (A). Note that each
58 4. COMBINATORICS

function f : A B such that {b1 , . . . , bik } 6 f (A) can be identified as a function

from A to B\{bi1 , . . . , bik }. Thus

N (ci1 cik ) = (n k)m

and
n
Nk = (n k)m , 0 k n.
k
Therefore
n
X n
=
C(m, n) = N (1)k (n k)m .
k
k=0

When m < n, obviously C(m, n) = 0. We then have

n
X n
(1)k (n k)m = 0.
k
k=0

Note that C(m, n) can be interpreted as the number of ways to place the
elements of A into n distinct boxes so that no one box is empty. We have
X
m
C(m, n) = .
i ++in=m
i1 , . . . , in
1
i1 ,...,in 1

We obtain the identity

X n
X n
m
= (1)k (n k)m .
i1 , . . . , in k
i1 ++in=m k=0
i1 ,...,in 1

Theorem 4.17. Let (n) be the number of integers x in [1, n] that are coprime
with n. Then for any positive integer n = pe11 perr , where p1 , . . . , pr are distinct
primes and e1 , . . . , er are positive integers, the Euler function (n) is given by
r
Y
1
(n) = n 1 .
pk
k=1

Proof. Let U = {1, 2, . . . , n}. For the distinct primes p1 , . . . , pr , we define

conditions c1 , . . . , cr . An element x U is called to satisfy condition ci if pi divides
x, 1 i k. Let Ai be the set of elements of U satisfying the condition ci . Then

n
Ai = pi , 2pi , 3pi , . . . , pi , 1 i r.
pi

For 1 i1 < < ik r, let q = pi1 pik , then

n
Ai1 Aik = q, 2q, 3q . . . , q .
q
 4.10. GENERALIZED INCLUSION-EXCLUSION FORMULA 59

n
Thus |Ai1 Aik | = pi1 pik for all 1 i1 < < ik r. Therefore
r
X X
(n) = n+ (1)k |Ai1 Aik |
k=1 1i1 <<ik n
Xr X n
= n+ (1)k
p i 1 pi k
k=1 1i1 <<ik n

1 1
= n 1 + +
p1 pr

1 1 1 1
+ + + + + +
p1 p2 p1 p 3 p1 pr pr1 pr

1 1 1 1
+ + + + +
p1 p2 p3 p1 p2 p4 p1 p2 pr pr2 pr1 pr

1
+ + (1)r
p 1 p 2 pr
Y r
1
= n 1 .
pk
k=1

Example For n = 36 = 22 32 ,

1 1
(36) = 36 1 1 = 12.
2 3
In fact, the integers from 1 to 36 that are coprime with 36 can be listed as
1, 5, 7, 11, 13, 17, 19, 23, 25, 29, 31, 35.
Example A permutation of {1, 2, . . . , n} is called a dearrangement if every i
(1 i n) is not placed at the ith position. We call an arrangement of {1, 2, . . . , n}
to satisfy condition ci if i is placed at the ith position. Let dn be the number of
dearrangements of {1, 2, . . . , n}. Then
dn = N (
c1 c2 cn )
X n X
= n! + (1)k (n k)!
k=1 1i1 <<ik n
n
X n
= (1)k (n k)!
k
k=0
n
X 1
= n! (1)k ' n!e1 .
k!
k=1

4.10. Generalized Inclusion-Exclusion Formula

Theorem 4.18. Let U be a finite set and |U | = N . Let c1 , . . . , cn be some
properties about the elements of U . Let Ek denote the number of elements of U
satisfying exactly m properties of c1 , . . . , cn , then
(4.5)
m m+1 m+2 n
nm
Em = Nm Nm+1 + Nm+2 +(1) Nn .
0 1 2 nm
60 4. COMBINATORICS

Proof. For each x U we count the contribution of x on both sides of (4.5)

and divide the situation into three cases.
Case I: the element x satisfies fewer than m conditions. In this case the con-
tributions of x on both sides are 0.
Case II: the element x satisfies exactly m conditions, say, ci1 , , cim . In this
case the contribution of x on the left side is 1; and the contribution of x on the
right side is also 1 becasue x is counted once in Nm and 0 times in Nk for all k > m.
Case III: the element x satisfies exactly r (> m) conditions, say, ci1 , , cir .
Then the contribution of x on the leftside is 0. On the right side the contributions
r r

of x for Nm , Nm+1 , . . ., Nr are m , m+1 , . . ., rr respectively; and the contri-
butions of x for Nk with k > r are all 0. So the contribution of x on the right side
is
m r m + 1 r
r

r
+ + (1)rm .
0 m 1 m+1 rm r
Now it is easy to see that
rm
X rm
X
i m+i r (m + i)! r!
(1) = (1)i
i=0
i m + i i=0
i!m! (m + i)!(r m i)!
rm
X r! (r m)!
= (1)i
i=0
m!(r m)! i!(r m i)!
rm
X r r m
i
= (1)
i=0
m i
rX rm
rm
= (1)i
m i=0 i
r
= [(1) + 1]rm = 0.
m
Thus the contributions of x on both sides are the same.
Theorem 4.19. For any positive integer n,
X
(d) = n.
d|n

Proof. Let S be the set of pairs (d, k) of integers such that

d|n, 1 k d, gcd(k, d) = 1.
Then S is a subset of the grid U = {1, 2, . . . , n} {1, 2, . . . , n}. The characteristic
function of S can be viewed as an n by n matrix whose (d, k) entry is 1 if (d, k) S
and 0 if (d, k) 6 S. For each fixed d such that d|n, the number of 1s in the dth row
is (d). Thus
X
|S| = (d).
d|n

Now it suffices to show that |S| = n. To this end we construct a bijection

between S and {1, 2, . . . , n}. Let f : S {1, 2, . . . , n} be defined by
f (d, k) = kn/d.
Since 1 k d and d|n, we have f (d, k) = kn/d n; that is, f is well-defined.
 4.10. GENERALIZED INCLUSION-EXCLUSION FORMULA 61

For (d0 , k 0 ) 6= (d, k), if f (d, k) = f (d0 , k 0 ), that is, kn/d = k 0 n/d0 , then kd0 = k 0 d.
Since gcd(d, k) = 1 and gcd(d0 , k 0 ) = 1, we have d|d0 , d0 |d, k|k 0 , and k 0 |k. Then
d = d0 and k = k 0 , that is, f is injective.
For any 1 m n, let gm = gcd(m, n), dm = n/gm , km = m/gm . Then
dm |n, km dm , and gcd(dm , km ) = 1, that is, (dm , km ) S. Now we have
f (dm , km ) = km n/dm = m, that is, f is surjective.

Exercises
(1) In how many ways to arrange the letters E, I, M, O, T, U, Y so that YOU,
ME and IT would not occur?
(2) Six passengers on a msall airplane are randomly assigned to the six seats
on the plane. On the return trip they are again randomly assigned seats.
(a) What is the chance that every passenger has the same seats on both
trips?
(b) What is the probability that exactly five passengers have the same
seats ob both trips?
(c) What is the probability that at least one passenger has the seat on
both trips?
(3) Show that for any positive integer n,
X n
(n) = (d) ,
d
d|n

where is the Mobius function defined by

1 if d = 1,
(d) = (1)k if d is a product of k distinct primes,

0 if d has a repeated prime factor.
(4) There are n couples to be arranged in a straightline. Find the number of
ways to arrange them so that
(a) there is no couple to be next to each other.
(b) men and women alternate and there is no couple next to each other.
(5) There are n couples to be seated at a round table. Find the number of
seating plans so that
(a) there is no couple to be next to each other.
(b) men and women alternate and there is no couple next to each other.
(6) There are n persons seated in a circle. Find the numbers of ways to be
re-seated so that no k consecutive persons in the original seating order
appear.
 CHAPTER 5

Recurrence Relations

5.1. Infinite Sequences

Definition 5.1. An infinite sequence is a function from the set of positive
integers to the set of real or complex numbers.

Example The game of Hanoi Tower is to play with a set of disks of graduated
size with holes in their centers and a playing board having three spokes for holding
the disks; see Figure ??. The object of the game is to transfer all the disks from
spoke A to spoke C by moving one disk at a time without placing a larger disk on
top of a smaller one. What is the minimal number of moves required when there
are n disks?








































 
 
 
 
 
 
 
 
 
 
 
 
 
 
   
 




















































































 





A B C

Solution. Let an be the minimum number of moves to transfer n disks from one
spoke to another. Then {an | n 1} defines an infinite sequence. The first few
terms of the sequence {an } can be listed as

1, 3, 7, 15, . . .

We are interested in finding a closed formula to compute an for arbitrary n.

In order to to move n disks from spoke A to spoke C, one must move the first
n 1 disks from spoke A to spoke B by an1 moves, then move the last (also the
largest) disk from spoke A to spoke C by one move, and then remove the n 1
disks again from spoke B to spoke C by an1 moves. Thus the total number of
moves should be

an = an1 + 1 + an1 = 2an1 + 1.

This means that the sequence {an | n 1} satisfies the recurrence relation

an = 2an1 + 1, n 1
(5.1)
a1 = 1.

Given a recurrence relation for a sequence with initial conditions, solving the
recurrence relation means to find a formula to express the general term an of the
sequence.
63
64 5. RECURRENCE RELATIONS

For the sequence {an | n 0} defined by the recurrence relation (5.1), if we

apply the recurrence relation again and again, we have
a1 = 2a0 + 1
a2 = 2a1 + 1 = 2(2a0 + 1) + 1 = 22 a0 + 2 + 1
a3 = 2a2 + 1 = 2(22 a0 + 2 + 1) = 23 a0 + 22 + 2 + 1
a4 = 2a3 + 1 = 2(23 a0 + 22 + 2 + 1) = 24 a0 + 23 + 22 + 2 + 1
..
.
an = 2n a0 + 2n1 + 2n2 + + 2 + 1 = 2n a0 + 2n 1.
Let a0 = 0. The general term is given by
an = 2n 1, n 1.

5.2. Homogeneous Recurrence Relations

Any recurrence relation of the form
(5.2) xn = axn1 + bxn2
is called a second order homogeneous linear recurrence relation.
Let xn = sn and xn = tn be two solutions, i.e.,
sn = asn1 + bsn2 and tn = atn1 + btn2 .
Then for constants c1 and c2
c1 sn + c2 tn = c1 (asn1 + bsn2 ) + c2 (atn1 + btn2 )
= a(c1 sn1 + c2 tn1 ) + b(c1 sn2 + c2 tn2 ).
This means that xn = c1 sn + c2 tn is a solution of (5.2).
Theorem 5.2. Any linear combination of solutions of a homogeneous recur-
rence linear relation is also a solution.
In solving the first order homogeneous recurrence linear relation xn = axn1 ,
it is clear that the general solution is xn = an a0 . This means that xn = an is a
solution. This suggests that, for the second order homogeneous recurrence linear
relation (5.2), we may have the solutions of the form
xn = rn .
Indeed, put xn = rn into (5.2); we have
rn = arn1 + brn2 or rn2 (r2 ar b) = 0.
Thus either r = 0 or
(5.3) r2 ar b = 0.
The equation (5.3) is called the characteristic equation of (5.2).
Theorem 5.3. If the characteristic equation (5.3) has two distinct roots r1 and
r2 , then the general solution for (5.2) is given by
xn = c1 r1n + c2 r2n .
If the characteristic equation (5.3) has only one root r, then the general solution
for (5.2) is given by
xn = c1 rn + c2 nrn .
 5.2. HOMOGENEOUS RECURRENCE RELATIONS 65

Proof. When the characteristic equation (5.3) has two distinct roots r1 and r2
it is clear that both xn = r1n and xn = r2n are solutions of (5.2).
Now assume that (5.2) has only one root r. Then a2 + 4b = 0 and r = a/2, i.e.,
a2 a
b= and r = .
4 2
n
We verify that xn = nr is a solution of (5.2). In fact,
a n1 a2 a n2 a n
axn1 + bxn2 = a(n 1) + (n 2) =n = xn .
2 4 2 2

Remark There is heuristic method to explain why xn = nrn is a solution when
the two roots are the same. If two roots r1 and r2 are distinct but very close to
each other, then r1n r2n is a soltuion. So is (r1n r2n )/(r1 r2 ). It follows that the
limit
rn r2n
lim 1 = nr1n1
r2 r1 r1 r2

would be a solution. Thus its multiple xn = r1 (nr1n1 ) = nr1n by the constant r1

is also a solution. Please note that this is not a mathematical proof, but a mathe-
matical idea.

Example Find a general formula for the Fibonacci sequence

fn = fn1 + fn2
f0 = 0

f1 = 1

Solution. The characteristic equation r2 = r + 1 has two distinct roots

1+ 5 1 5
r1 = and r2 = .
2 2
Then the general solutions is
!n !n
1+ 5 1 5
fn = c1 + c2 .
2 2
Set (
0 = c1
+ c2

1+ 5 1 5
1 = c1 2 + c2 2 .
We have c1 = c2 = 1 .Thus
5
!n !n
1 1+ 5 1 1 5
fn = , n 0.
5 2 5 2
Remark The Fibonacci sequence fn is an integer sequence, but it looks like a
sequence of irrational numbers from its general formula above.

Example Find the solution for the recurrence relation

xn = 6xn1 9xn2
x0 = 2

x1 = 3
66 5. RECURRENCE RELATIONS

Solution. The characteristic equation

r2 6r + 9 = 0 (r 3)2 = 0
has only one root r = 3. Then the general solution is
xn = c1 3n + c2 n3n .
The initial conditions x0 = 2 and x1 = 3 imply that c1 = 2 and c2 = 1. Thus the
solution is
xn = 2 3n n 3n = (2 n)3n , n 0.

Example Find the solution for the recurrence relation

xn = 2xn1 5xn2 , n 2
x0 = 1

x1 = 5

Solution. The characteristic equation

r2 2r + 5 = 0 (x 1 2i)(x 1 + 2i) = 0
has two distinct complex roots r1 = 1 + 2i and r2 = 1 2i. The initial conditions
imply that
c1 + c2 = 1 c1 (1 + 2i) + c2 (1 2i) = 5.
12i 1+2i
So c1 = 2 and c2 = 2 . Thus the solutions is
1 2i 1 + 2i
xn = (1 + 2i)n + (1 2i)n
2 2
5 5
= (1 + 2i)n+1 + (1 2i)n+1 , n 0.
2 2
Remark The sequence is obviously a real sequence. However, its general formula
involves complex numbers.

Example Two persons A and B gamble dollars on the toss of a fair coin; A has $70
and B has $30. In each play either A wins $1 from B or loss $1 to B; and the game
is played without stop until one wins all the money of the other or goes forever.
Find the following probabilities.
(a) A wins all the money of B.
(b) A loss all his moeny to B.
(c) The game continues forever.
Solution. Either A or B can keep track of the game simply by counting their own
money; their position (money) can be one of the numbers 0, 1, 2, . . . , 100. Let pn
be the probability that A reaches 100 at position n. After one toss, A enters into
either position n + 1 or position n 1. The new probability that A reaches 100 is
either pn+1 or pn1 . Since the probability of A moving to position n + 1 or n 1
from n is 12 . We thus have the recurrence relation

pn = 21 pn+1 + 12 pn1
p0 = 0

p100 = 1
 5.3. HIGHER ORDER HOMOGENEOUS RECURRENCE RELATIONS 67

The characteristic equation is r2 2r + 1 = 0; it has only one root r = 1. The

general solutions is
pn = c1 + c2 n.
1
Apply the boundary conditions p0 = 0 and p100 = 1; we have c1 = 0 and c2 = 100 .
Thus
n
pn = , 0 n 100.
100
n
Of course, pn = 100 for n > 100 is nonsense to the original problem. The probabil-
ities for (a), (b), and (c) are 70%, 30%, and 0, respectively.
The recurrence relation pn = 12 pn+1 + 12 pn1 can be directly solved. In fact,
the recurrence relation can be simplied to
pn+1 pn = pn pn1 .
Then pn+1 pn = pn pn1 = = p1 p0 . Since p0 = 0, we have pn = pn1 + p1 .
Apply the recurrence relation again and again, we obtain
pn = p0 + np1 .
n
Applying the conditions p0 = 0 and p100 = 1, we have pn = 100 .

5.3. Higher Order Homogeneous Recurrence Relations

For the higher order homogeneous recurrence relation
(5.4) xn+k = a1 xn+k1 + a2 xn+k2 + + ank xn , n0
the characteristic equation is
(5.5) rk = a1 rk1 + a2 rk1 + + ank+1 r + ank
or
rk a1 rk1 a2 rk1 ank+1 r ank = 0.
Theorem 5.4. For the recurrence relation (5.4), if its characteristic equation
(5.5) has distinct roots r1 , r2 , . . . , rk , then the general solution for (5.4) is
xn = c1 r1n + c2 r2n + + ck rkn
where c1 , c2 , . . . , ck are arbitrary constants. If the characteristic equation has re-
peated roots r1 , r2 , . . . , rs of multiplicities m1 , m2 , . . . , ms respectively, then the gen-
eral solution of (5.4) is a linear combination of the solutions
r1n , nr1n , . . . , nm1 1 r1n ;
r2n , nr2n , . . . , nm2 1 r2n ;
...;
rsn , nrsn , . . . , nms 1 rsn .
Example Find an explicit formula for the sequence given by the recurrence relation

xn = 15xn2 10xn3 60xn4 + 72xn5
x0 = 1, x1 = 6, x2 = 9, x3 = 110, x4 = 45
Solution. The characteristc equation r5 = 15r3 10r2 60r + 72 can be simpliefied
as
(r 2)3 (r + 3)2 = 0.
So there are two roots r1 = 2 with multiplicity 3 and r2 = 3 with multiplicity 2.
Then the general solution is given by
xn = c1 2n + c2 n2n + c3 n2 2n + c4 (3)n + c5 n(3)n .
68 5. RECURRENCE RELATIONS

The initial condition means that

c1
+c4 = 1


2c1 +2c2 +2c3 3c4 3c5 = 1
4c1 +8c2 +16c3 +9c4 +18c5 = 1



8c1 +24c2 +72c3 27c4 81c5 = 1

16c1 +64c2 +256c3 +81c4 +324c5 = 1
Solving the linear system we have
c1 = 2, c2 = 3, c3 = 2, c4 = 1, c5 = 1.

5.4. Non-homogeneous Equations

A recurrence relation of the form
(5.6) xn = axn1 + bxn2 + f (n)
(s)
is called a non-homogeneous recurrence relation. Let xn be a solution of
(5.6), called a special solution, then the general solution for (5.6) is
(5.7) xn = x(s) (h)
n + xn ,
(h)
where xn is the general solution for the corresponding homogeneous equation
(5.8) xn = axn1 + bxn2 .
Theorem 5.5. Let f (n) = crn for the non-homogeneous recurrence relation
(5.6). Let r1 and r2 be the roots of the characteristic equation of the corresponding
homogeneous equation (5.8).
(s) (s)
(a) If r 6= r1 , r 6= r2 , then xn can be of the form xn = Arn .
(s) (s)
(b) If r = r1 , r1 6= r2 , then xn can be of the form xn = Anrn .
(s) (s)
(c) If r = r1 = r2 , then xn can be of the form xn = An2 rn .
where A is a constant to be determined in all cases.
Proof. (a) Put xn = Arn into the recurrence relation (5.6), we have
Arn = aArn1 + bArn2 + crn .
We may assume r 6= 0; otherwise the recurrence relation is homogeneous. Then
A(r2 ar b) = cr2 .
Note that r2 ar b 6= 0 because r is not a root of the characteristc equation.
Thus when
cr2
A= 2
r ar b
the sequence xn = Arn is a special solution.
(b) Since r = r1 6= r2 , it is clear that xn = nrn is not a solution for its
corresponding homogeneous equation, i.e.,
nr2 a(n 1)r b(n 2) = n(r ar b) + ar + 2b = ar + 2b 6= 0.
Put xn = Anrn into (5.6); we have
Anrn = aA(n 1)rn1 + bA(n 2)rn2 + crn ,
If r 6= 0, we then have A(nr2 a(n 1)r b(n 2)) = cr2 . So
cr2
A=
ar + 2b
 5.5. DIVIDE-AND-CONQUER METHOD 69

the sequence xn = Anrn is a special solution.

(c) Since r = r1 = r2 , it follows that xn = n2 rn is not a solution of the
corresponding homogeneous equation, i.e.,
n2 r2 a(n 1)2 r b(n 2)2 = n2 (r2 ar b) + 2n(ar + 2b) a 4b = a 4b 6= 0.
Put xn = An2 rn into (5.6); we have

Arn2 n2 r2 a(n 1)2 r b(n 2)2 = crn .
If r 6= 0, we then have
cr2
A=
a + 4b
and the sequence xn = An2 rn is a non-homogeneous solution.

Example Consider the non-homogeneous equation

xn = 3xn1 + 10xn2 + 7 5n
x0 = 4

x1 = 3

Solution. The characteristic equation r2 3r10 = 0 has roots r1 = 5 and r2 = 2.

A special solution can be given by
cr2
xn = n5n = n5n+1
3a + 2b
and the general solution is
xn = n5n+1 + c1 5n + c2 (2)n .
The initial condition implies c1 = 2 and c2 = 6. Thus
xn = n5n+1 2 5n + 6(2)n .

5.5. Divide-and-Conquer Method

Suppose we have a job of size n to be done. If the size n is large and the job
is complicated, we may divide the job into smaller jobs of the same type and of
the same size, then conquer the smaller problems and use the results to construct
a solution for the original problem of size n. This is the nature of the so-called
Divide-and-Conquer method.

Example Suppose there are n (= 2k ) student files, indexed by the student id

numbers as
A = {a1 , a2 , . . . , an }.
Given a particular file a A, what is the number of comparisons needed in worst
case to find the position of the file a?
Solution. Let xn be the number of comparisons needed to find the position of the
file a in worst case. Then the answer depends on whether or not the files are sorted.
Case I: The files in A are not sorted. Then the answer is at most n comparisons.
Case II: The files in A are sorted in the order a1 < a2 < < an .
a1 a2 a n2 1 a n2 a n2 +1 an1 an
We may compare the file a with a n2 . If a = a n2 , the job is done by one comparison.
If a < a n2 , consider the subset {a1 , a2 , . . . , a n2 }. If a > a n2 , consider the subset
70 5. RECURRENCE RELATIONS

{a n2 +1 , a n2 +2 , . . . , an }. Then the number of comparisons is at most x n2 + 1. We

thus obtain a recurrence relation

xn = x n2 + 1
x1 = 1
Applying the recurrence relation again and again, we obtain
xn = x n2 + 1 = x 2n2 + 2 = x 2n3 + 3 = = x nk + k = x1 + k = k + 1.
2

k
Since n = 2 , we have k = log2 n. Therefore
xn = log2 n + 1.

Example Let S = {a1 , a2 , . . . , an } Z, where n = 2k and k 1. How many

number of comparisons are needed in worst case to find the minimum in S? We
assume that the numbers in S are not sorted.
Solution. The number of comparisons depends on the method we employed. If all
possible pairs of elements in S are compared, then theminimum
element will be
found, and the number of comparisons in worst case is n2 = n(n1)
2 = O(n2 ). Of
course this is not best possible.
There is another method to find a better solution. Let xn be the minimal
number of comparisons needed in worst case to find the minimum in S. Obviously,
xn = 1. For n = 2k and k 1, we may divide S into two subsets
n
S1 = {a1 , a2 , . . . , a n2 }, |S1 | = 2,
n
S2 = {a n2 +1 , a n2 +2 , . . . , an }, |S2 | = 2.
It takes x n2 comparisons to find the minimum m1 for S1 and the minimum m2 for
S2 . Then compare m1 with m2 to determine the minimum element in S. In this
way the total number of comparisons for S in worst case is 2x n2 + 1. We thus obtain
a recurrence relation
xn = 2x n2 + 1
x2 = 1
Apply the recurrence relation again and again; we have

xn = 2 2x 2n2 + 1 = 22 x 2n2 + 2 + 1

= 22 2x 2n3 + 1 + 2 + 1 = 23 x 2n3 + 22 + 2 + 1
= = 2k x nk + 2k1 + + 2 + 1
2

2k 1
= 2k1 + + 2 + 1 =
21
= n 1 = O(n).

We hope that we understand the nature of divide-and-conquer method by the
above examples. In order to solve a problem of size n, if the size n is large and the
problem is complicated, we divide the problem into a smaller subproblems of the
same type and of the same size d nb e, where a, b Z+ , 1 a < n and 1 < b < n.
Then we solve the a smaller subproblems and use the results to construct a solution
for the original problem of size n. We are especially interested in the case where
n = bk and b = 2.
 5.5. DIVIDE-AND-CONQUER METHOD 71

Theorem 5.6. (Divide-and-Conquer Algorithm)

(a) The time to solve the initial problem of size n = 1 is a constant c 0.
(b) The time to break the given problem of size n into a smaller same type
subproblems, together with the time to construct a solution for the original
problem by using the solutions for the a subproblems, is a function h(n).

Our concern here is to figure out the time complexity function f (n) which is
given by the recurrence relation

f (1) = c
f (n) = af ( nb ) + h(n), n = bk , k 1
Theorem 5.7. Let f : Z+ R be a function defined by the recurrence relation

f (1) = c
f (n) = af ( nb ) + c, n = bk , k 1
where a, b, c are positive integers and b 2. Then
(
f (n) = c(logb n + 1) if a = 1
logb a
f (n) = c(an a1 1) if a =
6 1
Proof. Applying the recurrence relation, we obtain
f (n) = af ( nb ) + c
af ( nb ) = a2 f ( bn2 ) + ac
a2 f ( bn2 ) = a3 f ( bn3 ) + a2 c
..
.
n n
ak2 f ( bk2 ) = ak1 f ( bk1 ) + ak2 c
k1 n n
a f ( bk1 ) = a f ( bk ) + ak1 c
k

Adding both sides of the k equations and canceling the common summands, we
have n
f (n) = ak f k + c + ac + a2 c + + ak1 c .
b
Since n = bk and f (1) = c, we further have

f (n) = c 1 + a + a2 + + ak .
If a = 1, then f (n) = c(k + 1). Note that n = bk implies k = logb n. Hence
f (n) = c (logb n + 1) .
k+1
c(a 1)
If a 6= 1, then f (n) = a1 .
Since k = logb n, it follows that
logb n log n logb a
ak = alogb n = blogb a = b b = nlogb a .
Therefore
c anlogb a 1
f (n) = .
a1

Exercises
(1) Find an explicit formua for each of the sequences defined by the recurrence
relations with initial conditions.
(a) xn = 5xn1 + 3, x1 = 3.
72 5. RECURRENCE RELATIONS

(b) xn = 3xn1 + 5n, x1 = 5.

(c) xn = 2xn1 + 15xn2 , x1 = 2, x2 = 4.
(d) xn = 4xn1 + 5xn2 , x1 = 3, x2 = 5.
(e) xn = 3xn1 2xn2 , x0 = 2, x1 = 4.
(f) xn = 6xn1 9xn2 , x0 = 3, x1 = 9.
(2) Show that if sn and tn are solutions for the non-homogeneous linear re-
currence relation
xn = axn1 + bxn2 + f (n), n 2,
then xn = sn tn is a solution for the homogeneous linear recurrence
relation
xn = axn1 + bxn2 , n 2.
(3) Let the characteristic equation for the homogeneous linear recurrence re-
lation
xn = axn1 + bxn2 , n 2
have two distinct roots r1 and r2 . Show that every solution can be written
in the form
xn = c1 r1n + c2 r2n
for some constants c1 and c2 .
(4) Let A1 , A2 , . . . , An+1 be kk matrices. Let Cn be the number of ways to
evaluate the product A1 A2 An+1 by choosing different orders in which
to do the n multiplications.
(a) Find a recurrence relation with an initial condition for the sequence
Cn . 2n
1
(b) Verify that the sequence n+1 satisfies your recurrence relation
1
n2n
and conclude that Cn = n+1 n . (The numbers Cn are called
Catalan numbers.)
(5) Find a general formula for the recurrence relation
xn = axn1 + b + cn
in terms of x0 , where a, b, c are real constants.
(6) Find an explicit formua for each of the sequences defined by the recurrence
relations with initial conditions.
(a) xn = 5x n3 + 5, x1 = 5, n = 2k , k 0.
(b) xn = xb n2 c + 3, x1 = 4, n 1.
(c) x2n = 2xn + 5 7n, x1 = 0.
(7) Let f (n) be a real sequence defined for n = 1, b, b2 , . . ., and satisfy the
recurrence relation
n
f (n) = af + h(n),
b
where b 2 is an integer. Show that
1+logb n1
X n
f (n) = f (1)nlogb a + ai h .
i=0
bi
(8) Let f (n) be a real sequence defined for n = 1, b, b2 , b3 , . . ., and satisfy the
recurrence relation
n
f (n) = af + a0 + a1 n + + ak nk ,
2
 5.5. DIVIDE-AND-CONQUER METHOD 73

where a, b, a0 , a1 , . . . , ak are real constatnts, a > 0 and b > 1. Show that

(a) If a = bi for some 0 i k, then
k
X bj aj j
f (n) = f (1)ni + ai ni logb n + n ni .
bj bi
j=1,j6=i
i
(b) If a 6= b for all 0 i k, then
Xk
bj aj j
f (n) = f (1)nlogb a + j
n nlogb a .
j=0
b a

Solution: For the recurrence relation

xn = axn1 + b + cn
we have
xn = axn1 + b + cn
= a(axn2 + b + c(n 1)) + b + cn = a2 xn2 + b(a + 1) + c(a(n 1) + n)
= a2 (axn3 + b + c(n 2)) + b(a + 1) + c(a(n 1) + n)
= a3 xn3 + b(a2 + a + 1) + c(a2 (n 2) + a(n 1) + n)
..
.
= an x0 + b(an1 + + a + 1) + c(an1 + 2an2 + + (n 1)a + n)
= an x0 + b(an1 + + a + 1) + c(an1 + + a + 1)
+c(an2 + + a + 1) + + c(a2 + a + 1) + c(a + 1) + c
If a = 1, then
cn(n + 1)
xn = x0 + bn + .
2
If a 6= 1, then
n
n b(an 1) a 1 an1 1 a2 1 a 1
xn = a x0 + +c + + + +
a1 a1 a1 a1 a1
n
b(a 1) c n
= an x0 + + a + an1 + a n
a1 a1
b(an 1) c(an+1 1) c(n 1)
= an x0 + + .
a1 (a 1)2 a1
Catland numbers We have C0 = 1, C1 = 1, C2 = 2. The product A1 A2 An+1
can be obtained by the multiplication of two matrices in n ways, i.e.,
A1 A2 An+1 = (A1 Ak )(Ak+1 An+1 ), 1 k n.
This yields the recurrence relation
n
X X
Cn = Ck1 Cnk = Ci Cj .
k=1 i+j=n1

Consider the generating function

X
F (x) = Cn xn .
n=0
74 5. RECURRENCE RELATIONS

Note that


X X
X
F (x) 1
F (x)2 = Ci Cj xn = Cn+1 xn = .
n=0 i+j=n n=0
x x

Then
xF (x)2 F (x) + 1 = 0.
We thus have
1 1 4x
F (x) = .
2x
Since
1
X

1 4x = 2 (4x)n
n=0
n

X 1
( 12 1) ( 12 n + 1) 2n
= 1+ 2
2 (1)n xn
n=1
n!
X
(1)(3)(5) (2(n 1) + 1) n
= 2 (1)n xn
n=0
n!
X
1 3 5 (2(n 1) 1) n n
= 2 x
n=0
n!
X
(2(n 1))! n
= 12 x
n=1
n!(n 1)!

X (2n)!
= 12 xn+1 ,
n=0
n!(n + 1)!
we conclude that

1 1 4x X (2n)! n
X 1 2n
F (x) = = x = xn .
2x n=0
n!(n + 1)! n=0
n + 1 n
1
2n
Therefore Cn = n+1 n .

Eulers Problem In how many different ways can a labeled convex n-gon be
divided into triangles by non-intersecting diagonals?
Solution. Let cn be number of ways for an (n + 2)-gon in the problem. Then c1 = 1,
c2 = 2, and c3 = 5. Consider a convex (n + 3)-gon V1 V2 Vn+3 .
In each decomposition of the (n + 3)-gon, the segment V1 Vn+3 is a side of
some triangle in the decomposition; and the third vertex of such a triangle is one
of the vertices V2 , V3 , . . . , Vn+2 . Let the thrid vertex be Vk+2 , 0 k n. Then
we have one (k + 2)-gon V1 V2 Vk+2 and one (n k + 2)-gon Vk+2 Vk+3 Vn+3 .
There are ck ways to divide V1 V2 Vk+2 into triangles and cnk ways to divide
Vk+2 Vk+3 Vn+3 into triangles. We thus have the recurrence relation
n
X
cn+1 = ck cnk ,
k=0

(2n)! ( 2n
n )
where c0 = 1. So cn = n!(n+1)! = n+1 .
 5.7. GROWTH OF FUNCTIONS 75

5.6. Searching and Sorting

Given an array A = ha1 , a2 , , an i and an object S, determine the position
of S in A, that is, find an index i such that ai = S (if such an i exists).
Algorithm SEQSEARCH
Step 1 Input A and S.
Step 2 For i = 1 to n do
Step 3 If ai = S, then output i and stop.
Step 4 Output S not in A and stop.

5.7. Growth of Functions

For functions f and g defined on the set Z+ of positive integers, if there exist
positive constant C and K such that
|f (n)| C|g(n)| for all n K,
then f is said to be big-Oh of g, write f is O(g). This means that f grows no
faster than g. We say that f and g have the same order if f is O(g) and g is
O(f ). If f is O(g), but g is not O(f ), then we say that f is of lower order than
g or g grows faster than f .
Example 5.1. For Example 6 in Lectures 12 and 13, the number of comparisons
f (n) is a function of integers, and in Case I, f is O(n); in Case II, f is O(log n).
For Example 7, the number of comparisons f (n) is also function of positive
integers, and for Solution I, f is O(n2 ), but for Solution II, f is O(n).
For two functions f and g defined on positive integers, f is O(g) if and only if
there exists a constant C such that
f (n)
lim C.
n g(n)
 CHAPTER 6

Binary Relations

6.1. Binary Relations

The notion of a relation between two sets of objects is quite common and
intuitively clear. Let X be the set of all living human females and Y the set of
all living human males. The wife-husband relation R can be defined from X to Y .
Thus, for x X and y Y , we say that x is related to y by the relation R if x is
a wife of y, and write xRy. To describe the relation R, we may take the collection
of all ordered pairs (x, y) such that x is related to y by R; the collection of related
ordered pairs is simply a subset of the product set X Y . This motivates the
definition of relations.
Definition 6.1. Let X and Y be nonempty sets. By a binary relation (or
just relation) from X to Y we mean a subset R X Y . If (x, y) R, we say
that x is related to y by R, written xRy. If (x, y) 6 R, we say that x is not related
For x X, we define
to y, and written xRy.
R(x) = {y Y | xRy} = {y Y | (x, y) R};
and for a subset A X, define
R(A) = {y Y | there exists x A such that xRy}.
If X = Y , we say that R is a binary relation on X.
Since binary relations from X to Y are subsets of X Y , one can define
intersection, union, and complement for binary relations. For a relation R X Y ,
the complementary relation of R is the binary relation R X Y , defined by
(x, y) 6 R;
xRy
and the inverse relation of R is the binary relation R1 Y X, defined by
yR1 x (x, y) R.

Example Consider a family A with five children, Amy, Bob, Charlie, Debbie, and
Eric. We abbreviate the names to their first letters so that A = {a, b, c, d, e}.
(a) The brother-sister relation Rbs is the set
Rbs = {(b, a), (b, d), (c, a), (c, d), (e, a), (e, d)}.
(b) The sister-brother relation Rsb is the set
Rsb = {(a, b), (a, c), (a, e), (d, b), (d, c), (d, e)}.
(c) The brother relation Rb is the set
{(b, b), (b, c), (b, e), (c, b), (c, c), (c, e), (e, b), (e, c), (e, e)}.

77
78 6. BINARY RELATIONS

(d) The sister relation Rs is the set

{(a, a), (a, d), (d, a), (d, d)}.

The brother-sister relation Rbs is the inverse of the sister-brother relation Rsb ;
1
that is, Rbs = Rsb . The brother or sister relation is the union of the brother
relation and the sister relation; that is, Rb Rs . The complementary relation of
brother or sister relation is the brother-sister or sister-brother relation; that is,
Rb Rs = Rbs Rsb .

Example
2 2
(a) The graph of the equation, x32 + y22 = 1, defines a binary relation on the
set R of real numbers; its graph is an ellipse.
(b) The relation less than, denoted <, is a binary relation on R, defined
by a < b if and only if a is less than b. As a subset of R2 = R R, the
relation is also given by the set {(a, b) R2 | a is less than b}.
(c) The relation greater than or equal to is a binary relation on R, defined
by a b if and only if a is greater than or equal to b. As a subset of R2 , the
relation is given by the set {(a, b) R2 | a is greater than or equal to b}.
(d) The divisibility relation | about integers, defined by a|b if and only if a
divides b, is a binary relation on the set Z of integers.

Example A function f : X Y can be viewed as a relation from X to Y ; it is a

relation f X Y such that |f (x)| = 1 for all x X.

Proposition 6.2. Let R be a binary relation from X to Y . Let A and B be

subsets of X.
(a) If A B, then R(A) R(B).
(b) R(A B) = R(A) R(B).
(c) R(A B) R(A) R(B).

Proof. (a) For any y R(A), there is an x A such that xRy. Since A B,
we have y R(B). Thus R(A) R(B).
(b) For any y R(A B), there is an x A B such that xRy. If x A,
then y R(A). If x B, then y R(B). In either case, y R(A) R(B).
Thus R(A B) R(A) R(B). On the other hand, it follows from (a) that
R(A) R(A B) and R(B) R(A B). Therefore R(A) R(B) R(A B).
(c) It follows from (1) that R(A B) R(A) and R(A B) R(B). Thus
R(A B) R(A) R(B).

Proposition 6.3. Let R1 and R2 be relations from X to Y . If R1 (x) = R2 (x)

for all x X, then R1 = R2 .

Proof. If xR1 y, then y R1 (x). Since R1 (x) = R2 (x), we have y R2 (x).

Then xR2 y. A similar argument shows that if xR2 y then xR1 y. Thus R1 = R2 .

Let R be a relation on a set X. A path of length k in R from x to y is a

finite sequence x = x0 , x1 , . . ., xk = y, beginning with x and ending with y, such
that
x0 Rx1 , x1 Rx2 , . . . , xk1 Rxk .
 6.2. REPRESENTATION OF RELATIONS 79

A path that begins and ends at the same vertex is called a cycle. For a fixed
positive integer k, we define a relation Rk on X as follows:
xRk y there is a path of length k from x to y.
We may also define a relation R on X by letting
xR y there is some path from x to y.
The relation R is sometimes called the connectivity relation for R. It is clear
that

[
R = R R2 R3 = Rk .
k=1
The reachability relation of R is the relation R on X defined by
xR y either x = y or xR y;
that is,

[
R = I R R2 R3 = Rk ,
k=0
where I is the identity relation on X, defined by xIy if and only if x = y. We
always assume that R0 = I for any relation R.

6.2. Representation of Relations

Binary relations are the most important relations among all relations. Ternary
relations, quaternary relations, and multi-relations can be studied by binary rela-
tions. We introduce two methods to represent a binary relation, one by a matrix
and the other one by a directed graph.
Definition 6.4. Let R be a binary relation from X = {x1 , x2 , . . . , xm } to
Y = {y1 , y2 , . . . , yn }. The matrix of R is an m n matrix MR = [aij ] whose
(i, j)-entry is
1 if (xi , yj ) R
aij =
0 if (xi , yj ) 6 R.
The matrix MR is called a Boolean matrix. If X = Y , MR is a square matrix.
For m n Boolean matrices M1 = [aij ] and M2 = [bij ], if aij bij for all
(i, j)-entries, we write M1 M2 . The matrix of the brother-sister relation Rbs
on the set A = {a, b, c, d, e} is the square matrix

0 0 0 0 0
1 0 0 1 0

1 0 0 1 0

0 0 0 0 0
1 0 0 1 0
and the matrix of the brother or sister relation is the square matrix

1 0 0 1 0
0 1 1 0 1

0 1 1 0 1

1 0 0 1 0
0 1 1 0 1
80 6. BINARY RELATIONS

Another way to describe a binary relation is to draw a directed graph. Let R

be a binary relation on a finite set V = {v1 , v2 , . . . , vn }. For each element vi V ,
we draw a solid dot and name it by vi , called a vertex. For two vertices vi and vj ,
if vi Rvj , we draw an arrow from vi to vj , called a directed edge; if vi = vj , the
arrow becomes a directed loop. The resulted graph is a directed graph, called the
digraph of R, denoted D(R). Note that the directed edges of a digraph may have
to cross each other when drawing the digraph on a plane. However, the intersection
points of directed edges are not considered to be vertices of the digraph. The in-
degree of a vertex v V is the number of vertices u such that (u, v) R, denoted
ideg (v); the out-degree of v is the number of vertices w such that (v, w) R,
denoted odeg (v). If R is a relation from X to Y , we define
odeg (x) = |R(x)| for all x X,
ideg (y) = |R1 (y)| for all y Y .
The digraphs of the brother-sister relation Rbs and the brother or sister
relation Rb Rs are demonstrated in the following.

b

 

  
  
  
  
  
  
  
  
   b
        
 
           
 
 




 
  
  

 
  
  

 
  
  

 
  
  

 
  
  

 
  
  

 
  
  

 
  
  




 


 


 


  
 



 
  
  

  
  

  
  

  
  

  
  

  
  

  
  

  
  

  
  

    
  

  
  

  
  

 
 

 c
                       a


 
  
  
  
 
 
 
 
 
  
  
  
 
     c
 


        
  
  
 
 

a  
 
  
  
  
   

  
  


 

 

  

  
  
 












     
  



  
  


  
  


  
  


  
  




 

 
  

 
 

 


 
 
 
 
 
 
 
 
 
 
 
   
 
 
 
 
 

  

  
  
  
  
  
    
   
 
  
  
    
 

 

 

 

 
  
 
    
 



     
 





 
 
 
 
 
 
 
   
  
 





     
  
  

  

   
  
  
 e d


   




e d
(a) Rbs (b) Rb Rs

Figure 1. The digraphs of relations Rbs and Rb Rs .

Proposition 6.5. For the digraph D(R) of a binary relation R on V ,

X X
ideg (v) = odeg (v) = |R|.
vV vV

If R is a relation from X to Y , then

X X
odeg (x) = ideg (y) = |R|.
xX yY

Proof. Trivial.

6.3. Composition of Relations

Definition 6.6. Let R be a relation from X to Y , and S a relation from Y to
Z. The composition of R and S is a relation S R from X to Z, defined by
x(S R)z there is an element y Y such that xRy and ySz.
If X = Y , then R is a relation on X; we have R2 = R R and Rk = Rk1 R for
k 2.
 6.3. COMPOSITION OF RELATIONS 81

Note that in the composition of R and S we consider R as the first relation

and S the second, and the notation S R is backward. However, many people
use R S as a name for what we have called S R. Such usage is inconsistent
with the notation for functional composition and causes some confusion. To avoid
misunderstanding and for aesthetic reason, we often write RS for S R.

Example For the brother-sister relation, sister-brother relation, brother re-

lation, and sister relation on A = {a, b, c, d, e}, we have
Rbs Rsb = Rb , Rsb Rbs = Rs , Rbs Rs = Rbs ,
Rbs Rbs = , R b Rb = R b , Rb Rs = .

Let X1 , X2 , . . ., Xn , Xn+1 be nonempty sets. Let Ri be a relation from Xi to

Xi+1 , 1 i n. We define a relation R1 R2 Rn from X1 to Xn+1 by
xR1 R2 Rn y
if and only if there is a sequence x = x1 , x2 , . . ., xn , xn+1 = y such that
x1 R1 x2 , x2 R2 x3 , . . . , xn Rn xn+1 .
Theorem 6.7. Let R1 be a relation from X1 to X2 , R2 a relation from X2 to
X3 , and R3 a relation from X3 to X4 . Then R1 (R2 R3 ) and (R1 R2 )R3 are relations
from X1 to X4 , and
R1 R2 R3 = R1 (R1 R3 ) = (R1 R2 )R3 .
Proof. For any x X1 and y X4 , we have
xR1 (R2 R3 )y x2 X2 s.t. xR1 x2 , x2 R2 R3 y
x2 X2 s.t. xR1 x2 ; x3 X3 s.t. x2 R2 x3 , x3 R3 y
x2 X2 , x3 X3 s.t. xR1 x2 , x2 R2 x3 , x3 R3 y
xR1 R2 R3 y.
Similarly, x(R1 R2 )R3 y xR1 R2 R3 y.
Proposition 6.8. Let R be a relation from X to Y , Ri (i I) a family of
relations from Y to Z, and S a relation from Z to W . Then
S S
(a) R
S iI R
= S iI RRi ;
i
(b) iI Ri S = iI Ri S.

Proof. (a) For any x(R i Ri )z, there exists y Y such that
S xRy and y(i Ri )z.
Then there is one j such that yRi z. Thus x(RRi )z, and so x( i RRi )z. Conversely,
for any x(i RRi )z, there is one i such that x(RRi )z. Then there exists y Y such
that xRy and yRi z. Of course, y(i Ri )z. Thus x(R i Ri )z. The proof for (b) is
similar.

For the convenience of representing composition of relations, we introduce two

operations and on real numbers. For a, b R, define
a b = min{a, b},

a b = max{a, b}.
82 6. BINARY RELATIONS

Lemma 6.9. For a, b, c R,

a (b c) = (a b) (a c),
a (b c) = (a b) (a c).
Proof. We only prove the first formula. The second one is similar.
Case 1: b c. If a c, then the left side is a(bc) = ac = c. The right side
is (a b) (a c) = b c = c. If b a c, then the left side is a (b c) = a c = a.
The right side is (a b) (a c) = b a = a. If a b c, then the left side is
a (b c) = a c = a. The right side is (a b) (a c) = a a = a.
Case 2: b c. If a c, then a(bc) = ac = a and (ab)(ac) = aa = a.
If b a c, then a (b c) = a c = a and (a b) (a c) = a c = a. If a b,
then a (b c) = a b = b and (a b) (a c) = b c = b.

For real numbers a1 , a2 , . . . , an , we define

n
^
ai = min{a1 , a2 , . . . , an },
i=1
n
_
ai = max{a1 , a2 , . . . , an }.
i=1
Let A = [aij ] be an m n matrix and B = [bjk ] an n p matrix. The Boolean
multiplication of A and B is an m p matrix A B = [cik ], whose (i, k)-entry is
n
_
cik = (aij bjk ).
j=1

Theorem 6.10. Let R be a relation from X = {x1 , . . . , xm } to Y = {y1 , . . . , yn }

and let S be a relation from Y to Z = {z1 , . . . , zp }. If MR , MS , and MRS are
matrices of the relations R, S, and RS respectively, then
MRS = MR MS .
Proof. Write MR = [aij ], MS = [bjk ], MR MS = [cik ], and MRS = [dik ].
It suffices to show that cik = dik for each (i, k)-entry of the matrices. In fact, if
cik = 1, it forces that aij bjk = 1 for at least one j, say j0 . Then aij0 = bj0 k = 1.
This means that xi Ryj0 and yj0 Szk . Thus xi RSzk by definition; so dik = 1. If
cik = 0, then aij bjk = 0 for all js; that is, there is no yj Y such that both
xi Ryj and yj Szk . Thus by definition xi is not related to zk by RS. Therefore
dik = 0. This completes the proof of cik = dik .

6.4. Special Relations

Most of time we are interested in some special relations satisfying certain prop-
erties. For instance, the less than relation on the set of real numbers satisfies the
so-called transitive property: if a < b and b < c, then a < c.
Definition 6.11. A binary relation R on a set X is called
(a) reflexive if xRx for all x in X;
(b) symmetric if xRy implies yRx;
(c) transitive if xRy and yRz imply xRz.
 6.4. SPECIAL RELATIONS 83

The relation R is called an equivalence relation if it is reflexive, symmetric, and

transitive; and in this case, if xRy, we say that x and y are equivalent.
The relation I = IX = {(x, x) | x X} is called the identity relation, and
X 2 is called the complete relation on X.

Example Many family relations are binary relations on the set of human beings.
(a) The brother relation Rb : xRb y x and y are both males and have the
same parents. (symmetric and transitive)
(b) The sister relation Rs : xRs y x and y are both females and have the
same parents. (symmetric and transitive)
(c) The brother-sister relation Rbs : xRbs y x is male, y is female, x and y
have the same parents.
(d) The sister-brother relation Rsb : xRsb x is female, y is male, and x and
y have the same parents.
(e) The generalized brother relation Rb0 : xRb0 y x and y are both males and
have the same father or the same mother. (symmetric)
(f) The generalized sister relation Rs0 : xRs0 y x and y are both females and
have the same father or mother. (symmetric)
(g) The relation R: xRy x and y have the same parents. (reflexive, sym-
metric, and transitive; equivalence relation)
(h) The relation R0 : xR0 y x and y have the same father or the same mother.
(reflexive and symmetric)

Example
(a) The less than relation < on the set of real numbers is a transitive
relation.
(b) The less than or equal to relation on the set of real numbers is a
reflexive and transitive relation.
(c) The divisibility relation on the set of positive integers is a reflexive and
transitive relation.
(d) Given a positive integer n; the congruence of modulo n is a relation
n on Z, defined by a n b if and only if b a is a multiple of n. The
standard notation for a n b is a b (mod n). The relation n is an
equivalence relation on Z.

Theorem 6.12. Let R be a relation on a set X. Then

(a) R is reflexive I R all diagonal entries of MR are 1.
(b) R is symmetric R = R1 MR is a symmetric matrix.
(c) R is transitive R2 R MR2 MR .

Proof. (a) and (b) are trivial.

(c) R is transitive R2 R. For any (x, y) R2 , there exists z X such
that (x, z) R and (z, y) R. Since R is transitive, we have (x, y) R. So
R2 R.
R2 R R is transitive. For (x, z) R and (z, y) R, we have (x, y)
2
R R. Then (x, y) R. So R is transitive.
Note that for any relations R and S on X, R S if and only if MR MS .
Also note that the matrix MR2 of R2 is MR2 . We thus have that R2 R if and
only if MR2 MR .
84 6. BINARY RELATIONS

6.5. Equivalence Relations and Partitions

The most important relations among binary relations are equivalence relations.
We will see that an equivalence relation on a set X will partition X into disjoint
equivalence classes.

Example Consider the congruence relation 3 on Z. For each a Z, define

[a] = {b Z | a 3 b} = {b Z | a b (mod 3)}.
It is clear that Z is partitioned into three disjoint subsets
[0] = {0, 3, 6, 9, . . .} = {3k | k Z},
[1] = {1, 1 3, 1 6, 1 9, . . .} = {3k + 1 | k Z},
[2] = {2, 2 3, 2 6, 2 9, . . .} = {3k + 2 | k Z}.
Moreover, [3k] = [0], [3k + 1] = [1], and [3k + 2] = [2] for all k Z.
Theorem 6.13. Let be an equivalence relation on a set X. For each x of
X, let [x] = {y X | x y}. Then
(a) [x] 6= for any x X,
(b) [x] = [y] if x y,
(c) [x] [y]
S = if x 6 y,
(d) X = xX [x].
Each subset [x] is called an equivalence class, and x is called a representative
of [x].
Proof. (a) Each [x] is obviously nonempty because is reflexive.
(b) For any z [x], we have x z by definition of [x]. Since x y, we have
y x by the symmetric property of . Then y x and x z imply that y z
by transitivity of . Thus z [y] by definition of [y]; that is, [x] [y]. Since is
symmetric, we have [y] [x]. Therefore [x] = [y].
(c) Suppose [x] [y] is not empty, say z [x] [y]. Then x z and y z. By
symmetry of , we have z y. Thus x y by transitivity of , a contradiction.
(d) This is obvious because x [x] for any x X.
Definition 6.14. A partition of a nonempty set X is a collection P = {Aj |j
J} of nonempty subsets of X such that
(a) Ai A
Sj = if i 6= j,
(b) X = jJ Aj .
Each subset Aj is called a block of the partition P.
Theorem 6.15. Let P be a partition of a set X. Let RP be the relation on X,
defined by
xRP y there exists a block Aj P such that x, y Aj .
Then RP is an equivalence relation on X, called the equivalence relation deter-
mined by P.
Proof. (a) For each x X, there exits one Aj such that x Aj . Then by
definition of RP , xRP x. So RP is reflexive. (b) If xRP y, then there is one Aj such
that x, y Aj . Again by definition of RP , yRP x. Thus RP is symmetric. (c) If
xRP y and yRP z, then there exist Ai and Aj such that x, y Ai and y, z Aj .
Obviously, z Ai Aj . Since P is a partition. It forces that Ai = Aj . Thus xRP z.
 6.5. EQUIVALENCE RELATIONS AND PARTITIONS 85

This shows that RP is transitive.

For an equivalence relation R on a set X, the collection PR = {[x] : x X}

of equivalence classes of R forms a partition of X, called the quotient set of X
under R. Let E(X) denote the set of all equivalence relations on X and P (X) the
set of all partitions of X. Define functions
f : E(X) P (X) by f (R) = PR ,
g : P (X) E(X) by g(P) = RP .
Then f and g satisfy the following properties.
Theorem 6.16. For any equivalence relation R on X and any partition P of
X,
(g f )(R) = R and (f g)(P) = P.
In other words, f and g are inverse of each other.
Proof. Note that (g f )(R) = g(f (R)) and (f g)(P) = f (g(P)). We have
x[g(f (R))(R)]y A f (R) s.t. x, y A xRy;
A f (g(P)) x X s.t. A = g(P)(x) A P.
Thus g(f (R)) = R and f (g(P)) = P.
Example 6.1. Let Z+ be the set of positive integers. Define the relation on
Z Z+ by
(a, b) (c, d) ad = bc.
Is an equivalence relation? If Yes, what are the equivalence classes?
Let R be a relation on a set X. The reflexive closure is a reflexive relation
r(R) on X such that
(a) R r(R);
(b) if R0 is a reflexive relation on X and R R0 , then r(R) R0 .
The symmetric closure of R is a symmetric relation s(R) on X such that
(a) R s(R);
(b) if R0 is a symmetric relation on X and R R0 , then s(R) R0 .
The transitive closure of R is a transitive relation t(R) on X such that
(a) R t(R);
(b) if R0 is a transitive relation on X and R R0 , then t(R) R0 .
Obviously, the reflexive, symmetric, and transitive closures of R need to be unique
respectively.
Theorem 6.17. For any relation R on a set X,
(a) r(R) = R I;
(b) s(R) = R R1 ;
S
(c) t(R) = R = k=1 Rk .
S
Proof. (a) and (b) are obvious. (c) Note that R k=1 Rk and
!

[ [ [ [ [ [
Ri Rj = Ri Rj = Ri+j = Rk Rk .
i=1 j=1 i,j=1 i,j=1 k=2 k=1
86 6. BINARY RELATIONS

S S
This shows that k=1 Rk is transitive and R k=1 Rk . Since any transitive S
relation which contains R must contain Rk for all positive integers k. Then k=1 Rk
is the transitive closure of R.
Theorem 6.18. Let R be a relation on a set X of n elements. Then
t(R) = R R2 Rn1 .
In particular, if R is reflexive, then t(R) = Rn1 .
Sn1
Proof. It suffices to show that Rl k=1 Rk for all l n; and this is equivalent
S l1
to showing Rl k=1 Rk for all l n. Let (x, y) Rl . There exist x1 , . . . , xl1
X such that all (x, x1 ), (x1 , x2 ), . . ., (xl1 , y) belong to R. Since l n, two
elements of x = x0 , x1 , x2 , . . ., xl1 ,xl = y must be the same, say, xi = xj
with i < j. Then (x0 , x1 ), . . ., (xi1 , xi ), (xj , xj+1 ), . . ., (xl1 , xl ) R imply that
Sl1 Sl1
(x, y) = (x0 , xl ) Rl+ij k=1 Rk . Thus Rl k=1 Rk .
If R is reflexive, we have Rk Rk+1 for all k 1. So t(R) = Rn1 .
Proposition 6.19. Let R be a relation on a set X. Then I t(R R1 ) is an
equivalence relation. In particular, if R is reflexive and symmetric, then t(R) is an
equivalence relation.
Proof. Since I t(R R1 ) is reflexive and transitive, we only need to show
that I t(R R1 ) is symmetric. For (x, y) I t(R R1 ), if x = y, obviously
(y, x) I t(R R1 ). If x 6= y, then (x, y) t(R R1 ). Thus (x, y) (R R1 )k
for some k 1; that is, there is a sequence x = x0 , x1 , . . . , xk = y such that
(xi , xi+1 ) R R1 , 0 i k 1. Since R R1 is symmetric, we have
(xi+1 , xi ) R R1 for all 0 i k 1. This means that (y, x) (R R1 )k .
So (y, x) I t(R R1 ). We have proved that I t(R R1 ) is symmetric.
Now if R is reflexive and symmetric, then t(R) = I t(R R1 ) = t(R). So
t(R) is an equivalence relation.

Let R be a relation on a set X. The reachability relation of R is a relation

R on X, defined by xR y if and only if either x = y or there exist x1 , x2 , . . . , xk
such that (x, x1 ), (x1 , x2 ), . . . , (xk , y) R; that is, R = I t(R).
Theorem 6.20. Let R be a relation on a set X. Let M and M be the matrices
of R and R respectively. If |X| = n, then
M = I M M 2 M n1 .
Moreover, if R is reflexive, then Rk Rk+1 for all k 1 and M = M n1 .
Proof. It follows from Theorem 6.18.

Let R be a relation on X = {x1 , . . . , xn }. If y0 , y1 , . . . , ym is a path in R, the

vertices y1 , . . . , ym1 are called interior vertices of the path. For each 1 k n,
we define a Boolean matrix Wk = [wij ], where wij = 1 if there is a path in R from
xi to xj whose interior vertices are contained in {x1 , . . . , xk }. Since the interior
vertices of any path in R is contained in X = {x1 , . . . , xn }, the (i, j)-entry of Wn
is equal to 1 if there is a path in R from xi to xj ; that is, Wn = MR . We set
W0 = MR . Then we have a sequence of Boolean matrices
W0 = MR , W1 , W2 , . . . , Wn .
 6.5. EQUIVALENCE RELATIONS AND PARTITIONS 87

We will give an algorithm, called Warshalls algorithm, to compute Wk from

Wk1 .
Let Wk1 = [sij ] and Wk = [tij ]. If tij = 1, then there must be a path
xi = y0 , y1 , . . . , ym = xj
from xi to xj whose interior vertices are contained in {x1 , . . . , xk }. We may assume
that the interior vertices y1 , . . . , ym1 are distinct. If xk is not an interior vertex
of this path, then all interior vertices must be actually contained in {x1 , . . . , xk1 },
so sij = 1. If xk is an interior of the path, say xk = yl , we then have two subpaths
xi = y0 , y1 , . . . , yl = xk and xk = yl , yl+1 , . . . , ym = xj
whose interior vertices are both contained in {x1 , . . . , xk1 }, so sik = 1 and skj = 1.
Thus
sij = 1 or
tij = 1
sik = 1 and skj = 1.

Warshalls Algorithm Working on the Boolean matrix Wk1 to produce Wk .

(a) If Wk1 has 1 in (i, j)-entry, so is Wk ; keep 1 there.
(b) If Wk1 has 0 in (i, j)-entry, then check the entries (i, k) and (k, j) in
Wk1 . If both entries are 1, then change the (i, j)-entry in Wk1 to 1.
Otherwise, keep 0 there.

Example Consider the relation R on A = {1, 2, 3, 4, 5}, defined by the Boolean

matrix
0 0 0 0 1
0 1 1 0 0

MR = 1 0 0 0 0 .
0 0 0 1 0
0 0 1 1 0
By Warshalls algorithm, we have

0 0 0 0 1 0 0 0 0 1 0 0 0 0 1
0 1 1 0 0 0 1 1 0 0 0 1 1 0 0

W0 =
1 0 0 0 0 W1 = 1
0 0 0 1 W2 =
1 0 0 0 1

0 0 0 1 0 0 0 0 1 0 0 0 0 1 0
0 0 1 1 0 0 0 1 1 0 0 0 1 1 0

0 0 0 0 1 0 0 0 0 1 1 0 1 1 1
1 1 1 0 1 1 1 1 0 1 1 1 1 1 1

W3 =
1 0 0 0 1 W4 =
1 0 0 0 1 W5 =
1 0 1 1 1 .

0 0 0 1 0 0 0 0 1 0 0 0 0 1 0
1 0 1 1 1 1 0 1 1 1 1 0 1 1 1
Definition 6.21. A binary relation R on a set X is called

(a) asymmetric if xRy implies y Rx;
(b) antisymmetric if xRy and yRx imply x = y.

EXERCISES
(1) Let R be a binary relation from X to Y . Let A and B be subsets of X.
(a) If A B, then R(A) R(B).
88 6. BINARY RELATIONS

(b) R(A B) = R(A) R(B).

(c) R(A B) R(A) R(B).
(2) Let R1 and R2 be relations from X to Y . If R1 (x) = R2 (x) for all x X,
then R1 = R2 .
(3) Let a, b, c R. Then
a (b c) = (a b) (a c),
a (b c) = (a b) (a c).
(4) Let R be a relation from X to Y , Ri (i I) a family of relations from Y
to Z, and
S S a relation
S from Z to W . Then
(a) R
S iI = S iI RRi ;
R i
(b) iI Ri S = iI Ri S.
(5) Let Ri (1 i 3) be relations on A = {a, b, c, d, e}, whose Boolean
matrices are given by

0 1 1 0 1 1 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1 1 0 1

M1 =
0 0 0 0 0 , M 2 = 0 0 0 0 0 , M3 = 0 1 1 0 1

0 1 1 0 1 1 0 0 1 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1 1 0 1
respectively.
(a) Draw the digraphs of the relations R1 , R2 , R3 .
(b) Find the Boolean matrices for the relations R11 , R2 R3 , R1 R1 ,
R1 R11 , R11 R1 , R1 R11 ; and verify that R1 R11 = R2 , R11 R1 =
R3 .
(c) Verify that R2 R3 is an equivalence relation and find the quotient
set A/(R2 R3 ).
(6) Let R be a relation on Z defined by xRy if x + y is an even integer.
(a) Show that R is an equivalence relation on Z.
(b) Find all equivalence classes of the relation R.
(7) Let X = {1, 2, . . . , 10} and let R be a relation on X such that aRb if and
only if |a b| 2. Determine whether R is an equivalence relation. Let
MR be the matrix of R; compute MR8 .
(8) A relation R on a set X is called a preference relation if R is reflexive
and transitive. Show that R R1 is an equivalence relation.
(9) Let n be a positive integer. The congruence relation of modulo n is
an equivalence relation on Z. Let Zn denote the quotient set Z/. For
any integer a Z, we define fa : Zn Zn by fa ([x]) = [ax]. Find the
cardinality of the set fa (Zn ).
(10) For a positive integer n, let (n) be the number of positive integers x n
such that gcd(x, n) = 1, called Eulers function. Let R be the relation
on X = {1, 2, . . . , n}, defined by
xRy if and only if x y, y|n, and gcd(x, y) = 1.
(a) For each y X, find the cardinality |R1 (y)|.
(b) Show that
X
|R| = (x).
x|n
 6.5. EQUIVALENCE RELATIONS AND PARTITIONS 89

(c) Show that |R| = n by proving that the function f : R X, defined

by f (x, y) = xn
y , is a bijection.
(11) Let X be a set of n elements. Show that the number of equivalence
relations on X is
Xn Xn
(l k)n
(1)k .
k!(l k)!
k=0 l=k
(Hint: equivalence relations are in one-to-one correspondence with parti-
tions.)