COMPSCI 230: Discrete Mathematics for Computer Science April 8, 2019

Lecture 22
Lecturer: Debmalya Panigrahi Scribe: Kevin Sun

1 Overview
In this lecture, we begin studying the fundamentals of counting discrete objects. We will also give
formalizations of rules and principles that we often use when counting objects.

2 Permutations and Combinations

In this section, we will introduce the foundations of combinatorics, the branch of mathematics that
deals with counting. In particular, we will study permutations and combinations, their relevant
formulas, and some basic identities involving these concepts.
Permuting n items: Suppose that we want to place n distinct items in a line, that is, we want to
order or permute the items. The most fundamental principle used in counting is the following: the
number of ways to permute n distinct items is the product of the first n positive integers, which
is denoted by n! (“n factorial”). This can be proven, somewhat informally, as follows: there are n
choices for the first item in line, (n − 1) for the second, and so on, until there’s only 1 choice left for
the last item in line. Since the choices can be made in sequential order, the total number of possible
permutations (orderings) is n(n − 1) · · · 1 = n!.
Permuting k of n items: Now instead of permuting all n items, suppose we only want to permute k
of them (where k is a positive integer less than n). By the same argument above, the number of ways
to do this is n(n − 1) · · · (n − (k − 1)). Notice that there are k terms in this product, because each
term corresponds to selecting an item to place next in the final ordering of k items. For convenience,
notice that a more concise way of writing this product is the following:

n!
n(n − 1) · · · (n − (k − 1)) = n(n − 1) · · · (n − k + 1) =
(n − k)!

Choosing k of n items: Now instead of permuting k items, suppose we only want to choose k items.
In other words, we are finding a subset of size k, rather than a sequence of length k. We denote
this value by (nk) (“n choose k”), and we often say a subset is a combination. In other words, a
combination can be thought of as a permutation in which order doesn’t matter.
As we just saw, there are n!/(n − k )! permutations of length k. However, each subset of size k is
represented as a permutation k! times. Therefore, we can conclude that:
 
n n!
=
k (n − k)!k!
Alternatively, we can reason about these formulas as follows. Let k be any positive integer at
most n. Then permuting n items is the same as choosing a subset of size k, ordering them, and then

22-1
ordering the remaining elements. This line of reasoning results in the following:
 
n
n! = · k! · (n − k)!
k
which is equivalent to the previous expression.
Remark: If n = 0, then the number of ways to order n items is somewhat ambiguous. Similarly,
if k = 0, then the number of ways to choose k items is somewhat ambiguous. To deal with these
ambiguities in a consistent way, we have the following conventions:
 
n
0! = 1 and = 1.
0

2.1 Stars and Bars

The notion of combinations is fundamental to combinatorics.To better familiarize ourselves with
combinations, we now look at one application known as “stars and bars”.
The setup is the following: suppose there are three children c1 , c2 , c3 , and we must distribute
10 identical candies among these three children. Each child can receive any number of candies,
including 0. For example, one possible distribution is (4, 3, 3): in this case, c1 receives 4 candies, c2
receives 3, and c3 receives 3. How many ways can we distribute the candies?
The key observation is the following: we can distribute the candies by arranging them in a line,
and then placing two “bars” somewhere along the line. For example, the (4, 3, 3) described above
can be modeled by the following:

F F F F F F F F F F

Each F represents a candy, and the two location of the two bars determines the distribution of
the candies. Notice that the following distribution is also possible:

F F F F F F F F F F

The above diagram corresponds to the distribution (0, 0, 10). In general, c1 receives the candies left
of the first bar, c2 receives the candies between the two bars, and c3 receives the candies right of the
second bar.
So we can see that distributing candies is identical to choosing the location of the two bars.
However, the number of ways to do this is not (11 2 ), because this ignores the possibility of placing
two bars next to each other. Instead, we should think of the process as follows: there are 12 empty
slots, pictured below.

We must place bars in two of the slots, and the remaining 10 slots will then represent the 10 candies
to distribute. For the distribution (5, 5, 0), the diagram becomes the following:

22-2
From this perspective, it becomes clear that the number of distributions of 10 candies to 3 children
is (12
2 ). In general, if there are n candies and k children, then there are n + k − 1 slots, and we must
place k − 1 bars. The remaining n candies, interspersed among the bars, represent a distribution.
Thus, the number of distributions is (n+k−k− 1
1 ).

2.2 Combinatorial Identities

We now state a couple of basic identities involving combinations.

Fact 1. Let n be a positive integer, and k be in {1, . . . , n}. Then the following identities hold:

n−1 n−1
         
n n n
= and = + .
k n−k k k−1 k

The latter is known as Pascal’s rule, named after the mathematician Blaise Pascal.

Remark: One proof of these identities is purely algebraic: if we simply use the formula
 
n n!
=
k (n − k)!k!

on every term, then a bit of algebraic manipulation proves the identity. However, in some sense,
this proof is not very elegant because it does not take advantage of the combinatorial interpretation
of the terms. Therefore, we now prove each identity by giving a combinatorial interpretation of
both sides. Under these interpretations, the validity of the equalities becomes clear.

Proof. Let A be a set with n elements, where n is a positive integer. As discussed above, for each
equality, we will give a combinatorial interpretation for both the left-hand side (LHS) and the
right-hand side (RHS).
In the first equality, the LHS counts the number of subsets of A that contain exactly k items.
Selecting k items is equivalent to excluding n − k items. The number of ways to exclude n − k items
from S is precisely the RHS of the equality.
In the second equality, the LHS still counts the number of subsets of A that contain exactly
k items. Since n is positive, we can fix a particular element of A which we denote by x. The set
of subsets of size k can be partitioned into two sets: those that contain x, and those that do not.
−1
The number of subsets of size k that contain x is (nk− 1 ). This is because once you choose x in the
subset, you are left to make k − 1 choices out of the remaining n − 1 elements. On the other hand,
the number of subsets of size k that exclude x is (n−k 1). The logic behind this expression is that if
you exclude x from the subset, then you are still left to make all the k choices, but only among the
remaining n − 1 elements. Summing these two values gives the number of ways to choose any
subset of size k, as desired.

A pictorial representation of Pascal’s rule (see Fact 1) is known as Pascal’s triangle. The triangle is
constructed as follows: the top row, which we consider row 0, contains a single 1. Each subsequent
row starts and ends with 1, and the internal terms are obtained by summing the closest two terms
in the previous row. The first 7 rows are pictured below, where n represents the row index:

22-3
 n =0 1
n =1 1 1
n =2 1 2 1
n =3 1 3 3 1
n =4 1 4 6 4 1
n =5 1 5 10 10 5 1
n =6 1 6 15 20 15 6 1
Now consider the values in row n. We claim that these values are precisely, in order, the following:
       
n n n n
··· .
0 1 n−1 n

The k-th term of this row (where k ∈ {0, 1, . . . , n}) is precisely (k−n 1)—the proof of this is a simple
induction argument that follows easily from Pascal’s identity.
We now state and prove another well-known identity; this one gives a combinatorial interpreta-
tion of the coefficients of a binomial expression. Due to this connection, the values (nk) are often
called the Binomial coefficients.

Theorem 2 (Binomial Theorem). For all n ∈ Z+ and a, b ∈ R,

n  
n n−k k
( a + b) = ∑
n
a b .
k =0
k

In other words, for every k ∈ {0, 1, . . . , n}, the coefficient of an−k bk in the expansion of ( a + b)n is (nk).

As an example, let us consider the case when n = 4 and k = 3. Then the coefficient we seek is
that of a1 b3 . There are precisely four ways to obtain a1 b3 by expanding ( a + b)4 : abbb, babb, bbab,
and bbba. Each of these ways corresponds to choosing the location of 3 b’s from four slots, and the
number of ways to do this is precisely (43).
Now we consider the general case. Let k ∈ {0, 1, . . . n}, and consider the term an−k bk . This term
is the product of (n − k ) a’s and k b’s. Furthermore, notice that ( a + b)n is simply

( a + b)( a + b) · · · ( a + b),

which is the product of n copies of ( a + b). Upon expansion of this product, we obtain the sum of a
n + 1 terms, and each term is the product of n variables, distributed as a’s and b’s.
Thus, we can think of “building” the coefficient of an−k bk as selecting k of the ( a + b) terms in
( a + b)n that will contribute a b. The number of ways to do this is precisely (nk), as desired.
Corollary 3. For all n ∈ Z+ , the following equality holds:
n  
n
2 = ∑
n

k =0
k

In other words, the sum of the values in the n-th row of Pascal’s triangle is 2n .

Remark: The above identity simply follows by a = b = 1 in Theorem 2. We give an alternate proof
below that relates this identity to the set of subsets of a set.

22-4
Proof. Let A be a set with n elements, and let Ak denote the subset of the power set 2 A containing
the subsets of A of size k. Then the sets A0 , A1 , . . . , An partition 2 A , which means the following
equalities hold:

2n = | 2 A | = | A 0 | + | A 1 | + · · · + | A n |
      n  
n n n n
= + +···+ =∑ .
0 1 n k =0
k

In other words, every subset of A has a size in {0, 1, . . . , n}, so to count the number of subsets of A,
we can count the number of subsets of each size over all possible sizes.

3 Cardinality Rules and Principles

In this section, we will see the formalization of counting strategies that we often take for granted:
the product rule, the sum rule, and the pigeonhole principle.

3.1 The Product Rule

Before stating this rule, we first introduce some notation. For any set S and positive integer k, we
let Sk denote the set of k-length sequences whose elements are in S, i.e.,

Sk = {(s1 , s2 , . . . , sk ) : ∀i ∈ {1, 2, . . . , k }. si ∈ S}.

Notice that S2 = S × S, but S3 6= (S × S) × S because the elements of S3 are of the form ( a, b, c),
while the elements of (S × S) × S have the form (( a, b), c). However, for notational convenience,
we often remove the internal parentheses and write ( a, b, c) instead. Using this convention, the set
Sk is the Cartesian product of S with itself k times.
The product rule states the following: If A and B are finite sets, then | A × B| = | A| · | B|. This rule
allows us to count the number of different ways to “combine” choices. For example, suppose we
must create a password that contains five letters of the alphabet (all lowercase), followed by three
digits (i.e., 0-9). Let A = { a, b, c, . . .} denote the set of letters, and B = {0, 1, . . . , 9} denote the set of
digits. Then to create a valid password, we must select five elements of A (order matters), as well
as three elements of B (order matters as well).
Thus, the set P of valid passwords is essentially the set A5 × B3 : for example, “apple123” is
a valid password, and ( a, p, p, l, e, 1, 2, 3) is an element of P (with internal parentheses removed).
By the product rule, we have | P| = | A5 × B3 | = | A5 | · | B3 | = 265 · 103 . Thus, there are 265 · 103
passwords that satisfy the given conditions.

3.2 The Principle of Inclusion-Exclusion

The principle of inclusion-exclusion (PIE) states the following: If A and B are finite sets, then
| A ∪ B| = | A| + | B| − | A ∩ B|. Note that if A ∩ B = ∅, then PIE states | A ∪ B| = | A| + | B|; this
special case is known as the sum rule. The name of this principle comes from the fact that we must
“exclude” the elements of A ∩ B after “including” each of them twice—once as an element of A,
and once as an element of B.

22-5
 Inclusion-exclusion allows us to count the number of objects that satisfy some property or
another, without over-counting the objects that satisfy both. For example, suppose 15 students in
the class own a cat, 20 own a dog, and 10 own both. Then the number of students that own a cat or
a dog (or both) is not 15 + 20 = 35, but rather, 15 + 20 − 10 = 25 because the 10 people that own
both are included in both the 15 cat owners and the 20 dog owners.

3.3 The Pigeonhole Principle

A pigeonhole is a small box in which a single pigeon can stand. If there are n pigeonholes and
over n pigeons, then clearly, some pigeonhole must accommodate at least 2 pigeons—this idea is
known as the pigeonhole principle. More formally, the pigeonhole principle states the following: If
A and B are finite sets such that | A| > | B| and f : A → B is a total function, then there exist a1 , a2 ∈ A
such that a1 6= a2 and f ( a1 ) = f ( a2 ). Thus, we can think of A as the set of pigeons, B as the set of
pigeonholes, and f as the function that matches each pigeon with a pigeonhole. If | A| > | B| then
some pigeonhole must accommodate at least 2 pigeons, which is precisely what the principle states.
The pigeonhole principle sounds quite obvious, yet when applied correctly, it can be surprisingly
powerful.
Example 1: Consider a subset S of {1, 2, . . . , 10}. We claim that if |S| ≥ 6, then there exist two
distinct elements of S such that one is a factor of the other. This proposition can be verified by
checking all (10
6 ) subsets, but we can provide a more elegant proof using the pigeonhole principle.
Notice that every positive integer n can be written as 2k · q(n), where k is a nonnegative integer
and q(n) is odd. Now consider writing the elements of S in this way. There are only 5 possible
values of q(n) (i.e., {1, 3, 5, 7, 9}), so there exists some odd integer q and a, b ∈ S such that a = 2α q
and b = 2β q, where α, β are distinct nonnegative integers. If α < β then a is a factor of b, and if
α > β, then b is a factor of a. In this example, we can think of the “odd component” of the elements
of S as the pigeons, and the set {1, 3, 5, 7, 9} as the pigeonholes.

4 Summary
In this lecture, we studied the fundamentals of combinatorics. We proved several combinatorial
identities and saw applications of cardinality rules and the pigeonhole principle.

22-6