The Binomial Theorem and Combinatorial Proofs: Shagnik Das
The Binomial Theorem and Combinatorial Proofs: Shagnik Das
The Binomial Theorem and Combinatorial Proofs: Shagnik Das
Shagnik Das
Introduction
As we live our lives, we are faced with innumerable decisions on a daily basis1 . Can I
get away with wearing these clothes again before I have to wash them? Which frozen
pizza should I have for dinner tonight? What excuse should I use for skipping the gym
today? While it may at times seem tiring to be faced with all these choices on a regular
basis, it is this exercise of free will that separates us from the machines we live with.2
This multitude of choices, then, provides some measure of our own vitality — the
more options we have, the more alive we truly are. What, then, could be more important
than being able to count how many options one actually has?4 As we have already seen
during our first week of lectures, one of the main protagonists in these counting problems
is the binomial coefficient, whose definition is given below.
Definition 1. Given non-negative integers n and k, the binomial coefficient nk denotes
1
In your homework you are asked to prove this theorem by induction. Here we will
give a different proof.
Proof. The left-hand size represents the product of n copies of x + y:
(x + y)n = (x + y)(x + y) . . . (x + y) .
| {z }
n times
We will now expand this expression, multiplying all the terms out together. Since
there are n linear factors, each monomial will collect one variable from each factor, and
hence will be of degree n. More specifically, each monomial will be of the form xk y n−k
for some k ∈ {0, 1, 2, . . . , n} representing how many times x was chosen.
The coefficient of xk y n−k is the number of ways of choosing x exactly k times. This
is equivalent to choosing a subset of k of the n factors from which to choose x (with y
n
being chosen from the rest). By definition, there are k
such subsets, and hence the
n
k n−k
coefficient of x y is k .
Summing up over the different monomials, we find
n
n
X n k n−k
(x + y) = x y .
k=0
k
Binomial Identities
While the Binomial Theorem is an algebraic statement, by substituting appropriate
values for x and y, we obtain relations involving the binomial coefficients. Such rela-
tions are examples of binomial identities, and can often be used to simplify expressions
involving several binomial coefficients. We provide some examples below.
Corollary 4. The following relations all hold.
(i) For all n ≥ 0, nk=0 nk = 2n .
P
2
Combinatorial Proofs
The Binomial Theorem thus provides some very quick proofs of several binomial identi-
ties. However, it is far from the only way of proving such statements. A combinatorial
proof of an identity is a proof obtained by interpreting the each side of the inequality
as a way of enumerating some set. If they are enumerations of the same set, then by
the principle of double-counting it follows that they must be equal. If they are different
sets, but you can build a bijection between the two, then the bijection rule shows they
must be equal.
Since the binomial coefficients are defined in terms of counting, identities involv-
ing these coefficients often lend themselves to combinatorial proofs. These proofs are
usually preferable to analytic or algebraic approaches, because instead of just verifying
that some equality is true, they provide some insight into why it is true. Moreover,
once one has identified a bijection between two sets, restricting the bijection to certain
subsets can often lead to several other identities.6
To illustrate the concept, we provide combinatorial proofs of the identities in Corol-
lary 4.
Proof of Corollary 4(i). For (i), let S be a set of n elements, and count 2S , which is
the collection of all subsets of S. On the one hand, if we write S = {s1 , s2 , . . . , sn }, we
can determine any subset X ⊆ S by asking n questions: “is s1 in X?”, “is s2 in X?”,
and so on until “is sn in X?”. Each question has two possible answers, and so by the
product rule there are 2n possible subsets.
On the other hand, we can classify the subsets based on their size: Sk is the
collection of subsets
of S of size k. As the subsets
P can range in size from 0 to n, we
n
have 2S = ∪˙ k=0 Sk . By the sum rule, 2S = nk=0 Sk . By definition, Sk = nk ,
and so we have n
n
S X n
2 = 2 = .
k=0
k
n
Proof of Corollary 4(ii). Let S be a set of nPelements, S = {s 1 , s2 , . . . , s n }. Since k
n
counts the number of subsets of S of size k, k odd k
counts the number of subsets of
n
P
S of odd size. Similarly, k even k enumerates the subsets of S of even size. To prove
the identity, we build a bijection f : 2X → 2X .
Given X ⊆ S, define
(
X ∪ {sn } if sn ∈ /X
f (X) = .
X \ {sn } if sn ∈ X
For every set X, we have f (f (X)) = X, since we either add and remove sn to X,
or remove and add back sn to X. In particular, f is invertible, and so a bijection.
Moreover, note that f (X) changes the size of X by exactly one, and so f maps odd
subsets to even subsets, and even subsets to odd subsets. Hence, restricting to the
collection of odd-sized subsets, f gives a bijection to the collection of even-sized subsets.
This shows that the two sides of the identity are indeed equal.
6
Our proof of Corollary 4(ii) below is an example of this phenomenon.
3
Proof of Corollary 4(iii). Let S be a set of n elements. We know that nk denotes the
number of subsets of S of size k. From (i), we know that 2k denotes the number of
n k
subsets of a set of size k. Hence, by the product rule, k 2 counts the number of ways
of choosing a subset B ⊆ S of size k, and then choosing a further subset A ⊆ B (of
arbitrary size). Taking a sum over all k between 0 and n then enumerates over all
possible sizes of the subset B, and hence over all possible choices of B. Thus the left-
hand side counts the number of ways of choosing sets A and B such that A ⊆ B ⊆ S.
Next we claim that 3n counts the number of ways of partitioning S into three sets,
so that S = C ∪D˙ ∪E.
˙ Indeed, for every element si ∈ S, we ask which subset si belongs
to. As there are three choices for each element (C, D or E), there are 3n possible
partitions.
To finish, we biject between the pairs (A, B) with A ⊆ B ⊆ S and the triples
(C, D, E) with S = C ∪D ˙ ∪E.
˙ We can define
Conclusion
Binomials coefficients are omnipresent in combinatorics, arising naturally in several
contexts. While they can sometimes be difficult to work with directly, binomial iden-
tities often allow us to simplify expressions involving these coefficients, and hence it is
useful to have a large collection of identities.7
The Binomial Theorem is a great source of identities, together with quick and short
proofs of them. However, given that binomial coefficients are inherently related to
enumerating sets, combinatorial proofs are often more natural, being easier to visualise
and understand. Furthermore, they can lead to generalisations and further identities.
There are also several identities that do not follow from the Binomial Theorem,
and you will meet a few on your homework. While it is again sometimes possible to
prove these by using the formula for the binomial coefficients and slogging through some
algebra, a combinatorial proof is usually preferable.
The above examples may have seemed rather mundane, with more work required
for little reward. However, there are several examples in enumerative combinatorics of
identities for which analytic proofs are known, but combinatorial proofs are desired for
the extra insight they would bring. After all, mathematics is more about explaining
why things are true rather than merely determining that they are.
7
Could a shortage of these equalities be referred to as an identity crisis?