MIT 18.

211: COMBINATORIAL ANALYSIS

FELIX GOTTI

Lecture 4: Binomial and Multinomial Theorems

In this lecture, we discuss the binomial theorem and further identities involving the
binomial coefficients. At the end, we introduce multinomial coefficients and generalize
the binomial theorem.

Binomial Theorem. At this point, we all know beforehand what we obtain when we
unfold (x + y)2 and (x + y)3 . We can actually use binomial coefficients to generalize
the formulas for the square and cube of a binomial expression.
Theorem 1. For any n ∈ N0 , the following identity holds:
n  
n
X n k n−k
(0.1) (x + y) = x y .
k=0
k

Proof. Observe that each of the summands of the form xk y n−k (for k ∈ J0, nK) that we
obtain after unfolding the left-hand side (x+y)(x+y) · · · (x+y) is the result of choosing
either x or y in each of the n factors x + y and multiplying out all the chosen variables.
Then for a fixed k, the number of copies of the summand xk y n−k we will obtain equals
the number of ways of choosing k copies of x out of the n copies of x + y (note that

choose k copies of x, we are forced to choose n − k copies of y). Thus, we will
once we
get nk copies of xk y n−k for every k ∈ J0, nK, and the identity (0.1) follows. 
Now we can evaluate the identity (0.1) to obtain interesting binomial identities. For
instance,
• nk=0 nk = 2n results from taking x = y = 1 in (0.1), and
P

• nk=0 (−1)k nk = 0 results from taking −x = y = 1 in (0.1).

P

Example 2. A lattice path of length n is a sequence of lattice points p0 , p1 , . . . , pn in

N20 such that p0 = (0, 0) and pi − pi−1 is either (1, 0) or (0, 1) for every i ∈ J1, nK. This
can be pictured as a northeastern path from p0 = (0, 0) to pn = (n − k, k) consisting
of n − k horizontal unit steps and k vertical unit steps. For instance, Figure 1 shows
a lattice path of length 13 with 5 vertical unit steps. Observe that the set of lattice
paths of length n is in bijection with the set of length-n binary sequences: label the
1
2 F. GOTTI

Figure 1. A 13-lattice path.

horizontal and vertical steps of a given lattice path by 0 and 1, respectively, and obtain
the desired binary sequence by reading the path in northeastern direction. Conversely,
given a binary sequence, we can recover its lattice path by sequentially interpreting

the 0’s and 1’s as horizontal and vertical unit steps, respectively. Hence there are nk
lattice paths of length n with k vertical steps. On the other hand, it is clear that there
are 2n lattice pathPof length

n.n Thus, counting lattice paths yields another way to
n n
argue the identity k=0 k = 2 .

Further Binomial Identities. We proceed to give further binomial identities that

are often useful.

Proposition 3 (Vandermonde Identity). For any m, n, k ∈ N0 , the following identity

holds:
  X k   
m+n m n
(0.2) = .
k i=0
i k − i

Proof. Suppose we have a group of m + n students, where m are freshmen and n are
sophomores. By definition, the left-hand side of (0.2) counts the number of k-teams we
can
form
m n

with the m + n students. On the other hand, for each i ∈ J0, kK, we can form
i
k−i
k-teams containing exactly i freshmen: we can first choose the i freshmen in
m n


i
ways and then complete the team choosing k − i sophomores in k−i
ways. Hence
the right-hand side of (0.2) also counts the number of k-teams we can form with the
m + n students. 

Here are two more binomial identities.

Proposition 4. For any n, k ∈ N0 , the following identities hold:

(1) n+1 n n




k+1
= k
+ k+1
;
n
(2) n+1 = j=k kj .

P

k+1
 COMBINATORIAL ANALYSIS 3
Proof. (1) The left-hand side counts all the lattice paths ending at (n − k, k + 1). Now
observe that the number of lattice paths ending at (n − k, k + 1) whose last step
is
vertical equals the number of lattice paths ending at (n − k, k), and so there are nk of
them. Similarly, the number of lattice paths ending at (n − k, k + 1) whose last step is

equals the number of lattice paths ending at (n − k − 1, k + 1), and so there
horizontal
n
are k+1 of them. Thus, the identity follows.
(2) The left-hand side counts the number of (k + 1)-subsets of [n + 1]. On the other
hand, for each i ∈ Jk + 1, n +
1K, we can count the number of (k + 1)-subsets of [n + 1]
whose maximum is i + 1 in ki different ways: first choose i + 1 and then complete the
(k + 1)-subset by choosing k elements from [i]. Since every (k + 1)-subset of [n + 1]
has a maximum in the set Jk + 1, n + 1K, the right-hand side also counts the number
of (k + 1)-subsets of [n + 1]. 

Multinomial Theorem. Our next goal is to generalize the binomial theorem. First,
let us generalize the binomial coefficients. For n identically-shaped given objects and
k colors labeled by 1, 2, . . . ,
k, suppose that there are ai objects of color i for every
n
i ∈ [k]. Then we let a1 ,...,a k
denote the number of ways of linearly arranging the n
given objects.
Proposition 5. Let a1 , . . . , ak be nonnegative integers, and set n = a1 + · · · + ak . Then
k 
n − j−1
  Y P 
n i=1 ai n!
(0.3) = = .
a1 , . . . , a k j=1
aj a1 !a2 ! · · · ak !

Proof. Suppose that we have n identically-shaped given objects of k colors with ai of

color i for each i ∈ [k]. We
can linearly arrange these objects as follows: out of n given
positions, choose a1 in an1 to place the objects of color 1, then out of the remaining
n − a1 positions choose a2 in n−a 1


a2
ways to place the objects of color 2, and so on,
where the step j consists in choosing aj positions in n−(a1 +···+a j−1 )


aj
ways from the
remaining n − (a1 + · · · + aj−1 ) positions. Therefore the first equality in (0.3) holds.
The second equality follows immediately after cancellation. 
We conclude by stating the multinomial theorem. By virtue of the first equality
in (0.3), the proof of the following theorem follows by mimicking that we gave for the
binomial theorem, and so it is left to the reader as a practice exercise.
Theorem 6. For any n ∈ N0 , the following identity holds:
 
X n
n
(x1 + · · · + xk ) = xa11 · · · xakk .
a +···+a =n
a1 , . . . , a k
1 k

Proof. Exercise. 
4 F. GOTTI

Practice Exercises
Pn n

n−1
Exercise 1. Prove that k=1 k k = n2 for every n ∈ N.

Exercise 2. [1, Exercise 4.24] Prove that the number of lattice paths from (0, 0) to
1 2n


(n, n) that never go above the line x = y is n+1 n .

Exercise 3. Prove that for all k, n ∈ N0 , the following identity holds:

      2
2n 2n − 2k 2k 2n n
= .
2k n−k k n k

3n


Exercise 4. [1, Exercise 4.34] Argue combinatorially that n,n,n
is divisible by 6.

References
[1] M. Bóna: A Walk Through Combinatorics: An Introduction to Enumeration and Graph Theory
(Fourth Edition), World Scientific, New Jersey, 2017.

Department of Mathematics, MIT, Cambridge, MA 02139

Email address: [email protected]