Scribd
Upload
15 views

Set Partitions

This document introduces set partitions and Stirling numbers of the second kind. It defines what a partition of a set is, denotes the Stirling number of the second kind S(n,k) as the number of partitions of [n] into k blocks, and provides some examples and properties of S(n,k). It also introduces Bell numbers and provides a recursive formula for calculating them.

Uploaded by

Made Widiarsa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
15 views

Set Partitions

This document introduces set partitions and Stirling numbers of the second kind. It defines what a partition of a set is, denotes the Stirling number of the second kind S(n,k) as the number of partitions of [n] into k blocks, and provides some examples and properties of S(n,k). It also introduces Bell numbers and provides a recursive formula for calculating them.

Uploaded by

Made Widiarsa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

MIT 18.

211: COMBINATORIAL ANALYSIS

FELIX GOTTI

Lecture 7: Set Partitions


In this section we introduce set partitions and Stirling numbers of the second kind.
Recall that two sets are called disjoint when their intersection is empty. A partition
of a set S is a collection π := {B1 , . . . , Bk } consisting of pairwise disjoint nonempty
subsets of S such that S = kj=1 Bj . For each j ∈ J1, kK, the set Bj is called a block of
S

the partition π, and we write |π| = k when π consists of k blocks. In addition, S(n, k)
denotes the number of partitions of [n] having k blocks, and it is called a Stirling
number of the second kind (we will define Stirling numbers of the first kind in coming
lectures).
Let us take a look at some particular cases of the S(n, k). By convention, we assume
that S(0, 0) = 1. Observe that S(n, k) = 0 when k > n. In addition, there is only one
partition of [n] consisting of 1 block (as such a block must be the whole set [n]) and
there is only one partition of [n] consisting of n blocks (as each block is forced to have
size one). Thus, S(n, 1) = S(n, n) = 1.

Example 1. We claim that S(n, 2) = 2n−1 − 1. Indeed, each partition of [n] into
two blocks, namely π := {B1 , B2 }, can be constructed by first choosing the subset B1
in 2n − 2 ways (as B1 cannot be neither empty nor the whose set [n]), which forces
B2 = [n] \ B1 , and then dividing our number of choices, 2n − 2, by 2 to account
for the fact that the order of the blocks inside the partition π is irrelevant. Hence
S(n, 2) = 2n−1 − 1.

Example 2. Let us verify now that S(n, n − 1) = n2 . Indeed, every partition of [n]

into n − 1 blocks must contain exactly one block of size 2, which completely determines
the rest of the blocks, namely the remaining n − 2 blocks of size 1. Therefore the set
of partitions of [n] into two blocks is in bijection with the set of subsets of [n] of size 2.
Hence S(n, n − 1) = n2 .

We can compute the number S(n, k) recursively, as the following theorem indicates.

Theorem 3. For any n, k ∈ N with k ≤ n, the following identity holds:


(0.1) S(n, k) = S(n − 1, k − 1) + kS(n − 1, k).
1
2 F. GOTTI

Proof. By the definition of the Stirling numbers of the second kind, the left-hand side
of (0.1) counts the set of partitions of [n] into k blocks. We will count the same set by
splitting it into two types of partitions: the partitions where n is itself a block and the
partitions where the block containing n has size at least two. To count the partitions
where n is a block by itself, we can take n out, choose a partition of [n − 1] into k − 1
blocks in S(n − 1, k − 1) ways, and enlarge the chosen partition to obtain a partition
of [n] into k blocks by adding {n} as the n-th block. To count the partitions where the
block containing n has size at least two, choose a partition of [n − 1] into k blocks in
S(n − 1, k) different ways, and for each of such choices create a partition of [n] into k
blocks in k different ways by placing n inside one of the k blocks. Putting all together,
we see that the number of partitions of [n] into k blocks is S(n − 1, k − 1) + kS(n − 1, k),
the right-hand side of (0.1). 
We can express the number of surjective functions between finite sets in terms of the
Stirling numbers of the second kind.
Proposition 4. For every n, k ∈ N, the number of surjective functions f : [n] → [k] is
S(n, k)k!.
Proof. To count the surjective functions f : [n] → [k], we can first fix a partition
π = {B1 , . . . , Bk } of [n] into k blocks in S(n, k) ways, then make a linear arrangement
w1 w2 · · · wk with the elements of [k] in k! ways, and then set f −1 (wi ) = Bi . Hence
there are S(n, k)k! surjective functions f : [n] → [k]. 
Stirling numbers of the second kind satisfy the following polynomial identity.
Proposition 5. For every n ∈ N, the following polynomial identity holds:
X n
n
(0.2) x = S(n, k)(x)k ,
k=0

where (x)k = x(x − 1) · · · (x − k + 1).


Proof. First, assume that x belongs to N. Observe that the left-hand side of (0.2)
counts the functions f : [n] → [x]. We can also count such functions as follows. For
each k ∈ J0, nK, we count the functions f : [n] → [x] with |f ([n])| = k, which amounts
to choosing a k-subset S of [x] in xk different ways and then counting the set of
surjective functions from [n] to S, which wePcan verify that is S(n,
Pn k)k! by mimicking
n x

the proof of Corollary 4. Hence there are k=0 k S(n, k)k! = Pk=0 S(n, k)(x)k , and
so (0.2) holds for every n ∈ N. Therefore the polynomial xn − nk=0 S(n, k)(x)k has
degree at most n and more than n different roots, which implies that it must be the
zero polynomial. Hence (0.2) must hold for every x ∈ R. 
For n ∈ N, the total number of partitions
Pn of [n] is denoted by B(n) and called a Bell
number. Then the equality B(n) = k=1 S(n, k) holds. We can compute Bell numbers
recursively using the following recurrence identity.
COMBINATORIAL ANALYSIS 3
Theorem 6. For every n ∈ N0 , the following identity holds:
n  
X n
(0.3) B(n + 1) = B(j).
j=0
j

Proof. By the definition of a Bell number, the left-hand side of (0.3) counts the set of
partitions of [n + 1]. We can count the same set as follows. For each s ∈ J1, n + 1K,
we can count the partitions of [n + 1] where the block B containing {n + 1} has size s:
n

first choose in s−1 ways the elements in B that are different from n + 1, then create
a partition of [n + 1] \ B in B(n + 1 − s) ways. Therefore
n+1   n   n  
X n X n X n
B(n + 1) = B(n + 1 − s) = B(j) = B(j).
s=1
s − 1 j=0
n − j j=0
j


Practice Exercises
Exercise 1. For n ∈ N with n ≥ 3, find a formula for S(n, 3).

n n
 
Exercise 2. Prove that S(n, n − 2) = 3
+3 4
for every n ∈ N with n ≥ 2.

Exercise 3. Argue combinatorially that the number of partitions of [n] with no two
consecutive numbers in any block is B(n − 1).

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]

You might also like