Enumerative Combinatorics . . . How do I love thee? Let me count
the ways! (E. B. Browning) Count the number of elements in a (finite) set, or the number of ways of arranging the elements of finite sets into definite patterns. For example: (1) how many ways are there to climb a staircase with 12 stairs if you take the stairs either 1 or 2 at a time? (2) how many ways are there to arrange the numbers from 1 to 10 around a circle so that no two consecutive numbers are adjacent? Such questions have ancient roots
magic squares (China, 2200 BCE)
permutations (China, 1100 BCE) n-permutations (Sefer Yetzirah, 200-500 CE) (movement of heavenly bodies) Pascal, Fermat (16-17th CE) (games of chance, or gambling) 1
MAT344
Professor S. Tanny
We begin with some puzzles and games.
A.
In the above, find the fewest number of moves required
to rearrange the balls from (a) to (b), where only 1 ball at a time can be moved.
Note: Count the number of moves that are required in your solution. Be sure to show why the task cannot be accomplished in fewer moves.
MAT344 B.
Professor S. Tanny
Tower of Hanoi: There are n rings on Tower 1, each of successively
smaller size. You are only allowed to move one ring at a time. You can never put a larger ring on top of a smaller ring. You are allowed to put rings on either of towers 2 or 3 temporarily. Find the minimum number of moves that are required to transfer all the n rings from the first tower to the second tower.
Note: Is it apparent that this can always be accomplished at all? Why?
A start at the solution: Let an be the required number. Think small!! For n = 1, we have a1 = 1. For n = 2 we can first show that a2 3 by demonstrating a way to accomplish the task in 3 moves. Then we observe that a2 > 2 since we need 2 moves to move the bottom ring. It follows that a2 = 3. What is the solution for general n? 3
MAT344
Professor S. Tanny
C. Game of SIM: In this game there are 6 dots in the plane and 2 players, Red and Blue. Each player alternately makes a move by joining a pair of dots by a line in his (her) colour, either Red or Blue. The winner is the first player to form a triangle in a single colour (note: the sides of the triangle must originate at the 6 dots, and not at the intersections of the interior lines drawn). Is this a finite game? What is the maximum number of moves? Would you rather go first or second? Is there always a winner? Is there a winning strategy for either player? If so, for which player?
Hint: To help you get the idea, think of another game between two players that is finite and determine the answers to the above questions for that game. 4
MAT344
Professor S. Tanny
Basic Counting Rules and Why They Work
We want to represent the letters of an alphabet with 2 digit binary strings. How many different letters can we represent? (This is the maximum size of the alphabet.)
22
01 00
10 11
Product Rule: If there are n1 ways to do a first thing, and n2 ways to
do a second (irrespective of the outcome for the first) then there are n1 n2 ways to do both things.
MAT344
Professor S. Tanny
Example: How many different Ontario licence plates can be issued, if
each consists of 3 letters followed by 3 numbers?
26
26
26
10
10
10
Think about this variation of the above question: How many different Ontario licence plates can be issued, if each consists of 3 letters followed by 3 numbers, where the three letters and the three numbers can come in any order?
MAT344
Professor S. Tanny
Heres a more formal mathematical approach to the Product Rule.
Suppose S is the set of solutions to some combinatorial problem. Suppose we can write S as a cross product of two sets, that is, in the form S = S1 S2. Then card (S) = card (S1) card (S2), where card (S) is the number of elements in S. More generally, if S = S1 S2 . . . Sn, then card (S) =
n Y
card (Si).
i=1
Exercise: Find the number of divisors of 21? Of 21,168?
(Hint: 21, 168 = 24 33 72). Example: We have 16 different books: 5 in Spanish, 4 in French, and 7 in English. In how many ways can we choose three books, no two of them in the same language? Since no two books can be from the same category, we must choose one book from each category. There are 5 ways to choose a Spanish book, 4 ways to choose a French book, and 7 ways to choose an English book. Using the product rule, we get 5 4 7 ways. 7
MAT344
Professor S. Tanny
Example: We have 16 different books: 5 in Spanish, 4 in French, and
7 in English. In how many ways can we choose a pair of books that are not both in the same language? There are 3 different possible pairs of books in different languages, namely, S F , S E and F E, and our choice will involve exactly one of these pairs. First we compute the number of ways we could draw a pair of each type. There are 5 Spanish books and 4 French books, so using the product rule there are 20 different choices for the S F pair. Similarly we count the number of choices for the other two pairs. But since we can choose only one of these pairs, we choose either S F (20 ways) or S E (35 ways) or F E (28 ways). So the total number of ways is the sum of these. Thus: S-F S-E F-E
20 35 28 83 8
MAT344
Professor S. Tanny
Sum Rule: If you can do one thing in n1 ways and a second thing in n2 ways, there are n1 + n2 ways that you can do either first thing or second thing, but not both. Here is a more formal mathematical approach to the sum rule: let S be the set of solutions to some mathematical problem. Suppose we can write S = S1 S2, where S1 S2 = . Then card (S) = card (S1)+ card (S2). More generally, if we have S =
n [
Si, and Si Sj = if i 6= j (that is,
i=1
the sets are pairwise disjoint), then card (S) =
n P i=1
card (Si).
Note: The collection {S1, S2, . . . , Sn} is called an n-partition of the set S. Exercise: Use mathematical induction to prove the general form of the product rule and the sum rule. 9
MAT344
Professor S. Tanny
Example: How many 3 letter sequences (so order counts) using the letters a, b, c, d, e, f are there of the following type: a) b) c) d)
allow repetition no repetition no repetition and must contain e allow repetition and must contain e
a) b) c) d)
63 (product rule) 6 5 4 (product rule) 60 = (5 4) + (5 4) + (5 4) (product and sum rules) 63 53 = 91 (What do 63 and 53 each count?)
Note: You can solve part d) by considering three separate cases,
i.e., sequences with 1e, 2e, or 3e. Try it! 10
MAT344
Professor S. Tanny
Exercise: Suppose two baseball teams A and B play a best of five
series; that is, the winner is the team to first win 3 games. Find the number of different possible outcomes in which team A wins the series. Represent the winner of a game by its letter. Then any outcome of the series in which A wins consists of an arrangement of A0s and B 0s with at least 3 A0s and at most 2 B 0s, and where A must be the last letter in the sequence. If the sequence has only 3 letters, then the sequence consists of 3 A0s, so there is only 1 such sequence. If it has 4 letters, it ends in A and there are 3 places to put the single B, so there are 3 sequences in all of this type. Finally, if it has 5 letters, it must end in A and there are 2 B 0s among the first 4 letters. Verify that there are 6 different possibilities. Using the sum rule we conclude that there is a total of 10 sequences in which A wins the series. 11
MAT344
Professor S. Tanny
Permutations We now count the number of ways to arrange the elements of a set. Recall that by definition a set contains distinct objects. For example, the set S = {1, 2, 3, 4} is a set with 4 elements (cardinality 4). Sometimes we also count arrangements of multisets. A multiset is a collection of not necessarily distinct elements, so it may contain multiple copies of the same element. For example, M = {13, 2, 35} denotes a multiset M containing 3 copies of the element 1, a single element 2, and 5 copies of the element 3. An arrangement of a multiset is equivalent to arranging a set but allowing some number of repetitions of one or more of the elements. Let S be a set with n elements. By the Product Rule there are n! = n (n 1) . . . 3 2 1 ways to arrange the elements of S. Further, there are n(n 1) (n r + 1) ways to create r-element sequences (that is, arrangements of r of the elements of S) drawn from the n elements of S. 12
MAT344
Professor S. Tanny
We introduce the following notation:
Set of the first n positive integers: [n] = {1, 2, , n} Alternate way to write n!: nn n! Falling factorial: nr = n(n 1) (n r + 1) = (nr)!
Rising Factorial:
r nr = n(n+1) (n+r1) = (n+r1)! = (n+r1) (n1)!
What is the value of 00?
By convention, 00 0! = 1. Heres one reason why: by definition, for any positive integer n, nn = n (n 1)n1. Take n = 1 : 11 = 1 = 1 00 00 = 1 13
MAT344
Professor S. Tanny
Counting the Number of Subsets of the Set [n]
For any nonnegative integer k, a k-subset of [n] is a subset of [n] with k elements. If k = 0 then there is only 1 subset, namely, the null set . How many k-subsets of [n] are there? An example: suppose n = 4 and k = 2. We enumerate all the 2-element subsets of [4]: 12 34
13 23
14 24
Thus, there are precisely 6 2-element subsets of [4].
How can we answer the question in general? 14
MAT344
Professor S. Tanny
Let x be the number of k-subsets of [n]. By what we showed earlier for
permutations, we know that each k-subset of [n] can be arranged in k! different ways. Thus, xk! counts the number of ordered k-subsets of [n]. But any such ordered k-subset is, by definition, just a k-permutation, and we know that there are nk k-permutations. Thus we conclude that ! k n x k! = nk , so x = nk! . Another way of writing this expression is , k which is called a binomial coefficient and read as n!choose k. Finally, n n! from this it follows immediately by algebra that = k!(nk)! . k !
What is
3 ? For n 0, what is 4
n ? 0
1 n0 n = 0 = = 1, 0 0 1 !
n0 15
MAT344
Professor S. Tanny
Example: A box contains 7 distinct balls numbered 1 through 7. Three
balls (1,2,3) are blue, two (4,5) are red, and two (6,7) are green. Find the number of selections of three balls which contain at least two blue balls. !
Solution 1: Choose the two blue balls in
3 2
= 3 ways. Choose the
!
remaining ball from the five remaining balls in
5 1
= 5 ways. By the
product rule, the answer is 3 5 = 15.
16
MAT344
Professor S. Tanny
Example: A box contains 7 distinct balls numbered 1 through 7. Three
balls (1,2,3) are blue, two (4,5) are red, and two (6,7) are green. Find the number of selections of three balls which contain at least two blue balls. Solution 2: Each selection will be one of the following: a)
3 blue balls - 1 way
b)
blue balls (1,2) plus one ball of another colour - 4 ways
c)
blue balls (1,3) plus one ball of another colour - 4 ways
d)
blue balls (2,3) plus one ball of another colour - 4 ways
Now add these up, using the sum rule, to get 13. Which answer is correct, and why?
17
MAT344
Professor S. Tanny
The triangle formula is a very important combinatorial identity involving
binomial coefficients: !
n+1 n n = + k k k1
We use a combinatorial argument to prove this identity. The strategy
is to show that both sides of the identity count the same thing in different ways. By definitions, the left hand side of the identity counts all of the ksubsets of the set [n + 1]. Now we show that the right hand side counts these k-subsets too.
18
MAT344
Professor S. Tanny
Fix your eye on any specific element of [n + 1], say the element (n + 1). For any k-subset of [n + 1], (n + 1) is either in or out of the k-subset. But a k-subset of [n +!1] where (n + 1) is out ! is a k-subset of [n], and n n we know there are such k-subsets, so counts all the k-subsets k k of [n + 1] where (n + 1) is out. Further, any k-subset of [n+1] where (n+1) is in must contain ! a unique n set of k 1 additional elements from [n]. There are ways to k1 ! n choose these k 1 elements so counts all k-subsets of [n + 1] k1 where (n + 1) is in. By the sum rule, the right hand side also counts all the k-subsets of [n + 1]. 19
MAT344
Professor S. Tanny
Here is an algebraic proof of the previous identity. This proof uses the factorial definition of the binomial coefficient: !
n n + k k1
n! n! + (k1)!(nk+1)! = k!(nk)!
h i n! 1+ 1 nk+1 (k1)!(nk)! k
h i n+1 n! (k1)!(nk)! k(nk+1)
(n+1)! = k!(nk+1)!
n+1 k
Notice that this proof does not work for k = 0 or k = n + 1? Why? To
cover these cases you must do them separately. As you can see, the algebraic proof is not nearly as elegant or illuminating as the combinatorial proof. In this course we give combinatorial proofs whenever we can. 20
