Discrete Mathematics (Spring 24) Homework 6 Solution

 
x11 x12 . . . x1n
1. Prove that the number of 2 × n arrays that can be made from
x21 x22 . . . x2n
the numbers 1, 2, . . . , 2n (without repetition), such that x11 < x12 < . . . < x1n , x21 <
x22 < . . . < x2n , and x1i < x2i for i = 1, . . . , n, is Cn .
Solution I: We describe a bijection between the arrays in the problem and Dyck
sequences (sequences with n 1’s and n −1’s whose partial sums are non-negative).
Given an array in the problem, define a1 , . . . , a2n by setting ai = 1 if i is on the first
row of the array, and setting ai = −1 otherwise. The sequence has n 1’s and n −1’s as
there are n numbers on each row. If a1 +. . .+ak < 0, then there are more numbers from
1, . . . , k that are on the second row (and occupy x21 , . . . , x2s by the increasing condition
for rows) than numbers from 1, . . . , k that are on the first row (and occupy x11 , . . . , x1r ).
But with s > r, x1s is not from 1, . . . , k, thus x1s > k ≥ x2s , a contradiction.
Conversely, given a Dyck sequence, construct an array by placing all indices i where
ai = 1 on the first row in an increasing order, and placing the rest of indices in an
increasing order on the second row. From the construction, both rows are increasing,
and if x1s > x2s = k for some s, then the number of indices from 1, . . . , k on the first
row is less than s, and we have a1 + . . . + ak < 0, a contradiction. Finally, these two
maps can be easily seen as inverses of each other so they are bijections.
Solution II: We show the number Cn′ of 2 × n arrays satisfies the Catalan recurrence.
The only legit array for n = 0 is the empty array so C0′ = 1 = C0 . Now consider n ≥ 2.
We can group the legit arrays according to the first value of k such that 1, 2, . . . , 2k are
placed in the leftmost 2 × k subarray (k can be n), in which case x1,1 = 1, x2,k = 2k,
and 2k + 1, . . . , 2n must be placed in the rightmost 2 × (n − k) array.
 
x1,k+1 − 2k + 1 . . . x1n − 2k + 1
For these arrays, is a legit array of 1, . . . , 2(n−k),
x2,k+1 − 2k + 1 . . . x2n − 2k + 1 
x1,2 − 1 x1,3 − 1 . . . x1,k − 1
and we claim that is a legit array of 1, . . . , 2(k −
x2,1 − 1 x2,2 − 1 . . . x2,k−1 − 1
1). The content being correct and the increasing condition of rows follow from the
construction.
  Suppose there is a violation of increasing condition along the column
x1,l − 1
, i.e. x1,l > x2,l−1 for some l ≤ k. Then x2,l−1 is less than x2,l < . . . < x2,k
x2,l−1 − 1
and x2,l−1 < x1,l < x1,l+1 < . . . < x1,k , while larger than x2,l−2 > . . . > x2,1 amd
x1,l > . . . > x1,1 , which forces x2,l−1 = 2(l − 1), moreover, 1, . . . , 2(l − 1) are placed in
the leftmost 2 × (l − 1) subarray, a contradiction that k is the smallest value for that to
happen. Conversely, every pair of such legit arrays of size 2×(k −1) and 2×(n−k) can

1
 be combined array of size 2 × n. Summing over all k and apply induction,
into a legit P
Cn = k=1 Ck−1 Cn−k = nk=1 Ck−1 Cn−k = Cn .
′
Pn ′ ′

2. The Bell number Bn is the number of partitions of the set {1, 2, . . . , n}. Show that
n  
X n
Bn+1 = Bn−k .
k=0
k

Solution: The L.H.S. is the number of partitions of the set {1, 2, . . . , n + 1}. We can
group these partitions according to the size (denoted by k + 1) of the block the element
n + 1 is in and count each group individually. For a given number k, there are nk ways

to choose which other elements of {1, 2, . . . , n} are in the block,
while the remaining
n − k elements form a partition themselves, so the group has nk Bn−k permutations.
Summing the sizes over all k = 0, 1, . . . , n gives the R.H.S.

3. Let Πn denote the set of all partitions of the set {1, 2, . . . , n} into non-empty sets.
Given two partitions π and σ in Πn , define π ≤ σ, provided that each part of π is
contained in a part of σ. Prove that the relation is a partial order on Πn (known as
the refinement order).
Solution: “Reflexivity”: let π ∈ Πn , then each part A ∈ π is contained in the part A
of π, so π ≤ π.
“Antisymmetry”: Suppose π ̸= σ are elements of Πn and π ≤ σ. Since they are not
equal, there must exists a part B of σ that is not a part of π. Pick any i ∈ B, and let A
be the part of π that contains i, then A must be contained in B, and the containment
is proper. If σ ≤ π, then B has to be contained in a part of π, which must be A by
considering i, but that is impossible as A ⊊ B.
“Transitivity”: Suppose π ≤ σ and σ ≤ µ. Let A be a part of π, then by the first
assumption, A is contained in some part B of σ, and by the second assumption, B is
contained in some part C of µ, so A is contained in the part C of µ. Since the choice
of A is arbitrary, π ≤ µ.

4. Continuing the convention above, let π, σ be two partitions in Πn . Define a relation

∼ on {1, 2, . . . , n} by declaring a ∼ b whenever there exists a sequence of elements
a = e0 , e1 , . . . , ek = b (k can be equal to 0) such that for each i = 1, . . . , k, either
ei−1 , ei are in the same part of π or in the same part of σ (or both). Show that
(1) ∼ is an equivalent relation, thus inducing a partition µ itself;
(2) µ ≥ π, σ; and
(3) whenever η ≥ π, σ, we must have µ ≤ η.
Solution: (1) “Reflexivity”: For any a ∈ {1, . . . , n}, the sequence a = e0 shows a ∼ a.
“Symmetry”: Suppose a ∼ b via the sequence a = e0 , e1 , . . . , ek = b, then the sequence
ek , , . . . , e0 also satisfies the condition, and shows that b ∼ a.
“Transitivity”: Suppose a ∼ b and b ∼ c via the sequences a = e0 , e1 , . . . , ek = b and
b = e′0 , e′1 , . . . , e′l = c, respectively. Then a = e0 , e1 , . . . .ek = e′0 , e′1 , . . . .e′l = c is a

2
sequence that shows a ∼ c.
(2) Let A be a part of π. Pick an arbitrary a ∈ A and let B be the part of µ that
contains a. Then for any b ∈ A, a, b is a sequence that shows a ∼ b thus b ∈ B. Since
b ∈ A is arbitrary, A ⊂ B; and since A ∈ π is arbitrary, π ≤ µ. Similarly, σ ≤ µ.
(3) Let η ≥ π, σ. Let M be a part of µ and pick a ∈ M . Let E be the part of η
that contains a, we claim that M ⊂ E, which is sufficient to show that µ ≤ η as M
is arbitrary. For any b ∈ M , there exists a sequence a = e0 , e1 , . . . , ek = b as in the
definition of ∼, we prove by induction that every ei in the sequence is in E. By the
choice of E, e0 = a ∈ E. Suppose ei ∈ E for some i, if ei , ei+1 are in the same part A
of π, then since π ≤ η, A is contained in a part of η, which must be E by considering
ei , so ei+1 ∈ A ⊂ E as well; the case that both ei , ei+1 are in the same part of σ is
analogous. This implies that b = ek ∈ E, and since b is arbitrary, we have M ⊂ E.