Enumerative combinatorics and posets:

Assignment # 2

Felipe Bedoya

May 4, 2020

Problem 1. Let Bn (q) be the set of subspaces of the vector space Fnq .
a Find a product formula for the cardinality of
Wn,k (q) = {V ∈ Bn (q) : dim(V ) = k}

b If we consider the formula before as a q-analog of a number. What would this number be?
Solution We want to know how many subspaces there are with dimension equal to k. We are
going to answer the question by showing that this is key in knowing the number of k independent
vectors we can have. Recall that the vector spaces is a n-product of a field Fq which has, as
expected, q elements, this means that the entire space has q n . Now, take any vector you want and
set it as v1 , i.e, the first generator of the given subspace. Notice that you have q n − 1 options.1
Let’s us think of how many possible choices we have for the second generator. Well, we have q n
elements from which q are generated by v1 . Thus, there are q n − q options. For the third position
we can have q n − q 2 since each of the q vectors generated by v1 can generate another q vectors.
Continuing with the process we have that such total number of selecting generators is:

k−1
Y
(q n − q i ) (1)
i=0

We can have the same amount by taking any given subspace of dimension equal to k and study all
possible sets of linearly independent vector. The process is actually the same, first we have q k − 1
possible choices, for our second option we have q k − q since multiples of the the first vector and so
on. According to the former argument, we have that:

Qk−1
(q n − q i )
W (n, k) = Qi=0
k−1
(2)
k i
i=0 (q − q )
1
If you take the zero vector then you have the trivial subspace.

1
Now factor as that expression and rewrite it as

(q k−1 )(q n − 1)(q n−1 − 1) · · · (q n−k+1 − 1)

(3)
(q k−1 )(q k − 1)(q k−1 − 1) · · · (q − 1)

That power of q can be eliminated, you can also multiply by (q − 1)k up and down and you would
i −1
get [i]q = qq−1 for each 1 ≤ i ≤ n.
Hence, the total number is

[n][n − 1] · · · [n − k + 1]
= [ nk ]q (4)
[k][k − 1] · · · [1]

Problem 2. Let λ = (λ1 , λ2,...,λk ) an integer partition and define Par as the number of partitions
that fit inside nm ` nm. We will proceed to show some related properties of

X
FParmn ,Boxes (q) = q Boxes(λ) (5)
λ∈Parmn

Solution We will describe the necessary insights for the statistic and we will answer all four
items in a mixed way. Firs, it will be shown that FParmn ,Boxes (q) = FParnm ,Boxes (q). This follows
immediately from a simple bijection. Suppose, as an example, the partition λ = (32 ) which can
be though as a 3 by 2 rectangle, and take the partition α = (3, 1) which clearly fits in λ. If you
rotate the young diagram 90 degrees and organize the blocks clockwise you would get λ0 = (23 )
and α0 = (2, 1, 1) which does not fit in λ but it does in λ0 . Notice that here we are just describing
the conjugate mapping between partitions, which is clearly a bijection.
Next, we want to show that

X
q |λ| = [ n+m
n ]q (6)
λ∈P armn

The idea is to proceed by induction on m (it would function perfectly well for n as well) for m =
0 we just have the empty partition. If m > 0, we consider the relation between the m and the
m-1 case. Consider λ ⊂ (mn ) where λ = (λ1 , λ2 , . . . , λk ) where λ1 is the greatest number. Either
λ1 < m and λ1 = m. For the former, λ ⊂ ((m − 1)n ). For the latter, the partition λ is equal
to (m, λ2 , . . . , λk ) where the rest of the partition can be denoted as λ0 . Since λ0 ⊂ (mn−1 ). Since
|λ| = |λ0 | + m we have that

2
 0
X X X
q |λ| = q |λ| + q |λ |+m (7)
λ∈P armn λ∈P ar(m−1)n λ∈P arm(n−1)

By applying the induction hypothesis,

X
q |λ| = + q m [ n+m−1
 n+m−1 
n−1 q n ]q = [ n+m
n ]q
λ∈P armn

The last equality hold because of the recurrence of the q-binomial. Notice that with this we answer
c) and d) since the asked recursion was key in the proof and we explained why that recursion holds.
For the special case q = 1 we can offer a more detailed insight. P It is clear that this is just the
q-analogue of the former expression and that FP arm×n ,boxes (1) = λ∈P arm×n 1boxes(λ) = |P arm×n |.
Thus, we aim at counting the number of partitions that fit inside (nm ). In that spirit, notice
that any partition λ ∈ P arm×n can be thought as a lattice path from (0, 0) to (m, n). Intuitively,
if you start at the south-west corner, where λ = (λ1 , λ2 , . . . , λk ) ∈ P arm×n , it follows that the
corresponding path is
N m−k E λk N E λk−1 −λk N E λk−2 −λk−1 N · · · E λ1 −λ2 N E n−λ1

The previous map is a way to code through words a given partition. Consider m = 4 and n = 2 and
the partition λ = (2, 2, 1). This can be mapped as N 1 E 1 N 1 E 1 N 1 E 0 N 1 which means graphically
is follow the border of the tabloid. Reciprocally, any path from (0, 0) to
(m, n) determine a unique
m+n
partition λ ∈ P arm×n given that it is well-known that there are n distinct paths.
With these reasoning we answer all items.

Problem 3. We want to show that

n n
k+1
q ( 2 ) [ nk ]q xk
Y X
(1 + xq j ) = (8)
j=1 k=1

Solution There is a proof by induction which is as follows. For n = 1 we get 1 + qt in both sides.
Also, by using the recursion of the q-binomial, we have

n n n
k+1 k+1 k+1
q ( 2 ) [ nk ]q xk = q ( 2 ) [ n−1 q( )+n−k [ n ] xk
X X X
k
k ]q x +
2
k q
1=0 k=1 k=1
n n−1 n n−1 n−1
k+1 k 
q ( 2 ) [ nk ]q xk = q (2) n−1
X Y X Y Y
j n
 k−1 j n
(1 + xq ) + q x k−1 q x = (1 + xq ) + q x (1 + xq j )
1=0 j=1 k=1 j=1 j=1

3
 P boxes(λ)
A combinatorial argument would be as follows: Let pk (q) = λq where P
the sum goes
through all partitions with k distinct parts each of size at most n and let Fn (x, q) = nk=0 pk (q)xk .
If we compute that number in two different ways we therefore show the relation between these
families of objects.
First, recall for the partitions in hand, it must be true that there is at most one part of size i for
1 ≤ j ≤ n. On the other hand, the term (1 + xq j ) shows whether or not the partition has a part
of size i. Thus, it follows that the left hand side of equation is actually Fn (x, q).
Finally, notice that in any such partition with exactly k parts we can fit inside the partition
(k, k − 1, . . . 2, 1), this is true there are k different parts where the smaller one is greater or equal to
Pk k(k+1) k+1


one. Such fitting partition has m=1 m = 2 = 2 boxes. It is also true that if we remove
the original partition, we are left with one that fits inside a box with dimension k × (n − k). By
the former question we know that it is counted by the q-binomial. Apply multiplicative principle
and we are done.

Problem 4. How many ways are there of arranging 16 football teams into four groups of four?
Solution ince we are not assuming any order the model can be represented by the partition 44 of
[16].Therefore, there are 16!
4!5
different options.

Problem 5. In the early versions of the Enigma machine, used in Germany in the 1930’s, the
plugboard swapped six pairs of distinct letters of the alphabet. In how many ways can this be
done (assuming 26 letters)?
26!
Solution As in the former question, we do it by using a 26 , 14 partition. Hence there hare 2!6 6!14!
combinations. This is just the size of the class given by the referenced partition.

Problem 6. Find the multiplicative expressions for:

(a) The chain Cn
(b) The Boolean algebra Bn
(c) The divisibility poset D12
(d) The divisibility poset D60
Solution (a) The chain Cn . There is just one element per level. Therefore, we have

n−1
X 1
F (P, t) = ti = (9)
i=0
1 − tn

(b) For the Boolean algebra Bn we have

4
 n  
n
X n i
F (P, t) = (1 + t) = t (10)
i=0
i

Since the elements with statistic i are just subsets of size i, the find out how many there are,
we just use nk numbers.

(c) The answer is the result of a straightforward calculation, and it is

(t + 1)(t2 + t + 1) (11)

(d) The result is

(t + 1)2 (t2 + t + 1) (12)

Problem 6. For a sequence of numbers {an }n≥0 we define its exponential generating function
(e.g.f.) to be the formal power series
∞
X xn
A(x) = an
n=0
n!
n
B(n) xn! .
P
Let B(n) be the n-th Bell number and F (x) = n≥0

(a) Use the recursion for the Bell numbers, B(0) = 1 and
X n − 1
B(n) = B(k), n≥1
k≥0
k

to prove that
dF (x)
= ex F (x).
dx

(b) Use the differential equation above to prove that F (x) is given by
x −1
F (x) = ee .

(c) Finally expand the formula in part (B) to give the explicit closed formula
1 X jn
B(n) =
e j≥0 j!

5
Solution (a) Observe that
X xk
F (x) = B(k)
k≥0
k!

Hence,
dF (x) x2 xn
= B(1) + B(2)x + B(3) + ... + B(n + 1) + ... (13)
dx 2! n
Now, we proceed to apply the recursion for the case n + 1, which leads to
dF (x) X xn
= B(n + 1)
dx n≥0
n!
!
X X n xn
= n B(i) .
n≥0 i=0
i n!

Now, we would like to show that F 0 (x) = ex F (x). Using the series expansion of the e, and the
multiplication rule of formal power series, it is true that
n  
! ! !
X X n xn X xn X xn
B(i) = B(n)
n≥0 i=0
i n! n≥0
n! n≥0
n!

Recall that when dealing with series equality, we must show the are alike term by term, i.e,
comparing the coefficient [xk ]. In the left-hand-side
k  
k
X 1 k
[x ] = B(k) (14)
i=0
k! i

In the right-hand-side the coefficients are given by

k
k
X 1
[x ] = B(k − i). (15)
i=0
i!(k − i)!

k


Note that by expanding i
the term k! vanishes in 14, obtaining the sum in 15. Hence,

dF (x)
= ex F (x).
dx

(b) This differential equation is homogeneous. It is solved by separation of variables in the following
manner
Z Z
1
ln F (x) = = ex dx = ex + c.
F (x)

6
We apply the exponential function to both sides arriving to
x +c
F (x) = ee

We determine c by using that B(0) = 1, i.e, using the initial value of the recursion
0 +c
1 = F (0) = ee ,

so c = −1. Hence, the exponential generating function of the Bell numbers is

x −1
F (x) = ee

as we wanted to show.
(c) By using the exponential expansion twice, it follows that
∞ ∞ ∞
ex −1 1 X ejx 1 X 1 X (xj)n
e = = ( ) (16)
e j=0 j! e j=0 j! n=0 n!

In order to obtain the result we compare the coefficient [xn ] in the above expression and in
X xn
F (x) = B(n) ,
n≥0
n!

In the former, the coefficient is given by

∞
n 1 X jn
[x ]F (x) =
e j=0 j!n!

whilst the latter,

B(n)
[xn ]F (x) =
n!
By merely equalizing terms it is readily seen that
∞
1 X jn
B(n) = .
e j=0 j!

Problem 7. Prove that F (P × Q, t) = F (P, t)F (Q, t)

Solution The resulting algebraic proof would be trivial once we clarify a couple of details. First,
let us show that if (x, y) l (x0 , y 0 ) then either x = x0 and y l y 0 or y = y 0 and x l x0 . This
holds because if we let x l x0 and y l y 0 , then it is true that (x, y) < (x, y 0 ) < (x0 , y 0 ) and
(x, y) < (x0 , y) < (x0 , y 0 ) given the order relation of a product.

7
With base on that we can show that y P and Q are graded, so is P × Q. Recall that in the latter,
the minimal are ordered pairs of elements from M in(P ) and M in(Q). The case for the maximal
is analogue. If we have (x1 , y1 ) l (x2 , y2 ) l · · · (xk−1 , yk−1 ) l (xk , yk ) then we have that

x1 l xi1 l · · · l xk
y1 l xj1 l · · · l yk
T
According the first statement we proved, it holds that {i1, i2, . . .} {j1, j2, . . .} = ∅ and its union
is equal to the indices until k − 1. This shows that the product is indeed graded since all maximal
chains have the same size. Finally, we define a graded function ρ(x, y) = ρ(x) + ρ(y). Notice that
it satisfies that ρ(x, y) = 0 if (x, y) ∈ M in(P × Q) and ρ(x0 , y 0 ) = ρ(x, y) + 1 if (x, y) l (x0 , y 0 )
according the the discussion of the beginning of the current solution.
Finally, we have that:

X
F (P × Q) = tρP ×Q ((x,y))
(x,y)∈P ×Q
X
= tρP (x)+ρQ (y)
(x,y)∈P ×Q
X
= tρP (x) tρQ (y)
(x,y)∈P ×Q
!
X X
= tρP (x) tρQ (y)
x∈P y∈Q
! !
X X
= tρP (x) tρQ (y)
x∈P y∈Q

= F (P, t) · F (Q, t)

Problem 8. Show that for any posets A and B we have that

Int(A × B) ∼
= Int(A × B∗) (17)

Solution It is true that for any elements (a1 , b1 ) ≤ (a2 , b2 ) of A × B holds

[(a1 , b1 ), (a2 , b2 )] = {(x, y) ∈ A × B : a1 ≤A x ≤A a2 , & b1 ≤B y ≤B b2 }

[(a1 , b1 ), (a2 , b2 )] = {(x, y) ∈ A × B ∗ : a1 ≤A x ≤A a2 & b2 ≤B ∗ y ≤B ∗ b1 }
[(a1 , b1 ), (a2 , b2 )] = [(a1 , b2 ), (a2 , b1 )]∗

8
Notice the former is an interval of A × B whilst the latter is an interval in A × B ∗ . Equivalently,
Int(A × B) = Int(A × B ∗ ).
Thus, the map ϕ : Int(A×B) −→ Int(A×B ∗ ) defined by ϕ([(a1 , b1 ), (a2 , b2 )] = [(a1 , b2 ), (a2 , b1 )]∗ is
noting else than the identity map, and therefore is clearly an which is clearly monotonic. Therefore
ϕ is an isomorphism and 17) holds.

Problem 9. Let G be a connected graph on n vertices such that G induces a subposet ΠG of Πn .

Show that ΠG ∼
= Bn−1 .
Solution The graph G is connected, then for every vertex there at least is an edge that joins
it together to another vertex (the vertices are the elements of [n]). If we let G = T where T
is a tree then we have a connected but acyclic graph, which means that if we remove an edge
then the graph is separated into two connected components, removing two vertices would lead to
three connected components and so on. Let’s analyze this process in more depth. The key to
the proposed bijection is to remember that the tree has n − 1 vertices and that we can remove
from 1 to the entire set of n − 1 edges without deleting the vertices. Remove the edge joining
i, j together. As a result, we have two main connected components, denoted as Gi and Gj . It is
clear that [n] = Gi t Gj . This partition of [n] is clearly unique. If we remove i, j, k then we have
the partition [n] = Gi t Gj t Gk . In the most extreme case (removing all edges), the connected
components are just {i} for 1 ≤ i ≤ n. The edges of partitions to be taken away are in bijection
to the numbers of subsets of [n − 1]. Hence, ΠT ∼ = Bn−1 .
Problem 10. Show that a finite locally connected poset with a 0̂ and 1̂ is graded.
Solution Consider a locally connected poset P with 0̂ and 1̂. Let us show that it is graded.
Suppose not, thus, there are two maximal chains with different length. Since P possess 0̂ and 1̂,
this implies that there is a non-empty interval that is not graded (if all intervals were graded, then
by connecting with 0̂ and 1̂ we would get a graded poset). There could be several, in such a case,
grab a minimal one ordered by inclusion, we denote it [x, y]. By hypothesis, P is locally connected
and since the chosen interval is not graded, (x, y) cannot be a singleton {z} or a chain, so there are
at least two maximal chains with different length, call then {z1 , . . . , zn } and {w1 , . . . , wm }. If we
have the case {z} and {w1 , . . . , wm }, it must be true that z is comparable to some wi . Otherwise,
(x, y) would not be connected. Assume z < wi . It is clear that in such a case we have found a new
and smaller non-graded interval.
Else, (x, y) must be connected, where we study to different chains. Notice that there should exists
zi , wj such that either zi < wj or zi > wj . If not, there would not be a way to go from elements
of the chain w to elements of the chain z. In that case we contradict the locally connectedness
hypothesis since we already analyzed the anticahins scenario. This implies that either [x, zi ] or
[x, wj ] is a smaller non-graded interval, which contradicts the fact that [x, y] is minimal by inclusion.

Problem 11. Find the number of ideals an given Pn = [n] [n0 ]

F

Solution e are going to proceed by recursion. Consider n = n0 = 1. In this case there are
three ideals ordered by inclusion, ∅, [1], [1, 10 ]. Now, let us see what happens if we consider n =

9
n0 = 2. All the ideals in the first case are ideals for the current one. The new elements are
[20 , 2], [10 , 2, 1], [20 , 2, 10 , 1], [20 10 1]. Consider Cn−1 as the number of ideals for the case n − 1. What
would happen if we add the nth element? Notice that the generated by n, together with any subset
of {10 , 20 , . . . , n0 } is and ideal since the elements of the latter are anti-chains. Hence we have 2n
new elements. It is also true that whenever we consider to include n0 , we must include the entire
set [n] but these are already counted by the term 2n . Hence, there no new ideals other than the
counted by this term.
Thus, the proposed recursion for the initial condition C1 = 3 is

Cn = 2n + Cn−1 (18)

Where Cn = 2n+1 − 1 is the value that solves it.

Problem 12. Determine the structure of J(P ) when P is a chain, an anti chain and a product of
chains
Solution hen P is an antichain, we know J(P ) is the Boolean algebra for the elements of the set
since no element covers any other. Hence, any subset of such elements would satisfy the definition
of ideal, which menas that the set of ideals is B|P | . For Cn , we have that J(P ) looks like Cn+1
since if ci belongs to any ideal, so does the the set {c1 , . . . , ci }. Hence, all the elements of J(P )
are ∅, {c1 }, {c1 , c2 }, . . . , Cn , having n + 1 elements in total, organized as chain by inclusion, since
the the generated of i + 1 covers the generated of i we have that the ideals of Cn is isomorphic to
Cn+1
For the product of chains Cn × Cm think of a m × n arrangement where the nodes (i, j) with
1 ≤ i ≤ m and 1 ≤ j ≤ n are elements of the product. Since the generated by any (i, j) are all
(k, l) with 1 ≤ k ≤ i and 1 ≤ l ≤ j we can count the ideals as the number of Young tabloids that
fit into the rectangle m × n since that collection of generators determines a unique partition. For
example, consider m = 3 and n = 4 (3 by 4 rectangle) and Cm × Cn , the generated by (4, 2)) and
(2, 3) looks like the partition (4, 4, 2) upside-down. Such number is given by what was explained
throughout the second problem.

10