C8.

3 Combinatorics
Sheet 1 — MT21
Sheet 1

Before doing this sheet, you should make sure you are familiar with Sperner’s Lemma, the
LYM Inequality, and Dilworth’s Theorem.

Section A
You should do these questions, but not hand them in. Solutions can be found on the course
page.
1. Write down all antichains contained in P(1) and P(2). How many different antichains
are there in P(3)?

Solution: In P(1) the antichains are ∅, {∅}, {1}.

In P(2) the antichains are ∅, {∅}, {1}, {2}, {12}, {1, 2}.
Lastly, in P(3) there are one antichain of size 0 and 8 antichains of size 1 (given by
the elements of P(3)). Any antichain of order 2 cannot contain ∅ or 123. There are
3 antichains of size 2 in [3] and 3 of size 2 in [3]



1 2
and 3 of size two between the
layers. Any antichain of size 3 must be contained in (equal to) a layer. This gives
1 + 8 + 9 + 2 = 20 antichains.

n


2. (a) Look up Stirling’s Formula. Use it to find an asymptotic estimate for n/2
of the
form (1 + o(1))f (n) when n is even.
n


(b) Now do the same for pn where p ∈ (0, 1) is a constant and pn is an integer. Write
your answer in terms of the binary entropy function

H(p) = −p log p − (1 − p) log(1 − p).

Solution: (a) Stirling’s approximation says that, as n → ∞

√
n! ∼ 2πn nn e−n , (1)

i.e. the ratio between the left and right hand side of (??) converges to 1 as n → ∞.
Suppose first that n is even. Then (with a little working) Stirling gives us
√ r
2πn(n/e)n
 
n n!
2 n
= 2
= (1 + o(1)) p
2 = 1 + o(1) · 2 .
n/2 (n/2)! 2πn/2(n/2e) n/2 πn

Mathematical Institute, University of Oxford Page 1 of 3

Alex Scott: [email protected]
 C8.3 Combinatorics: Sheet 1 — MT21

p
So we can take f (n) = 2/πn 2n . Note that this tells us that the largest binomial
√
coefficient has size around 2n / n.
(b) Applying Striling’s formula to n!, (pn)! and ((1 − p)n)! gives
  √
n
n 2πn n/e
= (1 + o(1)) √
n p
(1−p)n
pn 2πpn n/e ppn 2π(1 − p)n n/e (1 − p)(1−p)n
2H(p)n
= (1 + o(1)) p .
2np(1 − p)
n

√
Note that for p constant and n large we have pn
= 2(1+o(1))H(p)n , as the n factor can
be absorbed by the o(1) term in the exponent.

Mathematical Institute, University of Oxford Page 2 of 3

Alex Scott: [email protected]
 C8.3 Combinatorics: Sheet 1 — MT21

Section B
These questions are to be handed in.

3. Let k ≤ n/2, and suppose that F is an antichain in P[n] such that every A ∈ F has
|A| ≤ k. Prove that |F| ≤ nk .

4. Let (P, ≤) be a poset. Suppose that every chain in P has at most k elements. Prove
that P can be written as the union of k antichains.

5. Suppose F ⊂ P[n] is a set system containing no chain with k + 1 sets.

(a) Prove that
n
X |Fi |
n ≤ k,


i=0 i

where Fi = F ∩ [n](i) for each i.

(b) What is the maximum possible size of such a system?

6. Let A be an antichain in P[n] that is not of the form [n](r) . Must there exist a maximal
chain disjoint from A?

7. Let (P, ≤) be an infinite poset. Must P contain an infinite chain or antichain?

Mathematical Institute, University of Oxford Page 3 of 3

Alex Scott: [email protected]
 C8.3 Combinatorics: Sheet 1 — MT21

Section C
These questions are optional. Solutions can be found on the course page (except for
questions marked with a (*), which lie outside the course but are there if you want an
extra challenge!).

8. Let f (n) be the number of subsets A ⊂ P(n) such that A is an antichain.

n
(a) Prove that f (n) ≥ 2(bn/2c) for every n.
n
(b) (Tricky) Prove that, for every  > 0, f (n) ≤ 22 for all sufficiently large n.

n
Solution: (a) Each of the 2(bn/2c) subsets of [n]


is an antichain.
bn/2c
n


(b) Here are two solutions. Take a partition of P[n] into bn/2c chains. If A is an
antichain it can contain at most one set from each chain, giving at most n+1 possibilities
for how A intersects the chain. The number of antichains is therefore at most
n n √
(n + 1)(bn/2c) = 2log2 (n+1)(bn/2c) = 2O(2 (log n)/ n) = 2o(n) .
n

If you want a further challenge, you could try proving that there is a constant C > 0
such that
n
f (n) ≤ C (bn/2c) ,

but this is a bit harder!

9. Consider the poset P = (P(N), ⊆) (in other words, the elements of P are sets of natural
numbers, and A ≤ B if A ⊆ B). Show that P contains an uncountable antichain.

Solution: There are many ways to do this: for example, consider all sets that contain
exactly one element from each of the pairs {1, 2}, {3, 4}, . . . .

10. (*) Does the poset in the last question contain an uncountable chain?

Mathematical Institute, University of Oxford Page 4 of 3

Alex Scott: [email protected]