3TA Topics in Algebra in 2021

The Axiom of Choice and Zorn’s Lemma

An optional, non-examinable section

An attempt to base mathematics on set theory and logic led to the Zermelo-Fraenkel
axioms for set theory, commonly called ZF. The idea is to write down obvious axioms
we use when working with sets, while carefully avoiding anything that would lead to
contradictions, like Russell’s Paradox. For example, one has:

• Two sets with the same elements are equal.

• The empty set exists.

• One can form {x} and {x, y}.

• One can form unions of sets.

• One can form power sets.

• One can form an infinite set / the natural numbers.

This is not a complete or precise list, by any means. There is a further axiom, called
the Axiom of Choice. (Beware abbreviations: AC is Axiom of Choice, ACC is Ascending
Chain Condition; not the same thing.) Here it is:

(AC) Given a non-empty family A=∪{Ai }i∈I of non-empty sets, there exists a choice func-
tion for A, that is, a map f : I → Ai such that f (i) ∈ Ai for all i ∈ I.
i∈I

Adding AC to ZF gives “Zermelo Fraenkel with Choice” commonly called ZFC.

Note: If the index set I is finite, one can use the other axioms of ZF to find the function
f ; the axiom AC is not needed. The axiom AC is only needed for infinite I.
Informally, AC says that if you have an infinite collection of non-empty sets, you can
make a new set by choosing one element from each of them.
The axiom AC cannot be proved from the other axioms of ZF.

Is AC useful?

Yes. Unbelievably so. Giving just one example, it can be used to show that every vector
space has a basis. (This does not only refer to finite-dimensional vector spaces.) Without
AC, a vector space might have no basis at all, or have two bases with different cardinali-
ties.
Our primary use for AC will be via Zorn’s Lemma (see later, and Chapter 10 of our
textbook.)
Why do we not just assume that AC is obvious and natural?

It has surprising and non-intuitive consequences. For example, informally speaking, one
can use AC to show that a ball can be cut up into finitely many pieces and then re-
assembled to make two balls of the same size as the original. (Look up the “Banach-Tarski
Paradox.”)

Could adding AC to ZF introduce contradictions?

In the 1930’s, Gödel proved that adding AC to ZF does not introduce any contradictions.
However, in the 1960’s, Cohen proved that adding the negation of AC to ZF does not
introduce any contradictions, either.

To summarize:
There are good reasons for assuming the Axiom of Choice.
There are good reasons for not assuming the Axiom of Choice.

Here is one person’s opinion on the matter (E. Schechter, 1997):

“The Axiom of Choice and its negation cannot coexist in one proof, but they can cer-
tainly coexist in one mind. It may be convenient to accept AC on some days - e.g. for
compactness arguments - and to accept some alternative reality on other days, e.g. for
thinking about complete metric spaces.”

In many areas of mathematics, it turns out to be very interesting to see what one can prove
constructively, meaning without AC. Speaking for myself (me being Anneliese Schauerte),
my favourite such area is pointfree topology, where I do my own research. It is a particu-
larly lattice-theoretic approach to topology. (This may seem like a contradiction to you,
but it really isn’t.) Here AC and various weaker forms of AC arise naturally. So if one
simply assumed AC, one would be missing out on seeing the real, nuanced picture.

Whatever you end up thinking about AC, you should not accept or reject it without
understanding it thoroughly first. In the context of the work of 3TA, the most natural
version of it to concentrate on is Zorn’s Lemma, which we discuss next. The terminology
is unfortunate: it should maybe be called Zorn’s Axiom, but it was originally derived in
ZFC, so was a result. It has been called Zorn’s Lemma for so long, we can’t change it now.
Zorn’s Lemma

Zorn’s Lemma (ZL) is equivalent to the Axiom of Choice. This means ZF + AC implies
Zorn’s Lemma and ZF + Zorn’s Lemma implies AC. Here are various versions of ZL:

1. (ZL) Let P be a non-empty ordered set in which every non-empty chain has an
upper bound. Then P has a maximal element.

2. (WZL) Let P be a non-empty ordered set in which every non-empty chain has a
join. Then P has a maximal element.

3. (WWZL) Let P be a non-empty ordered set in which every chain has a join. Then
P has a maximal element.

You can see that what ZL is good for, is finding maximal elements in an ordered set. As
such, it fits very naturally into the material covered by this course. It is, however, used
very extensively in other areas of mathematics, like topology, analysis and algebra also.

Activities for the interested 3TA student

Well, there is a lot you could do; this is a massive and interesting area of mathematics.
However, as far as 3TA goes, what you could most usefully do is learn how to actually
apply Zorn’s Lemma. I have attached some tutorial-style questions for you, to help you
do exactly that.

For further reading, I recommend:

• Chapter 10 of our textbook, sections 10. 1 to 10.6. The rest of that chapter will
make a lot more sense after we have done Chapter 5. (We will do Chapter 5 as a
standard part of 3TA.)

• Have a look at Horst Herrlich’s book “Axiom of Choice.” Springer Lecture Notes in
Mathematics, No. 1876. Published by Springer in 2006. It is a beautifully written
and entertaining book, but intended for working mathematicians, so demanding for
under-graduates. (I got the Schechter quote above from this book.) Here are some
chapter headings: Hidden Choice, Unnecessary Choice, Disasters without Choice,
Disasters with Choice, Disasters either way, Beauty without Choice.
Optional tutorial questions on Zorn’s Lemma

1. As usual, let Σ∗∗ denote the set of finite and infinite strings of 0’s and 1’s, ordered
by u ≤ v iff u = v or u is a finite initial substring of v. Apply Zorn’s Lemma to
show that Σ∗∗ has a maximal element.
[You don’t actually need Zorn’s Lemma to find these maximal elements, of course.
I’m sure it’s perfectly obvious to you what they are. I just wanted to begin with
one really easy application of Zorn’s Lemma, so you get the hang of it.]

2. Prove (WWZL) =⇒ (WZL). (This is straightforward, because P and 1 ⊕ P have

the same maximal elements.)

3. In this question, we will work through the proof of the implication (WZL) =⇒ (ZL).

Theorem (WZL) =⇒ (ZL)

Proof:
Let P be a non-empty ordered set in which every non-empty chain has an upper
bound. We must show that P has a maximal element. (1)
Let A be the set of all non-empty chains in P , ordered by inclusion. (2)
Then A is not empty. (3)
Claim: every non-empty chain in A has a join. Let’s check this. (4)
Take {Xi∪}i∈I a chain in A (so in particular each Xi is a chain in P ). (5)
Let Z = Xi . (6)
i∈I
Then Z is also a chain. (7)
So Z is the join of {Xi }i∈I in A. (8)
We have now proved the claim that every non-empty chain in A has a join. (9)
Apply (WZL) to A to conclude that A has a maximal element, which we’ll denote
X0 . (10)
This means that X0 is a maximal (non-empty) chain of P . Let m be an upper
bound for X0 in P . (11)
Claim: m is a maximal element of P . Let’s check this: (12)
If x ∈ P and m ≤ x then X0 ∪ {x} is a chain. (13)
So X0 ∪ {x} = X0 by maximality of X0 . (14)
Thus x ∈ X0 and x ≤ m. Hence x = m. (15)
This concludes the proof of the claim that m is a maximal element of P . (16)

(a) Show that the hypotheses of (ZL) and (WZL) are different; i.e. give an example
of an ordered set P in which it is true that every non-empty chain has an upper
bound but false that every non-empty chain has a join.
(b) How do you know A is non-empty, as claimed in line (3)?
(c) Prove that Z is a chain, as claimed in line (7).
(d) How do you know that the m mentioned in line (11) exists?
(e) Prove that X0 ∪ {x} is a chain, as claimed in line (13).
4. In this question, we will prove Szpilrajn’s Theorem, which states that every order
has a linear extension. To say that (P, ≤) has a linear extension means that there ex-
ists an order ≼ on P such that (P, ≼) is a chain and, for all a, b ∈ P , a ≤ b =⇒ a ≼ b.

We already showed this for finite ordered sets in Tutorial 1: it is Exercise 1.29 (ii)
on p32 of the textbook. We now prove the version for infinite sets, using (ZL).
Proceed as follows.

Suppose (P, ≤) is an ordered set. In Tutorial 1, you checked that, if s∥t in (P, ≤),
there exists an order ≤1 on P such that s ≤1 t and, for all a, b ∈ P , a ≤ b =⇒ a ≤1 b.
We will take this as done; as you will see, we will need it here.

Suppose that (P, ≤) is an ordered set.

(a) Let E be the set of all orders ≼ on P such that, for all a, b ∈ P , a ≤ b =⇒ a ≼ b.
How do you know that E is non-empty?
(b) If you regard an order on P as a subset of P × P , you can think of E as a
collection of subsets of P × P ordered by inclusion. Take a non-empty chain
in E. Show that the union of the members of the chain forms an upper bound
for the chain, as required. (Quite a few things need to be checked here; make
sure you don’t omit any.)
(c) Apply (ZL) to get a maximal element ≼0 of E. Show that (P, ≼0 ) is the desired
linear extension of (P, ≤).