ZFC Set Theory

You have been using sets as if they were unproblematic ever since the introduction to sets: write some elements inside curly braces, form subsets, take unions. For everyday mathematics, this works fine. But it breaks down the moment you ask a deceptively simple question: can a set be an element of itself?

Answering that question with unrestricted set-formation leads directly to a contradiction — not a subtle one, but a one-line logical explosion. Zermelo–Fraenkel set theory with the Axiom of Choice, universally abbreviated ZFC, is the response mathematics settled on. Today ZFC is the standard foundation on which virtually all of mathematics rests.

The trouble with naïve set theory

In naïve set theory, a set is any collection you can describe by a property. For any predicate $\varphi(x)$, you can freely form

$$\{x \mid \varphi(x)\}.$$

This unlimited freedom is called the unrestricted comprehension principle. It is inconsistent.

Russell’s paradox (1901) is the simplest proof of this. Define

$$R \;\coloneqq\; \{x \mid x \notin x\}.$$

$R$ is supposed to collect every set that is not a member of itself. Now ask: is $R \in R$?

• If $R \in R$, then $R$ satisfies the defining condition $x \notin x$, so $R \notin R$. Contradiction.
• If $R \notin R$, then $R$ does not satisfy $x \notin x$, which means $R \in R$. Contradiction.

There is no consistent answer. $R$ cannot exist, which means unrestricted comprehension is broken.

The axiomatic approach

The fix is to replace unlimited freedom with a small, carefully chosen list of axioms — basic statements declared true — and derive everything else from them. Axioms must be:

• Consistent: they must not lead to contradictions (ZFC is believed to be consistent, though this cannot be proved inside ZFC itself — Gödel’s incompleteness theorems forbid it).
• Strong enough: sufficient to reconstruct all of ordinary mathematics.

ZFC consists of nine axioms from Zermelo and Fraenkel — two of which are axiom schemas, meaning each stands for an infinite family of axioms, one per predicate — plus the Axiom of Choice.

The nine ZF axioms

Extensionality

Two sets are equal when they have exactly the same elements.

$$\forall A\;\forall B\;\bigl(\forall x\;(x \in A \leftrightarrow x \in B)\bigr) \;\Rightarrow\; A = B$$

You already know this principle from the set introduction: $\{1,2,3\} = \{3,1,2\}$ because both contain exactly $1$, $2$, and $3$, regardless of order. Extensionality makes this the definition of equality — a set is fully determined by its members, and by nothing else.

Empty set

There exists a set with no elements.

$$\exists A\;\forall x\;(x \notin A)$$

Extensionality guarantees there is exactly one such set, which you denote $\emptyset$. Without this axiom, you could not even prove any set exists at all.

Pairing

For any two sets $a$ and $b$, there exists a set whose only elements are $a$ and $b$.

$$\forall a\;\forall b\;\exists A\;\forall x\;\bigl(x \in A \;\leftrightarrow\; x = a \;\lor\; x = b\bigr)$$

This justifies writing unordered pairs $\{a, b\}$. As a special case, taking $a = b$ gives singletons $\{a\}$.

Pairing also lets you define ordered pairs without any new primitive. The Kuratowski encoding is:

$$(a, b) \;\coloneqq\; \bigl\{\{a\},\;\{a, b\}\bigr\}.$$

You can verify that $(a, b) = (c, d)$ if and only if $a = c$ and $b = d$, so the encoding correctly captures the notion of order. From ordered pairs you can build relations, functions, and Cartesian products — all inside ZFC.

Union

For any set $\mathcal{F}$ of sets, there exists a set containing exactly the elements of the elements of $\mathcal{F}$.

$$\forall \mathcal{F}\;\exists A\;\forall x\;\bigl(x \in A \;\leftrightarrow\; \exists F \in \mathcal{F},\; x \in F\bigr)$$

The resulting set is the union $\bigcup \mathcal{F}$. When $\mathcal{F} = \{A, B\}$, this recovers the familiar $A \cup B$. The general form handles infinite families: $\bigcup \{A_1, A_2, A_3, \ldots\} = A_1 \cup A_2 \cup A_3 \cup \cdots$.

Power set

For any set $a$, there exists the set of all subsets of $a$.

$$\forall a\;\exists A\;\forall x\;\bigl(x \in A \;\leftrightarrow\; x \subseteq a\bigr)$$

This is precisely the power set $\mathcal{P}(a)$. The power set axiom is what allows you to keep climbing to larger infinite sets: starting from $\mathbb{N}$, you can form $\mathcal{P}(\mathbb{N})$, then $\mathcal{P}(\mathcal{P}(\mathbb{N}))$, and so on — each strictly larger than the last by Cantor’s theorem.

Separation (schema)

Given any set $a$ and any predicate $\varphi$, the subset of $a$ whose elements satisfy $\varphi$ exists.

$$\forall a\;\exists A\;\forall x\;\bigl(x \in A \;\leftrightarrow\; x \in a \;\land\; \varphi(x)\bigr)$$

This is a schema: there is one axiom for every predicate $\varphi$ in the language of set theory, giving infinitely many axioms at once.

Separation is the safe replacement for unrestricted comprehension. You can form $\{x \in a \mid \varphi(x)\}$, but you must start from an already-existing set $a$. This blocks Russell’s paradox: you can form $\{x \in a \mid x \notin x\}$ for any fixed set $a$, but there is no “set of all sets” to use as the starting point. The contradiction evaporates.

Set-builder notation $\{x \in S \mid P(x)\}$ from the introduction to sets is justified precisely by this axiom.

Infinity

There exists a set containing $\emptyset$ and closed under the operation $x \mapsto x \cup \{x\}$.

$$\exists A\;\Bigl(\emptyset \in A \;\land\; \forall x \in A,\; x \cup \{x\} \in A\Bigr)$$

This is the axiom that guarantees infinite sets exist. The set $A$ must contain the following chain of sets:

$$\emptyset,\quad \{\emptyset\},\quad \bigl\{\emptyset, \{\emptyset\}\bigr\},\quad \bigl\{\emptyset, \{\emptyset\}, \{\emptyset,\{\emptyset\}\}\bigr\},\quad \ldots$$

These are the von Neumann natural numbers: identify

$$0 \coloneqq \emptyset, \qquad 1 \coloneqq \{0\}, \qquad 2 \coloneqq \{0,1\}, \qquad 3 \coloneqq \{0,1,2\}, \qquad \ldots$$

so that every natural number is the set of all smaller natural numbers. The axiom of infinity is what lets you prove $\mathbb{N}$ exists as a completed set inside ZFC, rather than just as an informal notion.

Replacement (schema)

If $\varphi(x, y)$ defines a function on a set $a$ (each $x \in a$ corresponds to a unique $y$), then the image of $a$ under $\varphi$ is a set.

$$\forall a\;\Bigl[\bigl(\forall x \in a\;\exists!\, y\;\varphi(x,y)\bigr) \;\Rightarrow\; \exists B\;\forall y\;\bigl(y \in B \;\leftrightarrow\; \exists x \in a,\;\varphi(x,y)\bigr)\Bigr]$$

Like Separation, Replacement is a schema — one instance per predicate $\varphi$.

Replacement says: the image of any set under a definable function is itself a set. This lets you build sets like $\{\omega, \omega+1, \omega+2, \ldots\}$ (where $\omega$ is the first infinite ordinal) that lie beyond the reach of the earlier axioms. Separation can only cut sets down; Replacement can map them to entirely new objects.

Regularity (Foundation)

Every non-empty set contains an element disjoint from it.

$$\forall A\;\bigl(A \neq \emptyset \;\Rightarrow\; \exists x \in A,\; x \cap A = \emptyset\bigr)$$

Regularity rules out circular membership. If a set $a$ contained itself — $a \in a$ — then the singleton $\{a\}$ would violate regularity: its only element is $a$, but $a \cap \{a\} = \{a\} \neq \emptyset$. More generally, there can be no infinite chain $a_0 \ni a_1 \ni a_2 \ni \cdots$ descending forever.

In practice, you will almost never encounter sets that would violate regularity. Its main purpose is to tame the set-theoretic universe, making it amenable to proofs by well-founded induction on the membership relation $\in$.

The Axiom of Choice

The nine ZF axioms above form the system ZF. Adding the Axiom of Choice (AC) gives the full system ZFC.

For any family $\mathcal{F}$ of non-empty sets, there exists a function that picks one element from each member of $\mathcal{F}$.

$$\forall \mathcal{F}\;\Bigl[ \bigl(\forall F \in \mathcal{F},\; F \neq \emptyset\bigr) \;\Rightarrow\; \exists f\colon\mathcal{F} \to \bigcup\mathcal{F},\;\; \forall F \in \mathcal{F},\; f(F) \in F \Bigr]$$

The function $f$ is called a choice function for $\mathcal{F}$.

For finite families, choosing one element from each set is trivial — you just do it one by one in finitely many steps. The axiom is needed for infinite families, where you cannot carry out infinitely many choices without a systematic rule. AC asserts such a function exists even when no explicit rule is available.

Why AC stands apart

Kurt Gödel proved in 1938 that AC is consistent with ZF: you cannot derive a contradiction from ZF + AC as long as ZF itself is consistent. Paul Cohen proved in 1963 that AC is independent of ZF: you cannot prove AC from ZF alone either. Together, these results show that AC is a genuine additional assumption, not a consequence of the other nine axioms.

This independence is why mathematics texts sometimes mark theorems that require AC with a special symbol, and why some mathematicians deliberately work in ZF without AC to see how much breaks.

Equivalents of AC

Within ZF, many principles turn out to be exactly as strong as AC — each implies and is implied by AC:

Principle	Informal statement
Well-ordering theorem	Every set can be well-ordered
Zorn’s lemma	A partial order in which every chain has an upper bound has a maximal element
Tychonoff’s theorem	Any product of compact topological spaces is compact
Every vector space has a basis	Every vector space (including infinite-dimensional ones) admits a Hamel basis

You already saw the well-ordering theorem in the well-order checkpoint. Of these equivalents, Zorn’s lemma is the one you will encounter most often in algebra and analysis, even though its connection to “making choices” is less immediate on the surface.

What ZFC constructs

Starting from the ten axioms and $\emptyset$, you can build all of standard mathematics in a systematic sequence:

1. Natural numbers — $\mathbb{N}$ via the von Neumann encoding from the Axiom of Infinity.
2. Integers — $\mathbb{Z}$ as equivalence classes of pairs $(a, b) \in \mathbb{N}^2$ representing $a - b$.
3. Rationals — $\mathbb{Q}$ as equivalence classes of pairs $(p, q) \in \mathbb{Z} \times (\mathbb{Z} \setminus \{0\})$ representing $p/q$.
4. Reals — $\mathbb{R}$ via Dedekind cuts (subsets of $\mathbb{Q}$) or equivalence classes of Cauchy sequences.
5. Complex numbers — $\mathbb{C}$ as pairs of reals with the appropriate operations.
6. Functions and relations — as sets of ordered Kuratowski pairs.
7. Ordinals — as canonical well-ordered sets representing every possible order type.
8. Cardinals — measuring the “size” of infinite sets, forming the hierarchy $\aleph_0 < \aleph_1 < \aleph_2 < \cdots$

Every mathematical object you are likely to encounter in analysis, algebra, topology, or computer science can be reduced — in principle — to a pure set built from $\emptyset$ using the ZFC axioms.

Summary

• Naïve set theory is inconsistent: unrestricted comprehension leads to Russell’s paradox ($R = \{x \mid x \notin x\}$ yields $R \in R \leftrightarrow R \notin R$).
• ZFC replaces unlimited freedom with ten explicit axioms: nine from Zermelo–Fraenkel plus the Axiom of Choice.
• Extensionality defines set equality by membership; Empty Set guarantees $\emptyset$ exists; Pairing builds $\{a, b\}$ and Kuratowski ordered pairs.
• Union collects elements of a family of sets; Power Set produces $\mathcal{P}(a)$ and enables climbing to larger infinities.
• Separation (schema) is the safe replacement for unrestricted comprehension — you can only carve subsets out of an existing set, blocking Russell’s paradox.
• Infinity constructs $\mathbb{N}$ inside set theory via the von Neumann encoding $n = \{0, 1, \ldots, n-1\}$.
• Replacement (schema) maps sets to new sets under definable functions, reaching beyond what Separation alone can build.
• Regularity forbids circular membership chains and supports well-founded induction on $\in$.
• The Axiom of Choice asserts that choice functions exist for arbitrary families of non-empty sets. It is independent of ZF and equivalent to the well-ordering theorem and Zorn’s lemma.
• From these ten axioms and $\emptyset$, all of standard mathematics — $\mathbb{N}$, $\mathbb{Z}$, $\mathbb{Q}$, $\mathbb{R}$, $\mathbb{C}$, functions, ordinals, cardinals — can be constructed.