The ZFC Axioms

The Zermelo-Fraenkel axioms with Choice, collectively known as ZFC, form the standard axiomatic foundation for nearly all of modern mathematics. Born out of the crisis triggered by Russell’s paradox, these axioms replaced the informal notion of “a set is any collection” with a precise first-order theory that carefully controls which sets exist and how new sets can be formed. In this article we present the axioms, explore the cumulative hierarchy they give rise to, show how the natural numbers and basic arithmetic are constructed within the theory, and sketch the path from there to the integers, rationals, and real numbers — demonstrating that ZFC is powerful enough to serve as a universal foundation.

The Zermelo-Fraenkel Axioms

The story of ZFC begins with a catastrophe. In 1901, Bertrand Russell showed that the naive comprehension principle — the assumption that any property determines a set — leads to contradiction. The set $R = \{x : x \notin x\}$ can neither contain itself nor fail to contain itself. This paradox struck at the heart of Gottlob Frege’s recently published Grundgesetze der Arithmetik and made it clear that set theory needed to be rebuilt on firmer ground.

Ernst Zermelo published the first axiomatization in 1908, driven both by Russell’s paradox and by the need to place his own 1904 proof of the well-ordering theorem on a rigorous footing. His system introduced the key idea of restricting set formation: instead of forming sets from arbitrary properties, one can only separate elements from an already-existing set. In 1922, Abraham Fraenkel and, independently, Thoralf Skolem identified a gap in Zermelo’s system — the need for the axiom schema of replacement — and proposed refinements. The resulting theory, known as ZF, is formulated in first-order logic with a single binary relation symbol $\in$ (membership). When the Axiom of Choice is added, we obtain ZFC.

The axioms of ZF are as follows:

Axiom	Informal Statement
Extensionality	Two sets are equal if and only if they have the same elements.
Empty Set	There exists a set with no elements (derivable from Separation, but often stated explicitly).
Pairing	For any two sets $a$ and $b$, the set $\{a, b\}$ exists.
Union	For any set $a$, the union $\bigcup a$ of all elements of elements of $a$ exists.
Power Set	For any set $a$, the collection of all subsets of $a$ forms a set $\mathcal{P}(a)$.
Infinity	There exists a set containing $\emptyset$ and closed under the successor operation $x \mapsto x \cup \{x\}$.
Separation Schema	For any set $a$ and any first-order property $\varphi(z)$, the collection $\{z \in a : \varphi(z)\}$ is a set.
Replacement Schema	If $\varphi(x, y)$ defines a function on a set $a$, then the image $\{y : \exists x \in a\, \varphi(x, y)\}$ is a set.
Foundation (Regularity)	Every nonempty set has an $\in$-minimal element.

The Extensionality axiom says that a set is completely determined by its elements — there are no two distinct sets with exactly the same members. Formally: $\forall x\,\forall y\,[\forall z\,(z \in x \leftrightarrow z \in y) \to x = y]$. This is the identity criterion for sets and underlies the entire theory.

Pairing, Union, and Power Set provide the basic closure operations. From any two sets you can form their pair; from any collection of sets you can take the union of their elements; and from any set you can form the collection of all its subsets. Together with the empty set, these axioms let you build up an enormous variety of finite and infinite structures.

The Axiom of Infinity is what lifts set theory beyond the finite. It asserts the existence of at least one infinite set — specifically, a set $\omega$ satisfying $\emptyset \in \omega$ and $\forall x\,(x \in \omega \to x \cup \{x\} \in \omega)$. Without this axiom, all the other axioms would be satisfied by the hereditarily finite sets alone.

Separation (also called Comprehension or Aussonderung) is the safe replacement for naive comprehension. It allows you to carve out subsets of an existing set using a first-order property, but it cannot conjure a set from thin air. It is a schema — one axiom for each first-order formula $\varphi$. Replacement is strictly stronger: it says that the image of a set under any definable function is again a set. Replacement implies Separation, but it is kept as a distinct schema because of its additional power in constructing sets of high rank, such as $V_{\omega + \omega}$.

The Axiom of Choice (AC) stands apart and is traditionally listed separately, which is why we speak of ZF versus ZFC. It asserts that for every family of nonempty sets, there exists a function selecting exactly one element from each. Formally: for every set $a$ of nonempty sets, there is a function $f : a \to \bigcup a$ such that $f(x) \in x$ for each $x \in a$. The Axiom of Choice is equivalent to Zorn’s Lemma (every partially ordered set in which every chain has an upper bound contains a maximal element) and to the Well-Ordering Theorem (every set can be well-ordered). Weaker variants include Countable Choice and the Axiom of Dependent Choice, which suffice for most of analysis. Choice was historically controversial — it asserts existence without providing a construction — but it is now accepted as a standard part of working mathematics.

The Cumulative Hierarchy and Foundation

The axioms of ZFC do not merely list rules for forming sets; they implicitly describe a layered picture of the entire set-theoretic universe. This picture is the cumulative hierarchy, developed by John von Neumann in the 1920s, and it provides the conceptual and technical backbone of modern set theory.

The hierarchy is built by transfinite recursion on the ordinals. We begin with nothing and iteratively apply the power set operation:

$V_0 = \emptyset$

$V_{\alpha + 1} = \mathcal{P}(V_\alpha)$

$V_\lambda = \bigcup_{\alpha < \lambda} V_\alpha \quad \text{for limit ordinals } \lambda$

At the base, $V_0$ is empty. At the first successor stage, $V_1 = \mathcal{P}(\emptyset) = \{\emptyset\}$, a set with one element. Then $V_2 = \{\emptyset, \{\emptyset\}\}$ with two elements, and $V_3$ has four elements — and so on, with each finite stage $V_n$ having $2^{|V_{n-1}|}$ elements. The first limit stage $V_\omega = \bigcup_{n < \omega} V_n$ collects all the hereditarily finite sets — every set whose members, and members of members, and so on, are all finite. Beyond $V_\omega$, the hierarchy continues through all the transfinite ordinals: $V_{\omega + 1} = \mathcal{P}(V_\omega)$ is already uncountable, and the levels grow with breathtaking speed. The universe of all sets is defined as:

$V = \bigcup_{\alpha \in \mathrm{Ord}} V_\alpha$

where $\mathrm{Ord}$ denotes the proper class of all ordinals. Every set in ZFC appears at some level of this hierarchy, and the rank of a set $x$ is the least ordinal $\alpha$ such that $x \in V_{\alpha + 1}$. Rank provides a measure of complexity: the natural numbers have finite rank, the set of real numbers has rank $\omega + 1$, and the objects of ordinary mathematics live at relatively low levels of the hierarchy.

The Axiom of Foundation (also called Regularity) is what makes this picture work. It states that every nonempty set $x$ contains an element $y$ such that $x \cap y = \emptyset$ — in other words, $y$ is $\in$-minimal in $x$. Formally:

$\forall x\,[x \neq \emptyset \to \exists y\,(y \in x \land y \cap x = \emptyset)]$

An equivalent formulation is that there is no infinite descending chain $\cdots \in x_2 \in x_1 \in x_0$ — the membership relation is well-founded. Foundation rules out pathological sets: there is no set $x$ with $x \in x$, no pair of sets with $a \in b$ and $b \in a$, and no circular membership of any kind. Foundation guarantees that every set can be assigned a rank and thus appears somewhere in the cumulative hierarchy.

This is sometimes called the iterative conception of sets: sets are formed in stages, and at each stage you can only collect sets that were already available at earlier stages. The axioms of ZFC can be seen as formalizing this intuitive picture. The iterative conception provides a philosophical justification for the axioms — each axiom corresponds to a natural step in the stage-by-stage construction — and it explains why ZFC avoids the paradoxes: Russell’s set $R = \{x : x \notin x\}$ would have to contain sets of every rank and therefore could not appear at any particular stage.

Natural Numbers and Peano Arithmetic

One of the first accomplishments of ZFC is the construction of the natural numbers as specific sets. The standard encoding, due to von Neumann, identifies each natural number with the set of all smaller natural numbers:

$0 = \emptyset, \quad 1 = \{\emptyset\}, \quad 2 = \{\emptyset, \{\emptyset\}\}, \quad 3 = \{\emptyset, \{\emptyset\}, \{\emptyset, \{\emptyset\}\}\}, \quad \ldots$

In general, the successor of $n$ is defined as $S(n) = n \cup \{n\}$. Under this encoding, each natural number $n$ is a set with exactly $n$ elements, and $m < n$ if and only if $m \in n$. The elegance of the von Neumann construction is that it simultaneously defines the natural numbers and the ordering relation on them.

The Axiom of Infinity guarantees the existence of a set that contains $0 = \emptyset$ and is closed under the successor operation. By applying the Separation schema, we extract the smallest such set — the intersection of all sets containing $\emptyset$ and closed under successor. This minimal set is the set of natural numbers $\omega$ (also written $\mathbb{N}$).

The set $\omega$ satisfies the Peano axioms as formulated within ZFC:

1. $0 \in \omega$ (zero is a natural number).
2. If $n \in \omega$, then $S(n) \in \omega$ (the successor of a natural number is a natural number).
3. There is no $n \in \omega$ with $S(n) = 0$ (zero is not a successor).
4. If $S(m) = S(n)$, then $m = n$ (the successor function is injective).
5. Mathematical induction: if a set $A \subseteq \omega$ contains $0$ and is closed under successor, then $A = \omega$.

The induction principle follows from the definition of $\omega$ as the smallest inductive set. Since $A$ satisfies the closure conditions, it is an inductive set containing $\omega$ as a subset; but $\omega$ is contained in every inductive set, so $A = \omega$.

With $\omega$ in hand, we define addition by recursion on the second argument:

$m + 0 = m, \qquad m + S(n) = S(m + n)$

and multiplication similarly:

$m \cdot 0 = 0, \qquad m \cdot S(n) = m \cdot n + m$

These recursive definitions are justified by the recursion theorem, which itself follows from the properties of well-ordered sets in ZFC. The familiar laws of arithmetic — commutativity, associativity, distributivity — are then proved by induction. In this way, the entire structure of elementary number theory is derived from the axioms of ZFC, with no additional assumptions.

Formal Constructions within ZFC

The power of ZFC as a foundation for mathematics becomes visible when we extend beyond the natural numbers to construct the full number systems of analysis. Each step uses a standard set-theoretic technique — typically equivalence classes or subsets — applied to structures already built.

Before proceeding, we need the notion of an ordered pair. The Kuratowski definition encodes the ordered pair $(a, b)$ as the set $\{\{a\}, \{a, b\}\}$. This seemingly arbitrary definition has exactly the right property: $(a, b) = (c, d)$ if and only if $a = c$ and $b = d$. With ordered pairs in hand, we can define Cartesian products, relations, and functions as sets, making them first-class citizens of ZFC.

The integers $\mathbb{Z}$ are constructed from the natural numbers as equivalence classes of pairs. Consider the set $\omega \times \omega$ of all pairs $(a, b)$ of natural numbers, and define an equivalence relation by $(a, b) \sim (c, d)$ if and only if $a + d = b + c$. The intuition is that the pair $(a, b)$ represents the integer $a - b$. Each equivalence class is an integer: the class of $(3, 0)$ is the integer $+3$, the class of $(0, 2)$ is $-2$, and the class of $(5, 5)$ is $0$. Addition and multiplication are defined on representatives and shown to be well-defined (independent of the choice of representative).

The rationals $\mathbb{Q}$ are built from the integers by the same technique. Consider pairs $(p, q)$ with $p \in \mathbb{Z}$ and $q \in \mathbb{Z} \setminus \{0\}$, and define $(p, q) \sim (r, s)$ if and only if $p \cdot s = q \cdot r$. Each equivalence class is a rational number, representing the fraction $p/q$. The field operations extend naturally, and one verifies that $\mathbb{Q}$ is an ordered field.

The real numbers $\mathbb{R}$ require a more substantial construction. There are two classical approaches, both carried out entirely within ZFC. In the Dedekind cut construction, a real number is defined as a partition of $\mathbb{Q}$ into two nonempty sets $L$ and $R$ such that every element of $L$ is less than every element of $R$ and $L$ has no greatest element. The real number $\sqrt{2}$, for instance, is the cut where $L = \{q \in \mathbb{Q} : q < 0 \text{ or } q^2 < 2\}$. The alternative approach uses Cauchy sequences: a real number is an equivalence class of Cauchy sequences of rationals, where two sequences are equivalent if their difference converges to zero. Both constructions yield a complete ordered field, and one proves that any two complete ordered fields are isomorphic — so the real numbers are uniquely determined up to isomorphism.

This chain of constructions — from $\emptyset$ to $\omega$ to $\mathbb{Z}$ to $\mathbb{Q}$ to $\mathbb{R}$ — illustrates the foundational thesis of ZFC: essentially all of mathematics can be formalized as statements about sets built from the empty set using the ZFC axioms. Functions are sets of ordered pairs, topological spaces are sets equipped with distinguished collections of subsets, groups are sets with binary operations satisfying certain properties, and so on. This universality is both ZFC’s great strength and its raison d’etre. It does not mean that working mathematicians must think in terms of sets at every moment, but it does mean that there is a single coherent framework in which the consistency of all standard mathematical reasoning can, in principle, be verified.