Set Theory and Independence

An axiomatic foundation can organize a large body of mathematics without deciding every question formulated in its language. The incompleteness theorems establish a general limitation, but they do not identify the status of a particular problem about sets.

The continuum hypothesis required its own mathematical analysis. Gödel's constructible universe and Cohen's forcing supplied the relevant model constructions. Together they showed that, assuming ZFC is consistent, its axioms neither prove nor refute the hypothesis.

This result did more than leave a famous question unanswered. It made independence a method for understanding the strength and limitations of set-theoretic principles.

The cumulative hierarchy

The modern ZF axioms can be understood through a hierarchy of sets:

at a limit ordinal $\lambda$.

Each successor stage collects all subsets of the preceding stage. Limit stages collect what has appeared earlier. The hierarchy continues through the ordinals rather than stopping after finitely many or merely countably many steps.

In the surrounding theory, Foundation supports the assertion that every set appears at some stage. This account also depends on the other axioms and the machinery of transfinite recursion. Foundation alone does not construct the hierarchy.

The hierarchy is a conception of how sets are organized, not a finite algorithm that enumerates every set. Forming a power set is an existence principle, and the intended totality of subsets may be far richer than any effective list.

Choice and well-ordering

The axiom of choice states that every family of nonempty sets has a choice function selecting one member from each set.

For a finite family, such selections present little difficulty. For an arbitrary infinite family, the assertion supplies a simultaneous choice even when no explicit selection rule is given.

Zermelo used choice in his 1904 proof that every set can be well ordered. A well-order places the elements in an order in which every nonempty subset has a least element. Over the other usual set-theoretic axioms, the well-ordering principle and choice are equivalent.

Choice supports convenient general theorems in algebra and analysis. It also permits constructions that conflict with some intuitive expectations about size or measurability. Its status is therefore a question about which existence principles one accepts, not merely a matter of notation.

The distinction between ZF and ZFC records this choice explicitly: ZFC is ZF with the axiom of choice.

The continuum hypothesis

Cantor proved that the real numbers are uncountable. The continuum hypothesis asks whether there is an intermediate cardinality between the natural numbers and the reals.

In ZFC notation, it is

Here $\aleph_1$ is the least uncountable cardinal, and $2^{\aleph_0}$ is the cardinality of the power set of the natural numbers, equivalently the size of the continuum.

The generalized continuum hypothesis, GCH, makes the corresponding assertion

for every infinite cardinal $\kappa$.

Neither hypothesis is simply Cantor's theorem. Cantor establishes $2^\kappa>\kappa$; CH and GCH specify how large the jump is.

The model-theory note showed how structures can satisfy the same first-order information while differing externally. Independence arguments require a more particular achievement: construct models of the relevant axioms in which the target sentence takes different truth values.

Gödel's constructible universe

Gödel's work announced in 1938 and developed in his 1940 monograph constructs the inner model $L$.

Its stages resemble the cumulative hierarchy but use definable subsets:

$\operatorname{Def}(L_\alpha)$ contains subsets definable over the structure $(L_\alpha,\in)$, using first-order formulas and parameters from $L_\alpha$.

The restriction is substantial. The power-set operation collects all subsets; the constructible successor operation collects those described by the permitted definitions over that stage.

Gödel showed that $L$ satisfies the axioms of set theory together with choice and GCH, in the appropriate relative sense. The construction yields

It follows that choice and CH cannot be refuted from the weaker background if that background is consistent.

This does not show that the intended universe is $L$. The additional principle $V=L$ is a substantive assertion. An inner model that satisfies a sentence establishes a consistency result, not the truth of that sentence in every model or in every intended foundational interpretation.

Cohen's forcing: finite information about a new object

Paul Cohen introduced forcing in 1963 to obtain the other direction. The method constructs extensions of models while controlling the sentences that hold in the extension.

For intuition, begin with finite binary sequences ordered by extension. A longer sequence supplies more information about a prospective infinite sequence.

A set of conditions is dense if every condition can be extended to one in that set. For example, the conditions whose length exceeds $n$ form a dense set. Meeting each of these ensures that the eventual union specifies a bit at every natural-number position.

A generic filter meets the dense sets required by the ground model. Its union supplies an object that has been constrained through many finite requirements.

The word “generic” therefore does not mean random. It means satisfying a precisely specified collection of dense requirements. The forcing relation

connects finite conditions with statements that will hold in the extension.

The forcing theorem establishes the correspondence between this relation and truth in the generic extension. It is what allows the new model's properties to be proved rather than inferred from an informal picture of adding objects.

Why one new real is not enough

Adding a single generic real does not by itself refute CH. The number of reals and the preservation of cardinalities both matter.

A standard construction starts with a suitable ground model and uses finite partial functions

The generic object supplies a binary sequence for each index below the ground model's $\omega_2$. Dense requirements ensure that these sequences are distinct.

The forcing has the countable chain condition, established through combinatorial analysis of finite conditions. This provides the needed preservation of cardinalities. The extension therefore has at least $\aleph_2$ distinct reals while retaining the relevant distinction between $\aleph_1$ and $\aleph_2$.

Consequently,

and CH fails.

This sketch identifies two indispensable parts of the argument: producing many new reals and preventing the apparent increase from disappearing through a collapse of cardinals. A picture showing only one new sequence omits the central cardinal arithmetic.

The assumptions of the model construction

For exposition, forcing is often introduced over a countable transitive model $M$ of a suitable set theory. Its dense subsets can then be enumerated externally, allowing a generic filter to be constructed by successive choices.

The existence of such a model is stronger than merely asserting consistency of ZFC. The pedagogical assumption must not be silently substituted for the hypothesis of the final relative-consistency theorem.

The general metatheorem can be justified through formalized syntactic arguments, Boolean-valued methods, or appropriately organized model-theoretic reasoning. The resulting statement is that consistency of ZFC entails consistency of ZFC with the negation of CH.

Combined with Gödel's result, this establishes the independence of CH from ZFC, conditional on consistency.

The independence of choice from ZF requires another distinction. A full set-forcing extension of a model of ZFC still satisfies choice. To obtain models where choice fails, symmetric-model methods select suitable submodels rather than simply treating every generic extension as a model without choice.

Independence as a research method

After Cohen, forcing became a flexible way to investigate possible set-theoretic universes. Easton's theorem showed extensive freedom in the values of the power-set function on regular cardinals, subject to necessary restrictions and the theorem's hypotheses.

Solovay used a large-cardinal assumption to construct a model in which all sets of reals are Lebesgue measurable while retaining dependent choice. The full axiom of choice cannot coexist with that conclusion in the usual setting.

Large cardinals propose strong existence principles about the hierarchy. Determinacy principles instead assert the existence of winning strategies for certain infinite games. These approaches interact through deep theorems, but they should not be treated as interchangeable additional axioms.

The subject now asks both what different axioms permit and what reasons support adopting them. Structural coherence, explanatory power, interaction with existing mathematics, and philosophical views of the set universe all enter that assessment.

What independence leaves open

An independence theorem settles a question about derivability from specified axioms. It does not require everyone to conclude that the target statement has no determinate meaning or that every extension of the axioms is equally justified.

Nor does it show that set theory failed as a foundation. The axioms continue to support extensive mathematics, while independence results reveal which further commitments particular statements require.

The history of logic also contains systems designed to express necessity, possibility, time, and knowledge. Their semantics evaluate assertions across related alternatives rather than within one structure alone. The next note follows that development.

Sources and further reading

• Kurt Gödel, The Consistency of the Axiom of Choice and of the Generalized Continuum-Hypothesis with the Axioms of Set Theory (1940), following the 1938 announcements.
• Paul J. Cohen, “The Independence of the Continuum Hypothesis” (1963) and its 1964 continuation.
• William B. Easton, “Powers of Regular Cardinals” (1970).
• Robert M. Solovay, “A Model of Set-Theory in Which Every Set of Reals Is Lebesgue Measurable” (1970).
• The Continuum Hypothesis, for the distinction between independence, new axioms, and competing interpretations of the set universe.