The Continuum Hypothesis

The Continuum Hypothesis is not an isolated curiosity; it opened an entire research program. Forcing, invented to prove Part (b) above, became the single most important tool in modern set theory, used to prove the independence of dozens of other statements: the existence of a Suslin line, various forms of Martin's Axiom, whether every set of reals has the Baire property, and much of the combinatorics of cardinal invariants of the continuum (splitting numbers, bounding numbers, and the like, all of which sit provably between $\aleph_1$ and $2^{\aleph_0}$ with independence results of their own). In descriptive set theory, questions about definable sets of reals (Borel, analytic, projective sets) interact intimately with CH: the Perfect Set Property — every uncountable closed set of reals has a subset in bijection with $\mathbb R$ itself — holds for all closed sets in ZFC alone, extends to all analytic sets in ZFC, but for arbitrary sets of reals is exactly where CH-independent phenomena resurface. In category theory and topology, CH is used as a convenient hypothesis to construct pathological objects: Sierpiński used it (before independence was known, simply as an assumption) to construct a subset of the plane meeting every line in exactly countably many points, and to build a decomposition of $\mathbb R^3$ related to non-measurable sets. Some results in functional analysis (e.g. on the structure of ultrafilters on $\mathbb{N}$ and the Stone–Čech compactification $\beta\mathbb{N}$) genuinely change character depending on whether CH holds. In theoretical computer science, the study of oracle hierarchies and generalized computability occasionally invokes CH-style independence phenomena as a cautionary template for what "provably unresolvable questions" look like, and forcing techniques have inspired independence results for questions in the foundations of complexity theory. Philosophically, CH remains the flagship example driving the search for new axioms for set theory beyond ZFC. Gödel himself speculated that CH is simply false, and that mathematicians would eventually discover new, self-evident axioms (perhaps large cardinal axioms) that decide it. Hugh Woodin has developed extensive programs (the $\Omega$-logic program, and later the "ultimate $L$" program) aimed at identifying a canonical extension of ZFC under which CH would be settled — current work tends to suggest $2^{\aleph_0} = \aleph_2$ as the "correct" answer under certain forcing-axiom-based extensions, though this remains a live and contested research frontier, not a settled theorem.

Ask the tutor: How exactly does the Condensation Lemma guarantee that every constructible real number appears at some countable stage of Gödel's hierarchy?

Cantor's diagonal argument shows that no function $\mathbb{N} \to \mathbb{R}$ is surjective, so $\mathbb{R}$ is strictly bigger than $\mathbb{N}$. This alone doesn't tell you how much bigger. Cardinal arithmetic gives you a tower of infinities $\aleph_0 < \aleph_1 < \aleph_2 < \cdots$, built purely from the order structure of well-orderings, completely independent of any construction involving power sets. Separately, cardinal exponentiation gives you $2^{\aleph_0}$, built from subsets, a completely different kind of construction. CH is the assertion that these two independently-generated towers happen to meet at the first possible rung: that $2^{\aleph_0}$, whatever exact cardinal it turns out to be, is not just some cardinal above $\aleph_0$ but is literally the very next one, $\aleph_1$. A useful picture: think of the cardinals as rungs on a ladder, $\aleph_0, \aleph_1, \aleph_2, \ldots$, and think of $2^{\aleph_0}$ as a mysterious point that Cantor's Theorem places somewhere above $\aleph_0$ but does not pin down. ZFC's axioms are flexible enough that this point can be consistently glued to $\aleph_1$ (Gödel's universe $L$), or pushed arbitrarily far up the ladder — to $\aleph_2$, to $\aleph_{47}$, even to cardinals with uncountable cofinality far beyond any $\aleph_n$ (subject only to a technical constraint discovered later by König). The axioms tell you the point exists and lies strictly above $\aleph_0$; they do not tell you which rung, or even whether it lands on a labeled rung at all in a straightforward way.

Before stating the central theorem, it is worth being precise about what independence means, because this is where beginners most often get confused. Fix ZFC as our background theory. A sentence $\varphi$ in the language of set theory is independent of ZFC if neither $\varphi$ nor $\neg\varphi$ is provable from the ZFC axioms — assuming ZFC itself is consistent (has a model, or equivalently derives no contradiction). This is exactly analogous to how the Parallel Postulate is independent of the other axioms of Euclidean geometry: both it and its negation are consistent with the rest, giving Euclidean and non-Euclidean geometries as equally valid extensions. To prove independence of $\varphi$, one exhibits: 1. A model $M_1 \models \mathrm{ZFC} + \varphi$ (showing $\neg\varphi$ is not provable — if it were, it would hold in every model of ZFC, including $M_1$). 2. A model $M_2 \models \mathrm{ZFC} + \neg\varphi$ (showing $\varphi$ is not provable, symmetrically). Both constructions are carried out relative to the consistency of ZFC itself — by Gödel's Second Incompleteness Theorem, ZFC cannot prove its own consistency, so all such results are stated as "if ZFC is consistent, then ZFC + $\varphi$ is consistent," never as an unconditional consistency proof.

The two constructions are genuinely different in flavor: Gödel's $L$ is an inner model, obtained by restricting to a definable sub-universe of a given model, shrinking the collection of subsets allowed at each stage until CH becomes forced. Cohen's forcing is an outer model technique, growing a given model by generically adjoining new objects until CH is broken. That both directions are achievable, using dual strategies, is what makes the independence phenomenon feel less like a limitation on our knowledge and more like a genuine structural fact about the axioms.

Example 1 — Verifying CH holds in a toy analogue: constructibility from a countable base. To build intuition for why restricting to definable subsets caps the cardinality of the continuum, work a miniature, fully checkable version of the idea. Let $M_0 = \omega$ (the natural numbers with their usual structure), and define $M_{n+1}$ to consist of $M_n$ together with all subsets of $M_n$ that are definable over $(M_n, \in, +, \cdot)$ by a first-order formula with parameters in $M_n$ — exactly Gödel's operation, but starting from $\omega$ instead of $\varnothing$, and iterating only finitely.

• $M_0 = \omega$ is countably infinite.
• At each step, "definable subsets of $M_n$" means: fix the countably many first-order formulas $\varphi_1, \varphi_2, \ldots$ in the language, and for each formula and each finite tuple of parameters from $M_n$ (there are $\aleph_0$-many finite tuples from a countably infinite set, since $\bigcup_k M_n^k$ is a countable union of countable sets), form $\{x \in M_n : (M_n,\in,+,\cdot) \models \varphi_i(x, \bar a)\}$. This produces at most $\aleph_0 \times \aleph_0 = \aleph_0$ new subsets (a countable union, indexed by formula, of countable sets, indexed by parameter tuple).
• Hence $M_{n+1}$ has cardinality $|M_n| + \aleph_0 = \aleph_0$ whenever $|M_n| = \aleph_0$: by induction, every $M_n$ is countable.
• Set $M_\omega = \bigcup_n M_n$: a countable union of countable sets is countable (using countable choice, or here an explicit enumeration since each $M_n$ is explicitly indexed), so $M_\omega$ is countable.

This mirrors exactly the mechanism by which Gödel bounds $|L \cap \mathbb{R}|$: at every stage of the constructible hierarchy, you only ever add countably many new definable objects to a set of size $\aleph_1$ or less, and taking a union of $\aleph_1$-many such stages (rather than finitely or countably many, as here) yields a total of exactly $\aleph_1$. The full argument for $L$ replaces "iterate $\omega$-many times" with "iterate $\omega_1$-many times," replaces the countable base case with a well-ordered hierarchy over all ordinals, and uses the Condensation Lemma to push every real down into some countable stage $L_\alpha$ — this is the nontrivial step this toy model does not need to reproduce, since here we stopped at $\omega$ steps by fiat, but which in the real construction requires the elementary-submodel-and-collapse argument.

Example 2 — A concrete "in-between" candidate under $\neg$CH, and why it does not contradict Cantor's Theorem. Suppose we work in a model (like a Cohen extension) where $2^{\aleph_0} = \aleph_2$. Consider the set

More concretely: since $2^{\aleph_0} = \aleph_2 > \aleph_1$, ZFC guarantees the existence of a subset $A \subseteq \mathbb{R}$ with $|A| = \aleph_1$ — for instance, well-order $\mathbb{R}$ (possible by Choice) and let $A$ be the first $\aleph_1$ elements in that well-ordering. Such an $A$ satisfies

exhibiting explicitly a cardinality strictly between $\aleph_0$ and $2^{\aleph_0}$ — precisely the object CH asserts does not exist. This is not a counterexample to any theorem; it is exactly what $\neg$CH means, made concrete: in a model where the continuum is $\aleph_2$, the "gap rung" $\aleph_1$ is genuinely occupied by real sets of reals, and one can point to a specific one (the initial segment of length $\aleph_1$ under a well-ordering of $\mathbb{R}$) once a well-ordering has been fixed. The subtlety worth noting: this $A$ is not "explicit" in any constructive sense — its existence used the Axiom of Choice to well-order $\mathbb{R}$, and no formula picks out a canonical instance of $A$. This is typical: it is a theorem of Solovay-style results (beyond our scope here) that in the absence of full Choice, one can build models where every set of reals is Lebesgue measurable, and CH-type intermediate sets can fail to exist "definably" even when Choice-driven existence proofs like this one still succeed in full ZFC.