Cohen's Forcing

For nearly a century, Cantor's Continuum Hypothesis (CH)—a fundamental statement about the size of the infinite—resisted all attempts at proof or disproof from the standard axioms of set theory (ZFC). While Kurt Gödel had shown CH could not be disproven, its provability remained a gaping hole in the foundations of mathematics. This is the problem that Paul Cohen's revolutionary method, known as forcing, was invented to solve. Forcing is not merely a logical trick; it is a profound and creative toolkit for constructing new mathematical realities, extending our familiar universe of sets to create new ones where different mathematical truths hold.

This article will guide you through the workshop where these universes are built. In the first chapter, ​​"Principles and Mechanisms"​​, we will examine the machinery of forcing itself, exploring the core concepts of conditions, names, the forcing relation, and the generic filters that bring a new world into existence. We will see how this process, while radical, is carefully designed to preserve the fundamental laws of set theory. Following that, in ​​"Applications and Interdisciplinary Connections"​​, we will witness the power of this method in action. We will see how it definitively settled the status of the Continuum Hypothesis and evolved into an indispensable tool for exploring the vast landscape of mathematical possibility, from building models for Martin's Axiom to mapping the fine structure of the real number line.

Imagine you are a physicist. You have a set of fundamental laws—say, Newton's laws—and you want to know if they permit a certain phenomenon, like a perpetual motion machine. You can try to build one, and if you succeed, you've shown it's possible. If you can prove from the laws that building one leads to a contradiction, you've shown it's impossible. But what if you can do neither? This is the situation mathematicians found themselves in with Cantor's Continuum Hypothesis (CH). The standard axioms of set theory, Zermelo-Fraenkel with Choice (ZFC), seemed unable to either prove or disprove it.

So, what does a mathematician do? They take a page from the physicist's book: they try to build a universe. Kurt Gödel, in a stroke of genius, showed that you could find a universe within our own, a slimmed-down, orderly place called the ​​constructible universe ($L$)​​, where $\mathrm{CH}$ is true. This proved that ZFC could not disprove $\mathrm{CH}$, because if it could, then even this inner sanctum $L$ would have to obey that disproof, which it doesn't. This established one half of the independence of $\mathrm{CH}$.

Paul Cohen's great insight was to go in the other direction. Instead of building a smaller universe, he figured out how to build a bigger one. Forcing is the radical and beautiful machinery for constructing these new mathematical realities—universes that are extensions of the one we start with, but which contain new objects and, potentially, new truths. By building a universe where $\mathrm{CH}$ is false, Cohen completed the proof that $\mathrm{CH}$ is truly independent of our standard axioms. Let's walk through the workshop and see how these universes are made.

How can you describe a universe you haven't built yet? You can't just point to its new inhabitants. Cohen’s solution was to create a special language to talk about the properties of the new universe while standing firmly in the old one. This is the ​​forcing language​​.

The first ingredients are ​​conditions​​, which are finite pieces of information about the new objects we want to add. Think of them as blueprints. If we want to add a new real number (which can be seen as an infinite sequence of 0s and 1s), a condition might specify the first few digits. For example, in Cohen's original setup, the set of conditions, denoted $\mathbb{P}$, consists of all finite partial functions from the natural numbers $\omega$ to $\{0, 1\}$. A condition like $p = \{(0,1), (1,0), (2,1)\}$ is a piece of information specifying the first three bits of our new number. A "stronger" condition is one that contains more information; for example, $q = \{(0,1), (1,0), (2,1), (3,0)\}$ is stronger than $p$ because it extends it.

Next, every object in the new universe, from a simple number to a complex function, must have a ​​name​​ in our ground model. A name is a structured set that acts as a recipe, telling us how the object is to be constructed once the new "generic" information is available. For instance, we could create a name $\dot{x}$ that is intended to represent a new set of integers. The name might look like a collection of pairs: $\dot{x} = \{(\check{n}, p_n) : n \in \omega\}$, where $\check{n}$ is the canonical name for the integer $n$ and $p_n$ is a condition. The rule for interpreting this name is: the integer $n$ will be in our new set if the condition $p_n$ ends up being part of our final blueprint for the universe.

This leads us to the heart of the language: the ​​forcing relation​​, denoted by $\Vdash$. A statement like $p \Vdash \varphi$ means, "The condition $p$ contains enough information to force the statement $\varphi$ to be true in the new universe." For example, the condition $p=\{(0,1)\}$ forces the statement "the 0-th bit of the new real is 1". The miracle of forcing is that this entire relationship—this ability to decide what a piece of information guarantees about a future world—can be precisely defined within our current universe. We can reason about all possible extensions without ever leaving home.

The forcing language describes a whole multiverse of possibilities. To build one specific new universe, we must make a complete and consistent set of choices. This is the role of the ​​generic filter ($G$)​​.

Imagine you have a series of questions you must answer to specify your new universe. For Cohen's real number, these questions are "What is the 0-th bit?", "What is the 1st bit?", and so on. In the language of forcing, these questions correspond to ​​dense sets​​. A set $D$ of conditions is dense if, no matter what information you already have (some condition $p$), you can always find a stronger piece of information $d$ in $D$ that answers the question. For instance, the set of all conditions that specify the 10th bit is a dense set.

A ​​generic filter $G$​​ is a special collection of conditions that provides a cohesive set of answers to all such questions that can be formulated in our ground model. It's a "generic" description because it's not biased; it avoids any special properties that we could have singled out beforehand. When we say $G$ "meets" a dense set $D$, we mean their intersection is not empty ($G \cap D \neq \emptyset$), which is just a formal way of saying $G$ contains an answer to the question represented by $D$. It's worth noting that while $G$ must intersect every dense set, this intersection isn't necessarily a single point. If the dense set contains compatible pieces of information, the generic filter might contain several of them.

Here comes the sleight of hand that makes forcing possible. It turns out that we cannot actually find a generic filter for our entire universe $V$ that lives inside $V$. Trying to do so leads to a paradox related to Tarski's theorem on the undefinability of truth. It's like trying to find a map of the entire world that includes a map of itself—you get caught in an infinite regress. Cohen's solution was brilliant: don't try to extend the whole universe. Instead, start with a small, countable "toy" model of set theory, let's call it $M$. Because $M$ is countable, it only has a countable number of dense sets. We can then, working in our larger universe $V$, easily construct a filter $G$ that meets all of them. This $G$ is not in $M$, but it is in $V$. This is enough! By building the extension $M[G]$, we have created a new, perfectly valid model of set theory. If we can show that ZFC + ¬CH is true in this model, we have proved that ZFC + ¬CH is a consistent theory, which is exactly what's needed to establish independence.

We've built a new world, $M[G]$, by interpreting all our names using the information in our generic filter $G$. But is it a hospitable place for a mathematician? Does it still obey the fundamental laws of set theory, the ZFC axioms?

Amazingly, the answer is yes. The proof is a beautiful piece of self-reference. Let's take the Axiom of Separation as an example. It says that for any set $A$ and any property $\psi$, the collection of elements in $A$ that have property $\psi$ also forms a set. To prove this holds in $M[G]$, we take names for $A$ and the property $\psi$ in our ground model $M$. We then construct a new name, let's call it $\dot{B}$, specifically designed to be the name for our desired subset. The recipe for $\dot{B}$ is: a name $\rho$ gets put into $\dot{B}$ (with some condition $p$) if and only if $p$ forces that the object named by $\rho$ is in $A$ and has the property $\psi$.

Because the forcing relation ($\Vdash$) is definable in $M$, this entire recipe for $\dot{B}$ can be written down as a formula in the language of $M$. We can then use the Axiom of Separation in M to prove that this collection of pairs, $\dot{B}$, is a genuine set in $M$. We've used an axiom in the old universe to forge a name that guarantees the same axiom will hold in the new one! The forcing theorem then ensures that when we interpret $\dot{B}$ using our generic filter $G$, we get exactly the set we were looking for. This same style of argument works for all the other ZFC axioms, ensuring our new universe is structurally sound.

When performing surgery, a doctor wants to fix the problem without causing unintended damage. Forcing is similar. When we add new sets to our universe to make $\mathrm{CH}$ false, we want to be surgical. The goal is typically to add new real numbers to make the continuum large, while leaving the cardinal numbers, like $\aleph_1$, untouched. If we accidentally made $\aleph_1$ (the first uncountable number) become countable in our new universe, our statement "$2^{\aleph_0} > \aleph_1$" would be meaningless.

This is where a crucial property of forcing notions comes in: the ​​countable chain condition (ccc)​​. An ​​antichain​​ is a set of conditions that are mutually incompatible—like a set of blueprints for a house where each one specifies a different color for the front door. You can only choose one. A forcing notion has the ccc if every such set of mutually exclusive options is countable.

This seemingly technical property has profound consequences. It acts as a governor, preventing the forcing from being powerful enough to destroy large cardinals. The key idea is this: to make an uncountable cardinal like $\omega_1^V$ become countable in the new universe $V[G]$, you would need to add a new function that maps the natural numbers $\omega$ onto $\omega_1^V$. The ccc condition ensures that any such function named in the ground model cannot be cofinal. The range of any function we can define is "pinned down" by a countable number of conditions, and the union of values these conditions can force is too small to cover all of $\omega_1^V$. As a result, $\omega_1^V$ remains the first uncountable cardinal in the new universe.

By using a ccc forcing, such as Cohen's, we can add a vast number of new real numbers—say, $\aleph_2$ of them—while guaranteeing that $\aleph_1$ remains $\aleph_1$. In the resulting universe, we have $2^{\aleph_0} \ge \aleph_2 > \aleph_1$. We have successfully negated the Continuum Hypothesis, and we've done it with the precision of a master surgeon, creating a new, consistent mathematical world where the landscape of the infinite is demonstrably different.

Now that we have seen the nuts and bolts of the forcing machinery, we might be tempted to view it as a clever logical trick, a tool for proving that things are unprovable. But that would be like looking at a painter's brushes and pigments and seeing only tools for making a canvas not blank. The true power and beauty of forcing lie not in what it deconstructs, but in what it constructs. It is a toolkit for building universes, a method for exploring a vast, branching landscape of mathematical possibilities, each one a consistent world governed by the laws of ZFC. In this chapter, we will journey through some of these worlds, witnessing firsthand how forcing reshapes our understanding of the mathematical firmament.

For nearly a century after its formulation by Georg Cantor, the Continuum Hypothesis (CH)—the statement that there is no set whose size is strictly between that of the integers, $\aleph_0$, and that of the real numbers, $2^{\aleph_0}$—stood as one of mathematics' greatest unsolved problems. Kurt Gödel had shown in 1940 that one could not disprove CH within ZFC. The question of whether it could be proved remained open. Paul Cohen's invention of forcing in 1963 settled the matter in the most spectacular way possible: he showed that CH is also not provable from ZFC.

The core idea is one of profound elegance. We can start with any universe of sets $M$ that we believe in, and by forcing, we can "gently" add new real numbers to it, creating a larger universe $M[G]$. The trick is to do this so delicately that we don't disturb the existing structure of the cardinals. The countable chain condition (ccc) is the key property that ensures our forcing is gentle enough not to collapse $\aleph_1$ into a countable set, or $\aleph_2$ into a set of size $\aleph_1$.

Cohen's original construction showed how to add $\aleph_2$ new real numbers to a model of ZFC. Starting with a countable model $M$ of ZFC, one can use the Cohen forcing poset $\operatorname{Add}(\omega, \aleph_2^M)$ to build an extension $M[G]$. Because this forcing is ccc, cardinals are preserved, so $(\aleph_1)^M$ is still the first uncountable cardinal in $M[G]$. But now we have at least $\aleph_2^M$ real numbers. The result is a perfectly valid model of ZFC where $2^{\aleph_0} \ge \aleph_2$, and thus CH is false.

This immediately raises a tantalizing question: if we can make the continuum larger than $\aleph_1$, just how large can we make it? Forcing provides the answer: almost as large as we want. The technique is remarkably versatile. By choosing a forcing notion like $\operatorname{Fn}(\kappa, 2, \omega)$, we can add precisely $\kappa$ new real numbers for any cardinal $\kappa$ we choose.

However, the final size of the continuum depends not only on the forcing we use but also on the "raw material" of the universe we start with. To gain precise control and set $2^{\aleph_0}$ to be exactly $\aleph_2$, for instance, it is helpful to start from a "clean" and well-behaved universe, such as one that satisfies the Generalized Continuum Hypothesis (GCH). In such a universe, the cardinal arithmetic is simple, allowing us to prove that forcing with $\operatorname{Add}(\omega, \aleph_2)$ yields a model where $2^{\aleph_0} = \aleph_2$ precisely. This interplay between the ground model and the forcing extension reveals that forcing is not a blunt instrument but a precision tool, whose effects we can predict and control.

Gödel's discovery of the constructible universe, $L$, gave mathematicians a glimpse of a beautifully minimalist world. $L$ is an "inner model" of any universe of ZFC; it is a sub-universe built from the bottom up in the simplest way possible. In this rigid and orderly world, the Generalized Continuum Hypothesis is true. For a time, one might have wondered if $L$ represented the "true" universe of sets.

Forcing demolishes this notion of a single, absolute truth. We can take the pristine universe $L$ and, by forcing over it, create a new, larger universe $L[G]$ where GCH is spectacularly false. For example, using the techniques we've seen, we can create an extension of $L$ where $2^{\aleph_0} = \aleph_2$. What does this mean for Gödel's theorem that $L$ satisfies GCH? It means that truth is relative. From the perspective of the new, larger universe $L[G]$, the old universe $L$ is still there, existing as an "inner model." All the old theorems of $L$, including GCH, remain true inside $L$. But they are no longer true for the ambient reality of $L[G]$, which contains new sets that $L$ never dreamed of.

The situation is akin to the inhabitants of a two-dimensional "Flatland" ($L$) suddenly gaining the ability to perceive a third dimension. Their new world ($L[G]$) contains objects like spheres—the new sets—that are utterly alien to their old reality. Yet, their original Flatland still exists as a plane within this 3D space, and all the laws of 2D geometry are still valid on that plane. Forcing is the tool that gives us access to these higher-dimensional realities.

The ultimate expression of this freedom is Easton's Theorem, which shows that by using a more complex form of forcing, we can control the continuum function $\kappa \mapsto 2^\kappa$ for regular cardinals almost arbitrarily, subject only to basic constraints of logic and monotonicity. We can build a universe where $2^{\aleph_0} = \aleph_{17}$, $2^{\aleph_1} = \aleph_{42}$, and $2^{\aleph_2} = \aleph_{2001}$, all at the same time. The question of the continuum's size is not a question with a single answer; it is a parameter of the universe we choose to inhabit.

Beyond just proving independence, forcing is a powerful creative instrument for constructing new mathematical worlds with fascinating and often counter-intuitive properties. This allows us to test the relationships between different mathematical principles and explore the rich combinatorial structures that can exist.

Martin's Axiom: A User-Friendly Alternative to CH

The Continuum Hypothesis is a very strong statement. Martin's Axiom (MA) is a more subtle and broadly applicable principle. In essence, MA is a general-purpose existence theorem. It states that for any "well-behaved" (ccc) partially ordered set $\mathbb{P}$, we can find a generic filter for any collection of dense sets, as long as that collection is not too large (smaller than the continuum). CH implies MA, but they are not equivalent. What if we want the powerful consequences of MA without the rigidity of CH?

Forcing allows us to build such a world. The consistency of MA + $\neg$CH is one of the crown jewels of forcing theory. The proof is a masterpiece of construction. One starts with a model satisfying GCH and performs an "iterated forcing" of length $\aleph_2$. This is like a grand construction project: at each of the $\aleph_2$ stages, a bookkeeping device identifies a potential failure of MA and applies a specific forcing to "fix" it, adding a required generic filter. Simultaneously, at many of the stages, one also adds a new Cohen real. After $\aleph_2$ steps, we have systematically satisfied every instance of MA while also adding $\aleph_2$ new reals, resulting in a model of MA + ($2^{\aleph_0} = \aleph_2$).

The technical engine that makes this possible is ​​finite support iteration​​. This method allows us to chain together ccc forcings, with the crucial guarantee that the entire iterated construction remains ccc. This preservation property is the "magic" that ensures the process doesn't inadvertently collapse cardinals, allowing us to build these complex structures step-by-step without the whole edifice crumbling.

Cichon's Diagram: Charting the Continuum's Fine Structure

The question of the continuum's size is just the beginning. The real line, as studied in analysis and topology, has a rich and complex combinatorial structure. This structure can be measured by a collection of numbers known as ​​cardinal characteristics of the continuum​​. These are cardinals, typically between $\aleph_1$ and $2^{\aleph_0}$, that quantify the "difficulty" of certain infinite tasks.

For example, consider two such characteristics related to functions from $\omega$ to $\omega$:

• The ​​bounding number​​, $\mathfrak{b}$, is the smallest size of a family of functions that no single function can eventually dominate. It measures the difficulty of creating an "unpinnable" set of functions.
• The ​​dominating number​​, $\mathfrak{d}$, is the smallest size of a family of functions that eventually dominates every other function. It measures the difficulty of creating an "all-pinning" set.

In ZFC, one can prove $\mathfrak{b} \le \mathfrak{d}$. But are they equal? Forcing provides the answer: not necessarily! We can use specific forcing notions tailored to the task. For example, Hechler forcing is designed to add a single function that dominates all functions from the ground model. By iterating this forcing, we can construct a model where $\mathfrak{b} = \aleph_1$ but $\mathfrak{d} = \aleph_2$. We can also design forcings to manipulate other characteristics, such as the splitting number $\mathfrak{s}$ and the reaping number $\mathfrak{r}$, which relate to combinatorial properties of subsets of $\omega$. Forcing allows us to "pull apart" these characteristics, showing they correspond to genuinely distinct combinatorial notions and revealing an intricate landscape of possibilities for the structure of the real line, a landscape now famously mapped by Cichoń's diagram.

The development of forcing did not stop with Cohen. Modern set theorists have refined these techniques into an art form of incredible subtlety and power. One stunning example is the work of Saharon Shelah, who showed how to separate CH from another principle called the Diamond principle ($\Diamond_{\omega_1}$). The Diamond principle is a strong combinatorial statement that holds in Gödel's $L$ and implies CH. For a long time, it was thought to be inextricably linked to CH.

However, using a highly sophisticated technique called ​​revised countable support iteration​​ with forcings that are "countably closed" (and thus add no new reals), Shelah was able to construct a model of ZFC where CH holds but $\Diamond_{\omega_1}$ fails. This is akin to a master luthier tuning an instrument to produce a very specific, previously unheard chord. It demonstrates the almost surgical precision of modern forcing, allowing us to isolate and separate axioms that once seemed to be part of the same whole.

From its genesis as a tool to answer a single question, Cohen's forcing has blossomed into a rich and profound theory. It has transformed our view of mathematical truth from something static and absolute to something dynamic and relative. It gives us the power not just to observe the mathematical universe, but to take part in its creation, building new worlds that expand the horizons of what we know and what we can imagine.