Set Theory Models

What if our entire mathematical reality was just one of many possibilities? Standard set theory, often ZFC, provides the rules for the game of mathematics, but it doesn't describe a unique playing field. The study of set theory models explores the profound idea that there can be multiple, distinct "universes" of sets, each obeying the same fundamental axioms yet differing in crucial ways. This challenges the intuitive notion of a single, absolute mathematical truth and reveals a more pluralistic and dynamic landscape. This article delves into this fascinating world, explaining how seemingly solid concepts can become relative and how mathematicians can act as architects, building bespoke universes to test the very limits of logic.

The following chapters will guide you through this complex terrain. The "Principles and Mechanisms" section will introduce the core concepts that make this possible, including the relativity of truth via the Skolem paradox, the stabilizing idea of absoluteness, and the two master tools for universe-building: Gödel's constructible universe (L) and Cohen's technique of forcing. Subsequently, the "Applications and Interdisciplinary Connections" chapter will demonstrate the power of these models, showing how they are used to prove the independence of major mathematical questions and reveal surprising connections between set theory, topology, and computer science.

Imagine you are a biologist studying a colony of ants in a sealed glass box. You can see everything they do. You can see their tunnels, their food stores, their queen. From your god-like perspective, you know everything about their world. But what do the ants know? Their world is the glass box. They have no concept of "outside." They might have their own ant-physicists who declare universal laws, but those laws are only universal for the box. This simple idea—the difference between the view from inside and the view from outside—is the key to unlocking the entire, beautiful, and sometimes bewildering world of set theory models.

One of the most jarring and wonderful results in modern logic is the ​​Löwenheim-Skolem theorem​​. In essence, it tells us that if our foundational theory of mathematics (like the standard ​​Zermelo-Fraenkel set theory with Choice​​, or ​​ZFC​​) has any infinite model, or "universe," at all, then it must also have a model that is countable.

Stop and think about that. Our theory, ZFC, proves the existence of uncountable sets. Cantor's theorem, a pillar of mathematics, guarantees that the set of real numbers, $\mathbb{R}$, is staggeringly larger than the set of natural numbers, $\mathbb{N}$. Yet, the Löwenheim-Skolem theorem promises that there is a perfectly good mathematical universe, let's call its domain $M$, that satisfies all the axioms of ZFC, including Cantor's theorem, but which, from the outside, is itself a countable set. You could, in principle, list all the "sets" in this universe $M$ one by one: $m_1, m_2, m_3, \dots$.

This gives rise to the famous ​​Skolem paradox​​. How can a countable universe $M$ believe in uncountable sets? If we pick out the object that $M$ calls "the real numbers," let's call it $\mathbb{R}^M$, its underlying collection of points is a subset of the countable set $M$, and therefore must be countable from our outside perspective! So, we have a set, $\mathbb{R}^M$, that we know is countable, but the inhabitants of universe $M$ are unshakably convinced is uncountable.

The resolution is not a contradiction, but a profound lesson in humility. It all comes down to Alfred Tarski's definition of truth. A statement is true in a model if the witnesses for that statement exist inside the model's domain. When a mathematician in model $M$ says "$\mathbb{R}^M$ is uncountable," what they are really saying is, "There does not exist, within our universe M, any function $f$ that creates a one-to-one correspondence between our natural numbers and our real numbers." And they are absolutely right! The bijection that you and I can see from our vantage point outside the glass box simply isn't an object in $M$. The model $M$ is "missing" the very thing that would reveal the countability of its own set of reals. The notion of "uncountable" is not absolute; it's relative to the universe you inhabit. The paradox dissolves, leaving us with a crucial insight: mathematical truth, when formalized, is always truth relative to a model.

This relativity can be unsettling. Is everything up for grabs? If even cardinality is relative, can we be sure of anything? Thankfully, no. Some concepts are robust enough to mean the same thing inside and outside the box, provided the box is nicely constructed. This property is called ​​absoluteness​​.

The simplest and most important form of this is ​​$\Delta_0$-absoluteness​​. A ​​$\Delta_0$ formula​​ is a statement where all quantifiers are "bounded," meaning they only look inside sets that are already at hand (e.g., "for all $x$ in the set $y$..." or "there exists an $x$ in the set $y$..."). Imagine you're looking into a ​​transitive​​ model $M$—one where if a set $y$ is in the model, all of its elements are too. Transitivity is like the glass box being perfectly transparent, with no hidden compartments. For such a model, any $\Delta_0$ statement is absolute. It is true in $M$ if and only if it is true in our larger universe, $V$. The reason is simple: to check a $\Delta_0$ statement with parameters in $M$, you only need to look at elements of those parameters. Because $M$ is transitive, all those elements are also in $M$, so you and the inhabitant of $M$ are looking at the exact same collection of things.

The Skolem paradox doesn't violate this, because the statement "$X$ is uncountable" is not a $\Delta_0$ formula. It contains an unbounded quantifier: "there does not exist any function $f$..." This quantifier must search the entire universe for a witness. Our universe $V$ and the model's universe $M$ are different, so the search can yield different results.

This idea extends to slightly more complex formulas. A ​​$\Sigma_1$ formula​​ is one that simply asserts the existence of something, like "there exists a $y$ such that...". These formulas are ​​upward absolute​​: if a witness is found inside the transitive model $M$, then it certainly exists in the larger universe $V$ as well. How could it not? $M$ is a part of $V$. However, they are not, in general, downward absolute. Just because a witness exists in $V$ doesn't mean it's inside the smaller world $M$. This simple asymmetry is the engine behind many of set theory's most powerful results.

If we can have different models of set theory, where do they come from? Can we build one? Kurt Gödel gave a spectacular answer with his ​​constructible universe​​, denoted ​​L​​. The idea behind $L$ is one of radical minimalism: build a universe using only the tools you are explicitly given by the axioms, adding nothing extraneous.

The construction proceeds in stages, one for each ordinal number $\alpha$. We start with nothing and build layer by layer:

• ​​Stage 0​​: $L_0 = \emptyset$. We start with the empty set.
• ​​Successor Stage​​: To get from $L_\alpha$ to $L_{\alpha+1}$, we add every subset of $L_\alpha$ that can be defined using a first-order formula with parameters already in $L_\alpha$. We denote this collection of definable subsets as $\mathrm{Def}(L_\alpha)$, so $L_{\alpha+1} = \mathrm{Def}(L_\alpha)$. This step relies on the ​​Axiom of Power Set​​ (to have a collection of all subsets to pick from) and the ​​Axiom Schema of Separation​​ (to actually pick out the definable ones).
• ​​Limit Stage​​: For a limit ordinal $\lambda$ (like $\omega$, the first infinite ordinal), we simply gather everything we have built so far: $L_\lambda = \bigcup_{\beta < \lambda} L_\beta$.

Here, we encounter one of the most powerful and subtle axioms: the ​​Axiom Schema of Replacement​​. To form $L_\lambda$, we need to know that the collection of all previous stages, $\{L_\beta \mid \beta < \lambda\}$, is itself a set, so that we can apply the ​​Axiom of Union​​ to it. But $\lambda$ could be enormous! How can we be sure this collection isn't "too big" to be a set? Replacement is the guarantee. It says that if you have a set as your domain (here, the ordinal $\lambda$) and a definable function (here, $\beta \mapsto L_\beta$), the range of that function is also a set. It is this axiom that allows the construction of the cumulative hierarchy to climb to arbitrarily large heights.

The final constructible universe $L$ is the union of all the stages, $L = \bigcup_{\alpha \in \mathrm{Ord}} L_\alpha$. Because this construction uses all the ordinals, and the class of all ordinals ($\mathrm{Ord}$) is not a set, $L$ itself is a ​​proper class​​. It's a universe so vast it cannot be contained within a single one of its own sets. We call it an ​​inner model​​: a transitive, definable proper class that satisfies all the axioms of ZFC. Gödel showed that this minimalist universe also happens to satisfy the Axiom of Choice (AC) and the Continuum Hypothesis (CH).

Gödel's $L$ is a beautiful, rigid, and orderly universe. But what if we want to see what a messier universe looks like—one where the Continuum Hypothesis is false? For decades, this question was unanswered. Then, in the 1960s, Paul Cohen invented a revolutionary technique to do just that: ​​forcing​​.

Forcing is a method for starting with a model $M$ (say, Gödel's $L$) and gently adjoining a new object to create a larger universe $M[G]$. The key is that this new object, $G$, is "generic" over $M$. What does this mean?

Imagine the process as a dialogue between the old universe $M$ and us, the builders in the meta-universe.

1. ​​The Set of Rules​​: We start with a ​​partial order​​ $\mathbb{P}$ inside $M$. You can think of its elements as "conditions" or "pieces of information" about the new object we want to add. For example, if we want to add a new real number, the conditions might be finite strings of 0s and 1s, representing initial segments of its binary expansion.
2. ​​Forcing a Decision​​: A subset $D \subseteq \mathbb{P}$ is called ​​dense​​ if for any condition you might consider, there is a stronger, more specific condition within $D$. A dense set represents a question that must be answered about our new object (e.g., "Will the 10th digit be 0 or 1?").
3. ​​The Generic Object​​: We now need to build a consistent set of answers. A ​​filter​​ $G$ on $\mathbb{P}$ is a set of compatible conditions. An ​​$M$-generic filter​​ is a special kind of filter that doesn't live in $M$. It's chosen from the "outside" in such a way that it intersects every dense set that exists in M. In other words, our generic object $G$ provides an answer to every question that the original universe $M$ can think to ask.

The new universe, $M[G]$, is the smallest transitive model containing both $M$ and our new generic object $G$. By carefully choosing our partial order $\mathbb{P}$, we can force the new universe to have all sorts of amazing properties. To make the continuum larger than $\aleph_1$, Cohen designed a poset that adds $\aleph_2$ new real numbers. The new sets added by forcing serve as witnesses to statements that were false in the old model. This is how the independence of the Continuum Hypothesis was finally proven. We now have two equally valid models of ZFC: Gödel's $L$ where CH is true, and Cohen's $L[G]$ where CH is false.

All of this—Skolem's paradox, Gödel's L, Cohen's forcing—relies on using ​​first-order logic​​. This is a language where we can quantify over individuals (sets) but not over properties (classes of sets). At first glance, this seems like a limitation. Why not use ​​second-order logic​​, where we can directly quantify over all subsets and relations? A statement like "for every property $X$..." seems much more powerful.

And it is. But its power is its downfall. The meaning of a second-order statement like "$\exists X \psi(X)$" depends critically on what "all possible subsets $X$" means. As we saw with the Skolem paradox, the collection of all subsets of a set $D$, its power set $\mathcal{P}(D)$, is relative to the model. The universe $V$ might contain many more subsets of $D$ than the inner model $M$ does. The collection $\mathcal{P}^M(D)$ can be a proper subcollection of $\mathcal{P}^V(D)$.

This means that a second-order sentence can be true when interpreted in $M$ but false when interpreted in $V$. The notion of "validity" (truth in all possible structures) becomes non-absolute. Two mathematicians living in different set-theoretic universes could disagree on which second-order statements are logically true!

First-order logic, by being "weaker," is more robust. Its apparent paradoxes are not flaws, but features that reveal the deep, relative nature of mathematical existence. They force us to be precise about what we mean by "true" and "exists." They open the door to a pluralistic vision of the mathematical world, a stunning landscape of different, internally consistent universes, all of which we can explore with the beautiful and subtle tools of model theory.

We have spent some time laying down the axioms of set theory, the formal rules of the game. It might have felt like we were very carefully describing the properties of, say, the game of chess—how the knight moves, what a checkmate is, and so on. But the real fun of chess is not in reciting the rules, but in playing the game! What happens when we actually move the pieces? What surprising strategies and beautiful patterns emerge?

In this chapter, we will finally play the game. We will take the axioms and the concept of a "model" or "universe" of sets and see where they lead. And I must warn you, the journey is stranger and more wonderful than you might imagine. We will find that seemingly solid mathematical truths can become relative, that our intuition about counting can be led astray, and that questions about the familiar real number line can have answers that depend on the existence of infinities so vast they are difficult to describe. We will see how mathematicians, like cosmic architects, can build bespoke universes to test the limits of logic itself. Let us begin.

One of the most profound lessons we learn from studying models of set theory is that many mathematical concepts are not as absolute as they appear. Their properties can depend on the universe you are looking at them from.

The Skolem Paradox: Is ‘Big’ Really Big?

Let’s start with a classic mind-bender. We know ZFC proves that the set of real numbers, $\mathbb{R}$, is uncountable. That is, there is no surjection from the set of natural numbers $\omega$ onto $\mathbb{R}$. Now, the Löwenheim-Skolem theorem from first-order logic tells us something astonishing: if ZFC has a model at all, it must have a model that is countable from the outside.

Let that sink in. We have a universe of sets, let’s call it $N$, and we, from our vantage point in a larger universe $V$, can count every single set in $N$. There’s the first set, the second, the third... and we can list them all. And yet, this model $N$ is a perfectly good model of ZFC. So, inside $N$, all the theorems of ZFC are true. In particular, the people living inside $N$ will look at their version of the real numbers, $\mathbb{R}^N$, and they will prove, using Cantor's diagonal argument, that it is uncountable!

How can this be? How can a set like $\mathbb{R}^N$, which we know is countable from the outside (it's a subset of the countable model $N$), be "uncountable" to those inside? The resolution of this "Skolem's Paradox" is a beautiful lesson in relativity. The statement "$\mathbb{R}^N$ is uncountable" means "there exists no surjective function from $\omega$ to $\mathbb{R}^N$ within the model N." The bijection that we can see from our external viewpoint simply does not exist as an object inside the model $N$. Countability is not an absolute property of a set; it is relative to the universe of available functions. The model $N$ is just too poor to contain the very function that would reveal its countability.

A World Without Choice

The Axiom of Choice (AC) seems intuitive—if you have a collection of non-empty bins, you can surely pick one item from each. For a long time, mathematicians used it implicitly. But what happens if we decide to play the game of mathematics without it? We can build models of ZF (ZFC without Choice) where AC fails, and the world looks very different.

For instance, in a course on linear algebra, you learn that every vector space has a basis (a Hamel basis). The proof almost always invokes Zorn's Lemma, which is equivalent to the Axiom of Choice. Is AC really necessary? It turns out it is! The statement "every vector space has a basis" is, over ZF, fully equivalent to the Axiom of Choice. By building special models of ZF, like the one constructed by Robert Solovay, we can create mathematical universes where AC fails. In such a universe, it is a consistent statement that the set of real numbers $\mathbb{R}$, viewed as a vector space over the rational numbers $\mathbb{Q}$, does not have a Hamel basis. A fundamental theorem of algebra is not a universal truth, but a consequence of a specific, optional axiom.

The weirdness doesn't stop there. What could be more certain than the fact that a countable union of countable sets is countable? To prove this, you imagine laying out the countable sets in rows and then traversing the resulting grid diagonally. But this "process" hides a crucial step: to lay out the sets in rows, you must first have an enumeration (a counting function) for each set. Choosing one such enumeration for each of the countably many sets requires the Axiom of Countable Choice ($\mathsf{AC}_\omega$), a weaker form of AC. By building clever models known as symmetric or permutation models, we can construct a universe where $\mathsf{AC}_\omega$ fails. In this universe, it is possible to have a countable collection of pairs, whose union is, from the model's internal perspective, an uncountable set. Our most basic intuition about counting breaks down.

The Universe in a Mirror: Second-Order Logic and Large Cardinals

The real number line seems like a solid, well-understood object. But its deeper properties are surprisingly sensitive to the rest of the set-theoretic universe. This becomes apparent when we use second-order logic, where we can quantify not just over numbers, but over sets of numbers.

Consider the statement $\varphi$: "Every subset of the real numbers has the Baire property" (meaning it's 'close' to an open set in a precise way). Is this true? The answer is: it depends!

• In Gödel's constructible universe $L$, a relatively sparse model of ZFC, the Axiom of Choice allows us to construct "pathological" sets of reals that lack the Baire property. So in $L$, the sentence $\varphi$ is false.
• However, in Solovay's model, a lush universe where AC fails, every set of reals does have the Baire property. In this model, $\varphi$ is true.

The truth of a single sentence about $\mathbb{R}$ changes depending on the universe it's uttered in. The reason is that the quantifier "Every subset of $\mathbb{R}$" means different things in different models. In $L$, it means "every set of reals in L", while in Solovay's model, it means "every set of reals in Solovay's model". These are different collections of sets, and so the truth of the statement about them differs.

The connection gets even more breathtaking. The properties of the real number line can be tied to axioms about the existence of incredibly large infinite numbers, called "large cardinals". Consider the statement $\psi$: "There is no 'simple' (specifically, $\Delta^1_2$) way to well-order the real numbers."

• In the minimal universe $L$, where no large cardinals exist, there is a relatively simple, definable well-ordering of the reals. So $\psi$ is false.
• However, if we are in a universe where a "measurable cardinal" exists—an infinity with properties so strong its existence cannot be proven in ZFC—then it's a theorem that no such simple well-ordering of the reals can exist. In this universe, $\psi$ is true.

The truth of a statement about the structure of $\mathbb{R}$ is decided by an axiom about the existence of infinities far beyond it in the cosmic hierarchy. The real number line acts like a mirror, reflecting the properties of the entire universe of sets.

So far, we have been like tourists visiting strange new worlds. But the real power of model theory is that we are not just visitors; we are architects. We can construct these models to have specific properties, primarily to prove that certain statements are independent of our base axioms (like ZFC).

The two main tools for this are ​​inner models​​ and ​​forcing​​.

• An ​​inner model​​ is a sub-universe contained within a larger one. The most famous is Gödel's constructible universe, $L$. By showing that $L$ is a model of ZFC but also satisfies statements like the Continuum Hypothesis (CH), Gödel proved that if ZFC is consistent, then ZFC+CH is also consistent. The universe $L$ is a beautiful, minimalist reality built from the ground up, where the Axiom of Choice is not an axiom but a provable theorem derived from a canonical well-ordering of the entire universe.
• ​​Forcing​​ is a revolutionary technique developed by Paul Cohen. It allows us to start with a model of set theory and judiciously "add" new sets to it to create a larger model. For example, we can start with a model where CH is true (like $L$) and "force" it to contain many new real numbers, building a larger universe where CH is false. This proved that CH is independent of ZFC, finally solving Hilbert's first great problem. Forcing can be used to "collapse" cardinals, introduce sets with strange properties, or make the universe of sets wider or longer in carefully controlled ways.

By building one model where a statement $\varphi$ is true and another where it is false, we prove that $\varphi$ can neither be proven nor disproven from the ZFC axioms. This is the ultimate application of set theory models: mapping the boundaries of mathematical provability.

The study of set theory models doesn't just look inward; it builds surprising and beautiful bridges to other areas of mathematics and computer science.

Logic and Topology: A Compact Argument

What does logic have to do with geometry? At first glance, not much. But consider the Compactness Theorem of propositional logic. It states that if every finite subset of a collection of logical axioms has a model (a satisfying truth assignment), then the entire infinite collection has a model.

The proof of this theorem has a stunning topological translation. Imagine the space of all possible truth assignments for a countable set of formulas. This space can be viewed as an infinite product of discrete two-point spaces, $\{0, 1\}^{\mathcal{F}}$. By Tychonoff's theorem, a cornerstone of general topology, this space is compact. The condition that a valuation must be a "model" for a given set of axioms defines a closed subset of this space. The premise of the Compactness Theorem translates to the statement that the family of these closed sets has the finite intersection property. In a compact space, this guarantees that the total intersection is non-empty. Any point in that intersection is a truth assignment that satisfies all the axioms simultaneously. The existence of a logical model is guaranteed by the topological property of compactness!

Logic and Computation: The Unknowable Theory

Let's say we had access to a perfect, completed model of set theory, $V$. We could ask any question about sets, and the model would contain the answer. Could we, then, write a computer program to list all the true statements about this universe? Could we enumerate its complete theory, $Th(V)$?

The answer is a resounding no, and it connects set theory to the fundamental limits of computation discovered by Gödel and Turing. True arithmetic, the set of all true statements about the natural numbers, is not recursively enumerable—no computer program can list all and only these truths. Because we can translate statements about arithmetic into statements about sets, if we could enumerate the complete theory of $V$, we could also enumerate true arithmetic, which is impossible. This is a consequence of a deep result known as Tarski's Undefinability of Truth, which states that no sufficiently strong system can define its own truth predicate. Even if we live in a "platonic" mathematical reality, its complete description is fundamentally uncomputable.

The Limits of Change: Absoluteness

After seeing how much can change from one model to another, it's natural to wonder if anything is stable. Is all of mathematics built on shifting sands? Fortunately, no. There are profound "conservation laws" that limit the power of techniques like forcing.

Shoenfield's Absoluteness Theorem is a prime example. It states that certain kinds of sentences—those of a specific logical complexity known as $\Sigma^1_2$ and $\Pi^1_2$—are "absolute" between a model and its forcing extensions. This means that if such a sentence is true in your original universe, no amount of forcing can make it false, and vice-versa. This is remarkable. While forcing can add new reals, change the value of the continuum, and break the Axiom of Choice, it cannot change the truth of these particular sentences. There is a core of mathematical reality, at least concerning the projective hierarchy on the reals, that is robust and invariant. This provides a deep sense of unity and structure amidst the wild diversity of the set-theoretic multiverse.

Our journey has shown us that the world of sets is not a single, static museum of facts, but a dynamic, interconnected landscape of possibilities. By exploring different models, we learn the true meaning of our axioms, discover the limits of proof and computation, and uncover deep, unifying principles that tie together disparate branches of human thought. The game of mathematics is indeed a grand one, and its board is far larger than we ever thought.