Going Up Theorem

In the abstract world of commutative algebra, rings of numbers and polynomials form intricate structures. A central challenge lies in visualizing these structures and understanding how they relate to one another. One of the most powerful techniques is to translate the algebraic properties of a ring into the geometric language of a space, known as its prime spectrum. But what happens when one ring is contained within another? How does the geometry of the larger space relate to the smaller one? This article delves into this fundamental question, exploring the profound connection forged by integral extensions. We will first uncover the core principles, beginning with the Lying Over theorem and culminating in the powerful Going Up theorem, which provides a ladder for climbing between the prime ideal structures of the two rings. Following this, we will explore the remarkable applications of these ideas, demonstrating how they unify concepts in algebraic geometry, number theory, and the study of symmetries, revealing a deep structural coherence across mathematics.

Imagine you are a cartographer, but instead of mapping mountains and rivers, you are mapping the abstract landscapes of algebra. Our territories are not made of rock and soil, but of numbers and polynomials, structured into what mathematicians call ​​rings​​. How can we draw a map of such a place? It turns out that a beautiful way to visualize a ring $R$ is to create a geometric space from it, called its ​​prime spectrum​​, which we denote as $\mathrm{Spec}(R)$. The "points" of this space are not points in the usual sense, but special subsets of the ring called ​​prime ideals​​.

Think of the ring of integers, $\mathbb{Z}$. Its prime ideals are the sets of multiples of prime numbers, like $\{..., -6, -3, 0, 3, 6, ...\}$, which we write as $\langle 3 \rangle$, and also the "trivial" prime ideal containing only zero, $\langle 0 \rangle$. These are the points on our map of $\mathbb{Z}$.

Now, what happens if we have a larger ring $S$ that contains a smaller ring $R$? For example, the ring of Gaussian integers $\mathbb{Z}[i]$, consisting of numbers like $a+bi$, contains the ordinary integers $\mathbb{Z}$. This gives us two maps, one for $S$ and one for $R$. Naturally, we want to know how these two maps relate. Is there a way to project the map of $S$ onto the map of $R$? Indeed, there is. Any "point" $\mathfrak{q}$ in the space $\mathrm{Spec}(S)$ can be projected down to a point in $\mathrm{Spec}(R)$ simply by taking the intersection: $\mathfrak{p} = \mathfrak{q} \cap R$. This gives us a natural map between our algebraic landscapes, $f: \mathrm{Spec}(S) \to \mathrm{Spec}(R)$.

This chapter is the story of this map. We will explore when this map has beautiful properties, and what those properties tell us about the deep connection between the rings $R$ and $S$. The key to this entire story lies in a single, crucial condition: the extension from $R$ to $S$ must be ​​integral​​.

What does it mean for an extension $R \subseteq S$ to be integral? It means that every element in the larger ring $S$ is "tethered" to the smaller ring $R$ in a very specific way. Each element $s \in S$ must be a root of a monic polynomial—a polynomial whose leading coefficient is 1—with all its other coefficients coming from $R$.

The classic example is the extension $\mathbb{Z} \subseteq \mathbb{Z}[i]$. An element like $s = 1+i$ in $\mathbb{Z}[i]$ might seem to live freely, but it is actually bound to $\mathbb{Z}$ by the equation $s^2 - 2s + 2 = 0$. Since this is a monic polynomial with integer coefficients, $1+i$ is integral over $\mathbb{Z}$. In fact, every element of $\mathbb{Z}[i]$ is integral over $\mathbb{Z}$. This "integrality" condition acts like a set of invisible ropes, ensuring that $S$ doesn't stray too far from $R$. As we'll see, this tether has profound geometric consequences.

Our map $f: \mathrm{Spec}(S) \to \mathrm{Spec}(R)$ projects points downwards. A fundamental question arises immediately: does every point in the base space $\mathrm{Spec}(R)$ have at least one point in $\mathrm{Spec}(S)$ lying above it? In other words, is our map ​​surjective​​?

Without the tether of an integral extension, the answer can easily be no. But with it, the answer is a resounding "yes!" This is the content of the first major result on our journey, the ​​Lying Over Theorem​​. It states that if $R \subseteq S$ is an integral extension, then for any prime ideal $\mathfrak{p}$ of $R$, there exists a prime ideal $\mathfrak{q}$ of $S$ such that $\mathfrak{q} \cap R = \mathfrak{p}$. Geometrically, this means our projection map covers the entire base space; no point below is left without a corresponding point above it.

Let's return to our cartographic example, $\mathbb{Z} \subseteq \mathbb{Z}[i]$. Consider the point $\langle 29 \rangle$ in $\mathrm{Spec}(\mathbb{Z})$. Does anything lie over it? The number 29, a prime in $\mathbb{Z}$, can be factored in $\mathbb{Z}[i]$ as $29 = (5+2i)(5-2i)$. This algebraic fact has a geometric meaning: the point $\langle 29 \rangle$ "splits" into two distinct points above it in $\mathrm{Spec}(\mathbb{Z}[i])$: the prime ideal $\langle 5+2i \rangle$ and the prime ideal $\langle 5-2i \rangle$. You can check that both of these, when intersected with $\mathbb{Z}$, give you back exactly $\langle 29 \rangle$.

Not all points behave this way. The prime ideal $\langle 2 \rangle$ in $\mathbb{Z}$ has only one prime ideal lying over it in $\mathbb{Z}[i]$, namely $\langle 1+i \rangle$. We say that 2 ​​ramifies​​. The prime ideal $\langle 3 \rangle$, on the other hand, corresponds to the ideal $\langle 3 \rangle$ in $\mathbb{Z}[i]$, which remains prime. We say 3 ​​remains inert​​. The Lying Over theorem guarantees that something is up there; the arithmetic of the rings determines whether it's one point, two, or more.

The Lying Over theorem gives us a starting point. It tells us we can lift individual points. But what about paths? In geometry, we are interested not just in points, but in curves and dimensions. In our algebraic landscape, a "path" is a chain of prime ideals, where each is properly contained in the next: $\mathfrak{p}_0 \subsetneq \mathfrak{p}_1 \subsetneq \dots \subsetneq \mathfrak{p}_n$.

Suppose we have such a chain in our base space $\mathrm{Spec}(R)$. And suppose we pick a starting point $\mathfrak{q}_0$ in $\mathrm{Spec}(S)$ that lies over the start of our path, $\mathfrak{p}_0$. Can we always find a corresponding path in $\mathrm{Spec}(S)$ that starts at $\mathfrak{q}_0$ and lies perfectly over the path in $\mathrm{Spec}(R)$?

This is the question answered by the ​​Going Up Theorem​​. It guarantees that for an integral extension, we can always "go up". Given a chain of primes $\mathfrak{p}_0 \subsetneq \mathfrak{p}_1$ in $R$ and a prime $\mathfrak{q}_0$ in $S$ lying over $\mathfrak{p}_0$, there exists a prime $\mathfrak{q}_1$ in $S$ such that $\mathfrak{q}_0 \subsetneq \mathfrak{q}_1$ and $\mathfrak{q}_1$ lies over $\mathfrak{p}_1$. By applying this repeatedly, we can lift an entire chain of any length.

Let's see this in action. In $\mathbb{Z}$, we have the simple chain $\langle 0 \rangle \subsetneq \langle 29 \rangle$. As we saw, the prime ideal $\langle 5+2i \rangle$ in $\mathbb{Z}[i]$ lies over $\langle 29 \rangle$. The zero ideal $\langle 0 \rangle$ in $\mathbb{Z}[i]$ lies over the zero ideal $\langle 0 \rangle$ in $\mathbb{Z}$. Thus, we have successfully lifted the chain from the base ring to a chain in the larger ring: $\langle 0 \rangle \subsetneq \langle 5+2i \rangle$. This process can even be stacked. If we have a tower of integral extensions, like $\mathbb{Z} \subseteq \mathbb{Z}[i] \subseteq \mathbb{Z}[i][\sqrt{2}]$, we can "go up" in two steps, lifting a chain from $\mathbb{Z}$ all the way to $\mathbb{Z}[i][\sqrt{2}]$.

So, we can lift points, and we can lift paths. What is the grand consequence of all this? The Going Up theorem is usually paired with a sibling property called ​​Incomparability​​, which states that if two distinct primes in $S$ are nested ($\mathfrak{q}_0 \subsetneq \mathfrak{q}_1$), then their projections in $R$ must also be distinct and nested ($\mathfrak{p}_0 \subsetneq \mathfrak{p}_1$). You can't have two different levels on the ladder in $S$ projecting to the same level in $R$.

Together, Going Up and Incomparability lead to a stunning conclusion. Let's define the ​​Krull dimension​​ of a ring as the length of the longest possible chain of prime ideals. This is a measure of the "height" or "complexity" of our algebraic space. The theorems imply that for an integral extension of integral domains, the longest chain in $S$ has the same length as the longest chain in $R$. In other words: $\dim(S) = \dim(R)$ Integral extensions preserve dimension! This is a profound statement of rigidity. The tether of integrality is so strong that the fundamental geometric dimension of the two spaces must be identical.

This has beautiful consequences. For an integral domain, being a ​​field​​ (a ring where every nonzero element has a multiplicative inverse) is equivalent to having Krull dimension 0. Why? Because in a field, the only proper ideal is the zero ideal $\langle 0 \rangle$, so the longest chain has length 0. The dimension preservation immediately tells us that if $R \subseteq S$ is an integral extension of domains, then $S$ is a field if and only if $R$ is a field. If one has dimension 0, the other must too.

The power of these theorems makes it tempting to ask: is the "integral" condition really necessary? What if we have a ring extension that is not integral? The entire beautiful structure can collapse.

Consider the ring $S=k[x,y]$ (polynomials in two variables over a field $k$) and its subring $R = k[y, xy]$. The element $x \in S$ is not integral over $R$. Here, the Going Up property fails. One can construct a chain of prime ideals in $R$ that cannot be lifted. For instance, there is a prime ideal $\mathfrak{p}_1$ in $R$ and a prime $\mathfrak{q}_1$ in $S$ lying over it, but for a larger prime ideal $\mathfrak{p}_2$ containing $\mathfrak{p}_1$, no prime ideal $\mathfrak{q}_2$ exists in $S$ that both contains $\mathfrak{q}_1$ and lies over $\mathfrak{p}_2$. The ladder is broken. This shows that integrality is not just a technicality; it is the essential glue holding our geometric correspondence together.

Let's take one last look at our map $f: \mathrm{Spec}(S) \to \mathrm{Spec}(R)$, armed with our new understanding. The Lying Over and Going Up theorems are not just isolated algebraic curiosities. They have a powerful, unified meaning in the language of topology.

The Zariski topology is the natural topology on the prime spectrum. A central result, which encapsulates much of what we've discussed, states that for an integral extension, the map $f$ is a ​​closed map​​. This means that it sends closed sets in $\mathrm{Spec}(S)$ to closed sets in $\mathrm{Spec}(R)$.

Why is this so elegant? A continuous, surjective, and closed map is a very special thing in topology: a ​​quotient map​​. Intuitively, this means that the topology of the base space $\mathrm{Spec}(R)$ is completely and faithfully determined by the topology of $\mathrm{Spec}(S)$ via the projection map $f$. The base space isn't some distorted or incomplete shadow; it is a perfect quotient image of the larger space. The algebraic tether of integrality ensures that the geometric projection is as well-behaved and informative as one could possibly hope for. From a simple algebraic condition springs a deep and elegant geometric unity.

In our previous discussion, we uncovered the machinery of integral extensions and the elegant logic of the Going Up theorem. We saw that when one ring, $S$, sits "integrally" over another, $R$, a surprisingly tight bond forms between them. The Going Up theorem provides the beautiful guarantee that any ascending path of prime ideals you can trace in the simpler ring $R$ can be "lifted" to a corresponding path in the more complex ring $S$.

At first glance, this might seem like a niche rule for algebraists, a piece of intricate but isolated clockwork. But what is it for? Where does this abstract guarantee find its purchase in the broader landscape of science and thought? The answer is as profound as it is surprising. This single principle acts as a unifying thread, weaving together the geometric study of shapes, the arithmetic of numbers, and the subtle dance of symmetry. Let's embark on a journey to see how this one idea blossoms into a powerful tool across mathematics.

Imagine space. Not necessarily the three-dimensional space we live in, but a more abstract, topological space defined by the algebraic properties of a ring. In this world, the "points" are prime ideals, and the "dimension" of the space, its Krull dimension, is measured by the longest possible chain of nested prime ideals you can find. A one-dimensional space is like a line, where you can't have more than two nested points (ideals), say $(0) \subsetneq \mathfrak{p}$. A two-dimensional space is like a surface, admitting longer chains.

Now, suppose we have a one-dimensional space, let's call its ring $R$. What happens if we form a new space, with ring $S$, that is an integral extension of $R$? Geometrically, you can picture $\operatorname{Spec}(S)$ as being projected onto $\operatorname{Spec}(R)$. The "integral" property ensures this projection is well-behaved. The Going Up theorem, along with its cousin the Incomparability theorem, leads to a startlingly rigid conclusion: the dimension of $S$ must be the same as the dimension of $R$.

If $\dim(R) = 1$, then $\dim(S) = 1$. You simply cannot construct a "surface" that sits integrally over a "line." Any attempt to create a longer chain of prime ideals in $S$, for instance, a chain of length two like $\langle 0 \rangle \subsetneq \mathfrak{q}_1 \subsetneq \mathfrak{q}_2$, is doomed to fail. Such a chain would imply that $\dim(S)$ is at least two, but the theorems of integral extensions lash its dimension firmly to that of $R$. This isn't just a technicality; it's a conservation law for dimension. The process of integral extension, while potentially adding immense complexity, cannot arbitrarily inflate the underlying dimensionality of the space.

Let's take this geometric intuition into a different domain: the study of symmetry. Consider a set of polynomials in several variables, like $R = k[x_1, \dots, x_n]$. This ring corresponds to a simple $n$-dimensional space. Now, imagine a finite group of symmetries, $G$, that acts on this space, shuffling the variables around according to some rules. Some special polynomials are left completely unchanged—they are "invariant" under all the symmetries. These polynomials form their own ring, the ring of invariants, denoted $R^G$.

This ring of invariants, $R^G$, often has a structure that is far more complex and mysterious than the original polynomial ring $R$. Finding an explicit description of its generators was a central, and famously difficult, problem of 19th-century mathematics. Yet, the theory of integral extensions provides a stunningly simple insight. For any finite group $G$, the larger ring $R$ is always an integral extension of its invariant subring $R^G$.

And here’s the punchline: our conservation law for dimension applies directly. Because $R$ is integral over $R^G$, their dimensions must be equal.

This is a spectacular result. Without needing to compute a single invariant polynomial, we know that the "space of symmetries" has the exact same dimension as the original space it acts upon. The abstract machinery of the Going Up theorem cuts through the daunting complexity of invariant theory to reveal a simple, elegant, and powerful truth. It brings a hidden order to what might otherwise appear to be algebraic chaos.

From the world of geometric shapes and symmetries, let's turn to the discrete and ancient realm of numbers. When we move beyond the familiar integers $\mathbb{Z}$, we encounter new systems of arithmetic called number fields. Each number field $K$ has its own "ring of integers," $\mathcal{O}_K$, which are the numbers in $K$ that behave like whole numbers. For example, in the field $\mathbb{Q}(\sqrt{-5})$, the numbers include elements like $1 + \sqrt{-5}$, and its ring of integers is $\mathbb{Z}[\sqrt{-5}]$.

By its very definition, the ring of integers $\mathcal{O}_K$ is the integral extension of the ordinary integers $\mathbb{Z}$ inside the field $K$. This immediately forges a deep connection. We know that the ring $\mathbb{Z}$ has Krull dimension one; its chains of prime ideals are no longer than $\langle 0 \rangle \subsetneq \langle p \rangle$ for some prime number $p$.

Because $\mathcal{O}_K$ is an integral extension of $\mathbb{Z}$, its dimension must also be one. This simple fact, a direct consequence of the Going Up theorem, has monumental implications for the structure of arithmetic. In any integral domain of dimension one, every non-zero prime ideal is automatically a maximal ideal. There's no room for another ideal to be squeezed between a prime and the whole ring. This tells us that the "prime numbers" of these exotic rings (which are now prime ideals) are the fundamental, irreducible atoms of arithmetic, just as they are in $\mathbb{Z}$. This property is a defining characteristic of what are called Dedekind domains, which form the bedrock of modern algebraic number theory. The abstract language of rings and ideals, powered by the Going Up theorem, provides the essential grammar for the story of numbers.

Let us return to geometry, but now equipped with our most powerful tools. How do mathematicians study the intricate shapes—algebraic varieties—carved out in high-dimensional space by systems of polynomial equations? The ring of functions on such a shape, let's call it $A$, can be bewilderingly complex.

Here, one of the most profound results in algebra, the Noether Normalization Lemma, comes into play. It acts like a magical lens. It states that for any such ring $A$ (that is finitely generated over a field or over $\mathbb{Z}$), we can always find a simpler polynomial subring $P$ inside it, such that $A$ is a finite extension of $P$. "Finite" is an even stronger condition than "integral," but it implies it.

Geometrically, this is revolutionary. It means we can view any complicated affine algebraic variety, $\operatorname{Spec}(A)$, as a projection onto a simple, perfectly understood space, the affine space $\operatorname{Spec}(P)$. The integral nature of this projection ensures that it is wonderfully well-behaved. The Lying Over theorem guarantees that the map is surjective: every point in the simple base space has at least one point in the complicated shape lying above it. No part of the base is "missed."

Furthermore, because the extension is finite, something even stronger is true. The set of points lying above any single point in the base space—the "fiber" of the projection—is not just non-empty, it is a finite set of points. Our complicated, perhaps infinite, shape is revealed to be a "finitely-sheeted" cover of a simple grid. This reduces the study of any such shape to the study of a finite map onto the simplest possible space. The theory of integral extensions provides the dictionary for translating geometric properties back and forth between the complex shape and its simplified projection.

From a conservation law for dimension to a tool for understanding symmetry, from the foundation of number theory to a universal method for simplifying geometry, the Going Up theorem reveals its true character. It is not just a theorem; it is a fundamental perspective on mathematical structure, a testament to the deep and often surprising unity of seemingly disparate ideas.