Separated Sets in Topology

In our everyday experience, we intuitively understand when two objects are separate. Two fields with a path between them are distinct and do not overlap. However, in the precise world of mathematics, a more nuanced understanding of "separation" is required. What if the boundaries of these fields touch? Are they still truly separate? The simple idea of being "disjoint" is not enough to capture these subtle but crucial distinctions. The field of topology provides a powerful language to formalize this concept, leading to a deeper understanding of the nature of space itself.

This article delves into the topological concept of separated sets, moving beyond the basic notion of disjointness to explore a more refined measure of apartness. In the first section, "Principles and Mechanisms," we will establish the formal definition of separated sets and explore the hierarchy of topological spaces defined by their separation capabilities, including normal and completely normal spaces, and introduce the pivotal Urysohn's Lemma. Subsequently, in "Applications and Interdisciplinary Connections," we will see how these seemingly abstract ideas have profound consequences, building bridges to analysis through continuous functions and providing foundational principles for geometry, such as the Hyperplane Separation Theorem.

Imagine walking along a path. You see two fields, one on your left and one on your right. They are clearly separate; there is a path between them. They don't overlap. In the language of mathematics, we would say they are ​​disjoint​​. But what if the edge of one field is a stone wall, and the edge of the other is a wire fence, and they are built right up against each other? They are still disjoint—no blade of grass belongs to both fields—but they are "touching". Now imagine two other fields, where the land between them is a gentle, unowned slope. They don't touch, and even their boundaries don't touch. This subtle but crucial difference is at the heart of what mathematicians call ​​separated sets​​.

In topology, we make this idea precise. We don't just care about the points in a set; we also care about its boundary, or what we call its ​​closure​​. The closure of a set, which we denote as $\overline{A}$, is the set $A$ itself plus all of its "limit points"—the points you can get arbitrarily close to while staying within $A$.

Two sets, $A$ and $B$, are ​​separated​​ if neither one contains a point from the other's closure. Formally, this means $\overline{A} \cap B = \emptyset$ and $A \cap \overline{B} = \emptyset$. They are so far apart that you can't even stand on the very edge of one set and be touching the other.

Let’s make this concrete on the real number line, $\mathbb{R}$.

• Consider the sets $A = (0, 1)$ and $B = [1, 2)$. These sets are disjoint. But are they separated? The closure of $A$ is $\overline{A} = [0, 1]$. Now, let's check the condition: $\overline{A} \cap B = [0, 1] \cap [1, 2) = \{1\}$. The point $1$ is in $B$ and also in the closure of $A$. So, these sets are not separated. They are touching at the point $1$.
• Now consider $A = (0, 1)$ and $B = (1, 2)$. Their closures are $\overline{A} = [0, 1]$ and $\overline{B} = [1, 2]$. Let's check our conditions: $\overline{A} \cap B = [0, 1] \cap (1, 2) = \emptyset$, and $A \cap \overline{B} = (0, 1) \cap [1, 2] = \emptyset$. Both conditions hold! These sets are separated.

Notice that any pair of disjoint ​​closed sets​​ is automatically separated. If $A$ and $B$ are closed, then $\overline{A}=A$ and $\overline{B}=B$, so the separation conditions just become $A \cap B = \emptyset$. The idea of separated sets becomes truly interesting when the sets themselves are not closed, like our open intervals $(0, 1)$ and $(1, 2)$. This new definition gives us a more refined tool to measure the "apartness" of sets.

A central game in topology is building walls. Given two distinct sets, can we build a "wall" around each one such that the walls don't intersect? In topology, these "walls" are ​​open sets​​—regions that contain a little bubble of space around each of their points.

Let's start with the most straightforward case: separating two disjoint closed sets. Imagine two islands, $A$ and $B$. They are closed, and they are disjoint. Can we always find two disjoint open regions of the sea, $U$ and $V$, such that island $A$ is entirely within region $U$, and island $B$ is within region $V$?

A topological space where this is always possible for any pair of disjoint closed sets is called a ​​normal space​​ (or a ​​$T_4$ space​​, if it also satisfies a basic point-separation axiom called $T_1$). Our familiar Euclidean space $\mathbb{R}$ is a wonderful example of a normal space. Even if you have two sets of points that get tantalizingly close to each other, you can still build these walls. Consider the set of positive integers, $A = \{1, 2, 3, \dots\}$, and another set $B = \{1+\frac{1}{2}, 2+\frac{1}{3}, 3+\frac{1}{4}, \dots \}$. As $n$ gets large, the point $n \in A$ and the point $n + \frac{1}{n+1} \in B$ get arbitrarily close. Yet, because $\mathbb{R}$ is normal, we are guaranteed that there exist disjoint open sets containing $A$ and $B$. In fact, all metric spaces—spaces where we can define a notion of distance—are normal.

The great mathematician Pavel Urysohn discovered something even more beautiful. In a normal space, you can do more than just build walls. You can sculpt a continuous landscape. For any two disjoint closed sets $A$ and $B$, there exists a continuous function $f: X \to [0, 1]$ that is exactly $0$ on all of set $A$ and exactly $1$ on all of set $B$. It's like creating a smooth terrain that is at sea level ($0$) on island $A$, rises to a plateau of height $1$ on island $B$, and varies continuously everywhere in between. This powerful result is known as ​​Urysohn's Lemma​​.

We've seen that normal spaces allow us to separate disjoint closed sets. But what about our more general ​​separated sets​​? Can we always build walls around any two separated sets?

It turns out the answer is no, not in every normal space. This leads us to a stronger, more robust property. A space where you can build disjoint open walls around any pair of separated sets is called a ​​completely normal space​​ (or a ​​$T_5$ space​​ if it's also $T_1$).

This property has a stunning and deeply important consequence. A space is completely normal if and only if it is ​​hereditarily normal​​—that is, every single subspace of it is also a normal space.

This is a remarkable connection. The property of being able to separate any two separated sets (which feels like a local structural requirement) is perfectly equivalent to the property that the entire space and all of its descendants (subspaces) are well-behaved in the sense of normality. Normality itself is not a hereditary property; a normal space can contain a pathological subspace which is not normal. Complete normality is the cure. It's a guarantee of good behavior that is passed down through the generations.

Just as Urysohn's Lemma works for closed sets in normal spaces, a similar principle applies here: a space is completely normal if and only if for any two separated sets $A$ and $B$, you can define a continuous function $f: X \to [0, 1]$ that is $0$ on $A$ and $1$ on $B$. This establishes a beautiful hierarchy of separation:

• ​​Regular ($T_3$)​​: Can separate a point from a closed set.
• ​​Normal ($T_4$)​​: Can separate two disjoint closed sets.
• ​​Completely Normal ($T_5$)​​: Can separate any two separated sets.

And as you might guess, each level of this hierarchy is stricter than the last: every $T_5$ space is a $T_4$ space, and every $T_4$ space is a $T_3$ space.

Are all "reasonable" spaces completely normal? Not at all. Topology is filled with strange and wonderful counterexamples that test the limits of our intuition.

First, remember that whether sets are closed, open, or separated depends entirely on the ​​topology​​—the collection of open sets we define on our space. In the real line with its standard topology, the rational numbers $\mathbb{Q}$ and the irrational numbers $\mathbb{R} \setminus \mathbb{Q}$ are neither closed nor open. In fact, the closure of each is the entire real line, so they are far from being separated. If we change the rules and use a different topology, like the lower limit topology, the situation remains the same: their closures are still the whole space, so the question of separating them as closed sets is ill-posed.

More dramatically, there are spaces where even disjoint closed sets cannot be separated. The famous ​​Moore plane​​ (or Niemytzki plane) is one such example. It consists of the upper half of the Cartesian plane, including the $x$-axis. The topology is standard for points with $y > 0$, but for points on the $x$-axis, the open "bubbles" are open disks in the upper plane that are tangent to the axis at that point. In this strange space, the set of points on the x-axis with rational coordinates, $C_1 = \{(x, 0) | x \in \mathbb{Q}\}$, and the set with irrational coordinates, $C_2 = \{(x, 0) | x \in \mathbb{R} \setminus \mathbb{Q}\}$, are both disjoint closed sets. Yet, it is impossible to construct disjoint open sets around them. Any open "wall" built around the rationals on the axis inevitably "spills over" and touches any wall built around the irrationals. The Moore plane is therefore not normal, and consequently, it cannot be completely normal.

These examples are not just mathematical curiosities. They are the proving grounds where our definitions are forged and refined. They teach us that concepts like "separated" and "normal" are not absolute but are deeply tied to the structure of the space itself. By studying when and why we can build these conceptual walls, we gain a profound insight into the very nature of shape and continuity.

We have explored the formal definition of separated sets, a concept that might at first seem like a rather abstract piece of mathematical housekeeping. But to leave it at that would be like learning the rules of chess without ever seeing the beauty of a grandmaster's game. The real power and elegance of separated sets lie not in the definition itself, but in what it allows us to do. It is a key that unlocks a surprising number of doors, revealing deep and often beautiful connections between the seemingly disparate worlds of topology, analysis, and geometry. The ability—or inability—to separate sets tells us profound things about the very fabric of the space we are studying.

Let's embark on a journey to see how this simple idea of "placing a barrier" between two sets becomes a powerful tool for understanding our mathematical universe.

One of the most profound results in all of topology is a theorem that acts as a magical bridge between the world of spatial structure (topology) and the world of continuous change (analysis). This is the famous Urysohn's Lemma. Imagine a topological space as a landscape. We have two disjoint closed sets, say, two islands, $A$ and $B$. The space is "normal" if we can always find two disjoint open "moats" of water, $U$ and $V$, one surrounding each island. This is a statement purely about the layout of the space.

What Urysohn's Lemma tells us is something astonishing: if a space is normal, we can do more than just dig moats. We can actually build a continuous "landscape" over the entire space—a continuous function $f$ that maps every point to a real number between 0 and 1. This function can be constructed so that it has an "elevation" of exactly 0 everywhere on island $A$ and an elevation of exactly 1 everywhere on island $B$.

Think about what this means. It converts a topological property (the ability to separate with open sets) into an analytical one (the existence of a continuous function). Suddenly, we can use the powerful tools of calculus and analysis to study topological spaces. This ability to construct "separating functions" is a cornerstone of modern analysis and geometry. We can even get creative and transform this function; for example, by composing it with another function like $g(x) = \sin^2(\frac{\pi}{2}f(x))$, we create a new Urysohn function that separates not just the original islands, but the entire set of points at "sea level," $f^{-1}(0)$, from the set of all "mountain peaks," $f^{-1}(1)$.

Of course, this magic doesn't work in every space. There are strange topological worlds where this separation is impossible. Consider the so-called K-topology on the real number line. In this space, the set of points $B = \{1, 1/2, 1/3, \dots\}$ and the point $A=\{0\}$ are both closed and disjoint. Yet, they cannot be separated by disjoint open sets. Any open "blanket" we try to wrap around the set $B$ will inevitably have threads that get snagged on points arbitrarily close to 0, preventing us from creating a "safe zone" around 0 that is disjoint from the blanket. Such examples are not just curiosities; they are crucial for understanding why theorems have conditions and for appreciating the delicate structure of "nice" spaces where separation is possible.

Let's move from the abstract realm of general topology to the more familiar world of Euclidean geometry. Here, the idea of separation takes on a very concrete meaning, especially in the study of convex sets—shapes without any dents or holes, like a disk, a solid cube, or a half-plane. The idea of separating convex sets with a straight line (or a plane in 3D, a "hyperplane" in general) is fundamental to fields ranging from computer graphics and robotics to economics and machine learning.

The ​​Hyperplane Separation Theorem​​ is the geometric cousin of the topological separation axioms. It states that if you have two disjoint, non-empty convex sets in $\mathbb{R}^n$, you can always find a hyperplane that sits between them, with one set on one side and the other set on the other.

But the devil, as always, is in the details. Consider two closed disks in the plane that are tangent at a single point, like two coins touching. They are convex, but their intersection is not empty. Can they be separated? Yes, by the common tangent line. But can they be strictly separated, meaning the line touches neither? No. The point of tangency lies on any potential separating line, preventing strict separation. A similar, more subtle situation arises with two disjoint convex sets that get "infinitesimally close" to each other. Even if they don't touch, if the distance between them is zero, no "open slab" of positive thickness can fit between them, making strict separation impossible. These examples highlight the crucial difference between separation ($\le$) and strict separation ($\lt$).

The assumption of convexity is not a mere technicality; it is the heart of the theorem. If we drop it, separation is no longer guaranteed. Imagine two sets, $A = \{(x,y) \mid y > x^3\}$ and $B = \{(x,y) \mid y x^3\}$. These two regions are disjoint, but they are "intertwined" around the curve $y=x^3$ in such a way that no single straight line can be drawn to separate them. Any line you draw will inevitably cut through both regions. This illustrates a beautiful principle: convexity provides a kind of "global smoothness" that prevents sets from wrapping around each other in complex ways, thus permitting simple linear separation.

Returning to topology, we can ask what happens when we push the idea of separation to its limits.

What if we want to separate not just disjoint closed sets, but any two "separated sets" $A$ and $B$ (where $\overline{A} \cap B = \emptyset$ and $A \cap \overline{B} = \emptyset$)? Spaces that allow this are called ​​completely normal​​. This stronger property has beautiful consequences. For instance, in any space, distinct connected components—the fundamental "pieces" of the space—are always separated sets. Therefore, in a completely normal space, we are guaranteed that we can always place these distinct pieces into their own disjoint open neighborhoods.

We can also demand a stronger type of separation. Instead of a smooth ramp from 0 to 1 (a Urysohn function), what if we demand a function that acts like a light switch—it is exactly 0 on one set and exactly 1 on the other, mapping to the discrete space $\{0, 1\}$. Such a function carves the space into two disjoint open pieces. If a space is $T_1$ and has this powerful property for any two disjoint closed sets, it must not only be normal but also ​​totally disconnected​​. It is shattered into a "dust" of points, with no connected pieces larger than a single point. The kind of separation a space allows tells us about its very texture.

Finally, why stop at separating just two sets? What if we have an entire discrete collection of points, like stars scattered in the sky? Can we place a small, open "bubble" around each star such that no two bubbles overlap? A space where this is always possible is called ​​collectionwise Hausdorff​​. This is a powerful property for constructing functions and understanding complex spaces. And beautifully, it turns out that this ability to separate an entire collection of points is a direct consequence of the seemingly simpler property of being able to separate any two separated sets—that is, any completely normal space is also collectionwise Hausdorff.

Even in spaces that aren't normal, other properties can come to the rescue. In a Hausdorff space (where any two distinct points can be separated), if we take two compact sets that are disjoint, they can always be separated by disjoint open sets. Compactness acts as a kind of "topological finiteness," taming the sets and making them well-behaved enough to be separated, even when arbitrary closed sets in that same space might be inseparable.

From building functions to classifying geometric shapes and understanding the fundamental structure of space itself, the concept of separated sets is a simple, intuitive idea that blossoms into one of the most fruitful and unifying principles in mathematics. It shows us, once again, that asking the simplest questions—like "Can we put a fence here?"—can often lead us to the most profound and unexpected answers.