Closed Map

In the study of topology, functions are our primary means of comparing and transforming spaces. We are often interested in what properties these functions preserve. While continuous functions are celebrated for preserving the integrity of a space by preventing "tearing," another fundamental property exists: the preservation of boundaries and limit points. This article addresses the concept of ​​closed maps​​—functions that respect the "closedness" of sets. We will explore what it means for a function to have this powerful property and why it is not just a technical curiosity but a cornerstone concept with far-reaching implications.

This article will guide you through a comprehensive understanding of closed maps. In the first chapter, ​​Principles and Mechanisms​​, we will establish the formal definition, contrast closed maps with their open map counterparts, and uncover the profound relationship between closed maps, continuity, and compactness. In the second chapter, ​​Applications and Interdisciplinary Connections​​, we will see this theory in action, revealing how closed maps serve as a secret weapon for identifying homeomorphisms, describing physical rotations, and navigating the subtle complexities of infinite-dimensional analysis.

In our journey through the world of topology, we often talk about properties that are preserved by certain transformations. We've seen that continuous functions are the superstars—they preserve the "connectedness" of a space, preventing it from being torn apart. But there's another, equally profound, kind of preservation: the preservation of ​​closedness​​. This is the domain of ​​closed maps​​.

A function $f: X \to Y$ is called a ​​closed map​​ if it takes every closed set in its domain $X$ and maps it to a set that is also closed in the codomain $Y$. This definition seems simple enough, but its consequences are far-reaching and, at times, quite surprising. It tells us about a function's ability to respect boundaries and limit points, a fundamentally different property from continuity.

What does it really mean for a set to be "closed"? In familiar spaces like the real line $\mathbb{R}$ or the Euclidean plane $\mathbb{R}^2$, a closed set is one that contains all of its ​​limit points​​. Think of the interval $[0, 1]$. The points $0$ and $1$ are limit points—you can get as close as you want to them using points from within the set. Since $[0, 1]$ includes them, it's closed. The interval $(0, 1)$, on the other hand, is not closed because it doesn't contain its limit points $0$ and $1$.

A closed map, then, is a function that doesn't carelessly discard these limit points. If you give it a closed set, the resulting image will also contain all of its limit points.

Let's start with the simplest possible function: the ​​inclusion map​​. If you have a subset $A$ inside a larger space $X$, the inclusion $i: A \to X$ just says, "every point in $A$ is... well, that same point in $X$." It's almost a "do-nothing" function. When is this simple map a closed map? A moment's thought reveals a beautiful and direct connection: the inclusion map $i: A \to X$ is a closed map if and only if the subset $A$ is itself a closed subset of $X$. Why? A set is always closed relative to itself. The only closed set in the domain $A$ we need to worry about is $A$ itself. The map is closed if the image of $A$, which is just $A$ again, is closed in the larger space $X$. This provides our first, most basic litmus test for closedness.

You might be wondering if "closed map" is just another name for "open map" (a map that sends open sets to open sets), or perhaps they are two sides of the same coin. Let's investigate with a classic example.

Consider the projection map $p_1: \mathbb{R}^2 \to \mathbb{R}$ that takes a point $(x, y)$ and gives you its "shadow" on the x-axis, $p_1(x, y) = x$. This map is wonderfully continuous and, as it turns out, it's an ​​open map​​. If you take any open disk in the plane, its shadow on the x-axis is an open interval.

But is it a closed map? Let's test it. Consider the set $C = \{(x, y) \in \mathbb{R}^2 \mid xy = 1\}$. This is the graph of a hyperbola, a perfectly good closed set in the plane. What is its shadow on the x-axis? The projection $p_1(C)$ is the set of all $x$-coordinates that appear in the set. Since we can find a $y$ for any $x$ as long as $x \neq 0$ (by setting $y = 1/x$), the image is the entire real line except for zero: $\mathbb{R} \setminus \{0\}$. Is this set closed? No! The point $0$ is a limit point of this set (consider the sequence $1/n \to 0$), but $0$ is not in the set itself. The projection map failed to preserve the closed nature of the hyperbola; it "lost" a limit point in the process.

So, we have our answer: open and closed maps are fundamentally different beasts. A map can be open but not closed. In fact, we can also find maps that are closed but not open, and maps that are neither! This distinction is not just a curious edge case; it's central to understanding the geometric behavior of functions.

Checking every single closed set to see if its image is closed seems like an impossible task. Is there a more powerful principle at play? A condition that, if met, automatically guarantees a map is closed? Fortunately, yes. And it comes from one of the most powerful ideas in all of topology: ​​compactness​​.

Intuitively, a space is ​​compact​​ if it's "self-contained" and "bounded." The interval $[0, 1]$ is compact; the entire real line $\mathbb{R}$ is not. You can't "run off to infinity" within a compact space. The other ingredient we need is for the codomain to be a ​​Hausdorff​​ space—one where any two distinct points can be put into their own separate, non-overlapping open "bubbles." Nearly all the spaces you're familiar with, like $\mathbb{R}$, $\mathbb{R}^2$, and spheres, are Hausdorff.

Here is the theorem, and it's a thing of beauty:

​​Any continuous function from a compact space to a Hausdorff space is a closed map.​​

Why is this true? Think of it this way. A closed subset of a compact space is itself compact (it's a self-contained piece of a self-contained whole). Continuity guarantees that the image of this compact piece is also compact. And finally, in a well-behaved Hausdorff space, every compact set is automatically closed. The chain of logic is flawless:

Closed subset of compact domain $\implies$ Compact subset of domain $\xrightarrow{\text{continuous map}}$ Compact subset of codomain $\implies$ Closed subset of Hausdorff codomain.

Let's see this superpower in action. Consider the function $f: [0, 1] \to S^1$ (the unit circle in $\mathbb{R}^2$) given by $f(t) = (\cos(2\pi t), \sin(2\pi t))$. This function continuously wraps the unit interval around the circle. Is it a closed map? The domain $[0, 1]$ is compact. The codomain $S^1$ is Hausdorff. The function is continuous. Therefore, by our theorem, it must be a closed map. We don't have to lift a finger to check any specific sets.

This theorem also explains why some maps fail. The projection $p_1: \mathbb{R}^2 \to \mathbb{R}$ wasn't a closed map. Could we have predicted this? Its domain, $\mathbb{R}^2$, is not compact. So the theorem doesn't apply, and failure is a possibility. Or consider wrapping the half-open interval $[0, 2\pi)$ around the circle. This domain isn't compact, and indeed, this map is not closed. The superpower of compactness is not to be taken lightly!

Just as we can combine numbers with addition and multiplication, we can combine functions through composition, restriction, and pasting. Do these operations preserve the property of being a closed map?

• ​​Restriction:​​ If you have a closed map $f: X \to Y$ and you restrict your attention to a closed subset $A \subseteq X$, is the new map $f|_A: A \to Y$ also a closed map? The answer is a resounding yes. If a function behaves well on the whole space, it will continue to behave well on a well-behaved (closed) portion of it.

• ​​Composition:​​ If $f: X \to Y$ and $g: Y \to Z$ are both closed maps, is their composition $g \circ f: X \to Z$ also a closed map? Yes. Let $C$ be a closed set in $X$. Since $f$ is closed, its image $f(C)$ is a closed set in $Y$. Now, since $g$ is a closed map, it takes the closed set $f(C)$ and maps it to the set $g(f(C))$, which must be closed in $Z$. The property is passed along the chain like a baton in a relay race. However, if even one link in the chain is broken, the whole thing can fail.

• ​​Pasting:​​ Suppose you build a function $f$ by "pasting" together two other functions, $f|_A$ and $f|_B$, where $X = A \cup B$. If both $A$ and $B$ are closed sets, and both restrictions $f|_A$ and $f|_B$ are closed maps, then the entire function $f$ is a closed map. This "Pasting Lemma" for closed maps is a powerful tool for building complex closed maps from simpler pieces.

Throughout our discussion, the specific topologies on the domain and codomain have been lurking in the background, quietly dictating the outcome. Let's bring them to the forefront.

Consider the floor function $f: \mathbb{R} \to \mathbb{Z}$, which maps a real number to the greatest integer less than or equal to it. Let's give $\mathbb{R}$ its usual topology and $\mathbb{Z}$ the ​​discrete topology​​, where every subset is considered open (and therefore also closed). Is the floor function a closed map? Yes, trivially so! Take any closed set $C$ in $\mathbb{R}$. Its image $f(C)$ is some subset of the integers. But in the discrete topology on $\mathbb{Z}$, every subset is closed. So $f(C)$ is guaranteed to be closed, no matter what $C$ we started with. The nature of the codomain's topology preordained the result.

Finally, we must address a subtle but important distinction. Is being a "closed map" related to having a "closed graph"? The ​​graph​​ of a function $f: X \to Y$ is the set of points $\Gamma_f = \{(x, f(x))\}$ inside the product space $X \times Y$. It's a standard result that if $f$ is continuous and $Y$ is Hausdorff, then its graph $\Gamma_f$ is a closed set. This might lead us to believe the two concepts are intertwined.

They are not the same. It is possible to construct a function that is a closed map, but whose graph is not a closed set in the product space. These examples often live in the strange world of non-Hausdorff or non-standard topologies, but they serve as a crucial reminder of the precision of topological language. A "closed map" is a statement about the function's behavior on subsets of its domain. A "closed graph" is a statement about the shape of the function itself as an object living in a larger product space.

Understanding closed maps opens up a new dimension in our analysis of functions. It's not just about continuity's promise of "no tearing." It's about a function's respect for boundaries, its interaction with compactness, and the subtle, powerful ways it can transform the geometric landscape.

We have spent some time getting to know the formal definition of a closed map, a function that diligently carries closed sets from its domain to closed sets in its codomain. At first glance, this might seem like a rather niche and technical property. Why should we care if a map preserves this particular feature? Is it just a definition for definition's sake, a curiosity for the abstract-minded topologist? The wonderful answer is no. Like a master key that unlocks many different doors, the concept of a closed map reveals its power and beauty by providing elegant solutions and profound insights across a surprising variety of mathematical landscapes. It is a unifying thread that ties together abstract topology, the geometry of rotations, and the subtleties of infinite-dimensional analysis.

Perhaps the most immediate and celebrated application of closed maps is in answering a central question of topology: when are two spaces truly the "same"? The gold standard for topological equivalence is the homeomorphism—a continuous bijection whose inverse is also continuous. The first two conditions, continuity and bijectivity, are often straightforward to check. But proving that the inverse function is continuous can be a tedious and difficult task. It requires us to show that the preimage of every open set under the inverse is open, which is equivalent to showing that the original function is an open map.

This is where the closed map property provides a brilliant alternative. For a bijective function, showing it is a closed map is exactly equivalent to showing its inverse is continuous. So, the problem is transformed: how can we easily tell if a map is closed?

Here, nature gives us a remarkable gift, a theorem of stunning power and simplicity. If you have a continuous map $f$ from a ​​compact​​ space $X$ to a ​​Hausdorff​​ space $Y$, then $f$ is automatically a closed map!. Think about what this means. Compactness, the topological notion of being "solid" and having no "missing points" at infinity, and the Hausdorff property, the gentle condition that any two distinct points can be separated by their own open neighborhoods, are properties of the spaces, not the map itself. If your spaces have these common and desirable characteristics, any continuous function between them gets the 'closed map' property for free.

The practical upshot is enormous. If you have a continuous bijection between a compact space and a Hausdorff space, you know instantly it's a homeomorphism. You don't need to construct the inverse. You don't need to check if it's an open map. You just check the spaces themselves. It feels almost like cheating! This single theorem acts as a powerful detective's tool, cutting through computational fog to reveal the true topological nature of a function with astonishing efficiency.

To truly appreciate this, let's build some intuition. What does a closed map even look like? Consider a simple, geometric function $f: \mathbb{R}^2 \to \mathbb{R}^2$ defined by $f(x,y) = (|x|, y)$. This map "folds" the entire plane along the y-axis, mapping the left half-plane onto the right. Is this a closed map? A moment's thought shows that it is. If you take any closed set $C$ in the plane, its image $f(C)$ is just the union of its portion in the right half-plane and the reflection of its portion in the left half-plane. Since reflection is a homeomorphism (it preserves all topological properties) and the union of two closed sets is always closed, the image is guaranteed to be closed.

Now contrast this with a map that isn't closed, such as the simple projection $p(x,y) = x$ which squashes the plane onto the x-axis. Consider the set of points where $xy=1$, which forms a hyperbola—a perfectly good closed set. When we project this onto the x-axis, we get the entire x-axis except for the origin. This resulting set, $\mathbb{R} \setminus \{0\}$, is famously not closed. The projection map failed to preserve the 'closed' property. This sort of counterexample is invaluable; it shows that being a closed map is a non-trivial property, making the cases where it holds all the more special.

Is our "compact-to-Hausdorff" superpower just a toy for playing with abstract planes? Far from it. Let's look at the real world of physics and engineering. Consider the set of all possible rotations in three-dimensional space. This set forms a beautiful mathematical object called the special orthogonal group, $SO(3)$. Now, pick a direction—say, the North Pole on a globe, represented by the vector $e_1 = (1, 0, 0)^T$. The set of all possible directions forms the 2-sphere, $S^2$.

There is a natural map $p: SO(3) \to S^2$ that takes a rotation and tells us where it sends the North Pole. This map is fundamental to describing orientations. Is it a closed map? Let's use our secret weapon. The space of rotations $SO(3)$ is compact—it is a closed and bounded subset of the 9-dimensional space of $3 \times 3$ matrices. The sphere $S^2$ is a subspace of our familiar Euclidean $\mathbb{R}^3$ and is therefore perfectly Hausdorff. The conditions of our theorem are met! We can declare, with no further calculation, that this physically meaningful map is a closed map. A deep result about the geometry of rotations is proven in two lines using a purely topological argument. This is a sterling example of the unity of mathematics.

The story gets even more interesting when we venture into the wild territory of infinite-dimensional spaces, the home of functional analysis. Here, words can have subtle and different meanings, and our intuition must be retrained.

Consider two spaces of infinite sequences: $\ell^1$, the space of sequences whose absolute values sum to a finite number, and $c_0$, the space of sequences that converge to zero. Every sequence in $\ell^1$ must converge to zero, so $\ell^1$ is a subspace of $c_0$. Let's examine the simple inclusion map $T: \ell^1 \to c_0$ that just takes a sequence in $\ell^1$ and views it as a sequence in $c_0$.

In functional analysis, there is a concept called a "closed graph operator." An operator has a closed graph if the set of all pairs $(x, T(x))$ forms a closed set in the product space. For linear operators between the nice spaces of functional analysis (Banach spaces), the celebrated Closed Graph Theorem tells us that an operator has a closed graph if and only if it is continuous. Our inclusion map $T$ is easily shown to be continuous, so it is a closed graph operator.

But is it a ​​closed map​​ in the sense we've been discussing? Does it map closed sets in $\ell^1$ to closed sets in $c_0$? The answer is a resounding no. To see this, consider the entire space $\ell^1$. It is certainly a closed set within itself. Its image under $T$ is just $\ell^1$ again, but now sitting inside the larger space $c_0$. And this image is not closed in $c_0$. We can construct a sequence of points in $\ell^1$ that converge to a point that is in $c_0$ but not in $\ell^1$. The classic example is the harmonic sequence $y_n = 1/n$. This sequence converges to zero, so it's in $c_0$. But it is not in $\ell^1$ because the harmonic series $\sum 1/n$ diverges. However, one can easily find a sequence of finitely-supported (and thus $\ell^1$) sequences that converge to $(1/n)$ in the topology of $c_0$. This means $(1/n)$ is a limit point of the image of $T$, but it is not in the image. The image is not closed.

This example is a crucial lesson. It demonstrates that in the world of infinite dimensions, "closed graph" and "closed map" are fundamentally different concepts. It shows that even the most natural inclusion maps can fail to be closed, highlighting the subtlety and richness of these infinite-dimensional structures.

Finally, let's return to the topologist's workshop, where new spaces are constantly being built by cutting and gluing. One of the most useful constructions is the ​​mapping cylinder​​. Given any continuous map $f: X \to Y$, the mapping cylinder $M_f$ is formed by taking the "prism" $X \times [0, 1]$ and gluing the end $X \times \{1\}$ to the space $Y$ according to the map $f$. This procedure effectively transforms the map $f$ into an inclusion, as $Y$ now sits inside the newly created space $M_f$.

A natural question arises: how does $Y$ sit inside this larger construction? Is its embedding "nice"? One way to measure "niceness" is to ask if the inclusion map $j: Y \to M_f$ is a closed map. The answer is remarkable: it always is, regardless of the spaces $X$ and $Y$ or the continuous map $f$ you started with.

The reason is a small piece of logical elegance. For a set $C$ in $Y$ to be considered closed inside the mapping cylinder $M_f$, its preimage before the gluing must be a closed set in the disjoint union of $(X \times [0,1])$ and $Y$. This preimage consists of two parts: the set $C$ itself inside $Y$, and the set of points in the prism that get glued to $C$, which is precisely $f^{-1}(C) \times \{1\}$. Since $f$ is continuous and $C$ is closed, $f^{-1}(C)$ is closed in $X$. Thus, both pieces of the preimage are closed, their union is closed, and so the image $j(C)$ is closed in $M_f$. This property, which comes "built-in" to the construction, is a vital technical lemma that makes the mapping cylinder an indispensable and well-behaved tool in algebraic topology, particularly in the study of homotopy. The fact that the concept of a closed map appears so naturally and powerfully in such a fundamental construction underscores its deep importance to the field.

From a simple definition, we have taken a journey across mathematics. The concept of a closed map is no mere technicality. It is a unifying principle, a labor-saving device, a source of subtle distinctions, and a guarantor of structural integrity. It reminds us that sometimes the most powerful ideas are those that tell us what is preserved on a journey from one space to another.