The Ordered Square

The unit square is one of the most familiar objects in mathematics, a simple shape we learn to draw as children. But what happens when we impose a completely different structure on it, one that follows the logic of a dictionary rather than distance? This article introduces the ordered square, the unit square endowed with the lexicographical order topology. This seemingly simple change creates a topological space with bizarre and counterintuitive properties, challenging our fundamental understanding of concepts like continuity, connectedness, and distance. We will bridge this gap between geometric intuition and topological reality. The first chapter, "Principles and Mechanisms," will deconstruct the ordered square, explaining how its dictionary-like order gives rise to its strange open sets and core properties. Following this, "Applications and Interdisciplinary Connections" will demonstrate its power as a topologist's laboratory, serving as a critical counterexample that refines our understanding of major theorems and reveals the subtle assumptions that underpin our perception of space.

Now that we have been introduced to the curious entity known as the ordered square, let's take a journey into its inner workings. How is it built, and why does it behave so differently from the familiar, friendly square we draw on paper? Like a physicist disassembling a watch to understand time, we will take apart the ordered square, piece by piece, to reveal the beautiful and strange logic that governs it.

Imagine you have a square piece of paper, the unit square $S = [0,1] \times [0,1]$. Normally, we think about points on this paper in terms of their Euclidean distance. But let's try a completely different approach. Let's imagine the points are words in a dictionary.

The ​​lexicographical order​​, or dictionary order, does just that. To compare two points, $P_1 = (x_1, y_1)$ and $P_2 = (x_2, y_2)$, we first look at their "first letters"—the $x$-coordinates. If $x_1$ comes before $x_2$ (i.e., $x_1 x_2$), then we declare that $P_1$ comes before $P_2$. The $y$-coordinates don't even matter, just as you don't look at the second letter of "apple" to know it comes before "banana".

Only if the first letters are the same ($x_1 = x_2$) do we proceed to look at the "second letters"—the $y$-coordinates. In that case, $P_1$ comes before $P_2$ if $y_1 y_2$.

This simple rule imposes a complete and total order on every single point in the square. We can march from the very first "word," $(0,0)$, all the way to the very last, $(1,1)$, without any ambiguity. The point $(0.2, 0.9)$ comes before $(0.3, 0.1)$. The point $(0.5, 0.4)$ comes before $(0.5, 0.6)$. It's as if we are scanning a book page: we read from left to right, and for a given column of text, we read from top to bottom (if we imagine the y-axis increasing upwards).

A new order demands a new sense of "nearness," a new topology. In an ordered space, the most basic kind of ​​open set​​ is an "open interval," the set of all points that lie strictly between two other points, say $P_A$ and $P_B$.

What do these open intervals look like? The answer holds the key to the entire structure.

Suppose we take two points on the same vertical line, like $P_A = (0.5, 0.2)$ and $P_B = (0.5, 0.8)$. The open interval $(P_A, P_B)$ consists of all points $(x,y)$ such that $(0.5, 0.2) (x,y) (0.5, 0.8)$. By the rules of our dictionary order, this can only be true if $x=0.5$ and $0.2 y 0.8$. The result is simply a vertical line segment: $\{0.5\} \times (0.2, 0.8)$. So far, so good.

But what if the points have different $x$-coordinates? Let's take $P_A = (0.3, 0.9)$ and $P_B = (0.6, 0.1)$. The open interval $(P_A, P_B)$ contains all points "after" $(0.3, 0.9)$ and "before" $(0.6, 0.1)$. This includes:

• The rest of the vertical line at $x=0.3$: $\{0.3\} \times (0.9, 1]$.
• All the points on the vertical lines for $x$ between $0.3$ and $0.6$: the entire rectangular strip $(0.3, 0.6) \times [0,1]$.
• The beginning of the vertical line at $x=0.6$: $\{0.6\} \times [0, 0.1)$.

This is a bizarre shape—not at all like a simple open disk in the usual plane. But it leads to a startling and crucial observation. Consider the set of points on a single vertical line, but excluding the endpoints, for example, the set $V_{0.5} = \{0.5\} \times (0,1)$. In our ordered square, this is precisely the open interval between the points $(0.5, 0)$ and $(0.5, 1)$. Therefore, $V_{0.5}$ is an open set!

Think about that for a moment. An entire line segment (without its ends) is considered an "open blob." This is profoundly different from our Euclidean intuition, where a line has no interior and cannot be open. In the ordered square, these vertical fibers are the fundamental open units.

Let's test this new structure by seeing how it relates to the coordinates themselves. We can define two natural "projection" maps: one that takes a point $(x,y)$ and tells you its $x$-coordinate, $\pi_1(x,y) = x$, and another that tells you its $y$-coordinate, $\pi_2(x,y) = y$. In the standard Euclidean topology, both of these maps are perfectly continuous—small changes in the input point $(x,y)$ lead to small changes in the output $x$ and $y$. What happens here?

The map $\pi_1(x,y)=x$ turns out to be ​​continuous​​. If you take a small open interval $(a,b)$ on the $x$-axis, its preimage—the set of all points in the square that get mapped into it—is the vertical strip $(a,b) \times [0,1]$. As we saw, this is a nice open set in the ordered square, a union of those open vertical slices. So far, so reasonable.

Now for the shocker: the map $\pi_2(x,y)=y$ is ​​not continuous​​. To see why, let's take an open set in the codomain $[0,1]$, say the interval $V = [0, 0.5)$. Its preimage is the set $U = [0,1] \times [0, 0.5)$. This set $U$ is not open in the ordered square. For example, the point $P=(1, 0.25)$ is in $U$, but any open neighborhood of $P$ must contain points with $x$-coordinates less than 1. Such a neighborhood, for instance the open interval from $(0.9,1)$ to $(1,0.3)$, will contain points like $(0.95, 0.8)$ which are not in $U$ because their $y$-coordinate is greater than $0.5$. The open sets of the ordered square refuse to be confined to horizontal strips. They always want to "bleed" vertically. The topology tears apart horizontal connections, making the projection onto the y-axis discontinuous.

As a beautiful counterpoint, if we look at the subspace topology on a single vertical line, say $\{0.5\} \times [0,1]$, it behaves exactly as we'd expect. The open sets it inherits from the big square are just the usual open intervals on that line segment. The strangeness of the ordered square isn't within the vertical lines, but in how they are stitched together—or rather, not stitched together horizontally.

This brings us to one of the most consequential properties of the ordered square. We established that each vertical slice $V_x = \{x\} \times (0,1)$ is an open set. Now, consider the whole collection of them: $\{\,V_x \mid x \in (0,1)\,\}$.

This is a family of non-empty open sets. They are all disjoint from one another—a point in $V_{0.3}$ has an $x$-coordinate of $0.3$, so it can't be in $V_{0.4}$. And crucially, the family is ​​uncountable​​, because there is one such set for every real number $x$ between 0 and 1.

This single fact acts like a wrecking ball for several "nice" properties we often take for granted.

• ​​Not Second-Countable:​​ A space is second-countable if its entire topology can be generated from a countable list of basic open sets. Our familiar Euclidean plane is second-countable (think of all open disks with rational centers and radii). But a second-countable space cannot possibly contain an uncountable swarm of disjoint open sets. You'd run out of basic open sets to put inside them! Therefore, the ordered square is ​​not second-countable​​.
• ​​Not Separable:​​ A space is separable if it contains a countable set of points that is "dense," meaning it gets arbitrarily close to every other point (like the rational numbers on the real line). If our space were separable, each of our uncountable disjoint open sets would have to contain at least one point from this countable dense set, which is impossible. Thus, the ordered square is ​​not separable​​.

With its "shredded" structure of disjoint open slices and torn horizontal connections, one might guess the ordered square is a disconnected mess. But here, nature throws us a curveball.

The space possesses a deep, underlying coherence due to a property it shares with the real number line: the ​​least upper bound property​​. This axiom states that any non-empty collection of points that has an upper bound must have a smallest possible upper bound (a supremum) that is also in the space. There are no "holes" or "gaps" in the order. An ordered set with this property is called a ​​linear continuum​​.

This property is the secret glue holding the square together. A fundamental theorem of topology states that any linear continuum is ​​connected​​. You cannot partition it into two disjoint non-empty open sets. Despite its appearance, the ordered square is a single, unbroken whole.

Furthermore, this same property is the key ingredient in proving that the ordered square is ​​compact​​. Just like a closed interval on the real line, it has a "boundedness" and "closedness" that forces any infinite collection of open sets covering it to have a finite sub-collection that still does the job.

Let us now assemble the pieces of our puzzle. We have constructed a space that is:

• ​​Normal, Regular, and Hausdorff:​​ It satisfies a whole hierarchy of separation axioms. Any two points can be separated by open sets, and any point can be separated from a closed set. In this regard, it is quite well-behaved.
• ​​Connected and Compact:​​ It is a single, unbroken, bounded entity.

But, at the same time, it is:

• ​​Not Path-Connected:​​ Although the space is connected, you cannot "walk" from one point to another. A continuous path, say from $(0,0)$ to $(1,1)$, would be a continuous image of the interval $[0,1]$. Such an image cannot possibly contain an uncountable number of disjoint open sets. Yet any path from a point with $x=0$ to a point with $x=1$ would have to cross every single one of our vertical open slices $V_x$. The two facts are irreconcilable. There is no such path.
• ​​Not Second-Countable:​​ As we saw, it's filled with an uncountable crowd of disjoint open sets.

This last point delivers the final, beautiful conclusion. The great Urysohn's Metrization Theorem states that a topological space has a distance function (is metrizable) if and only if it is Regular, Hausdorff, and Second-Countable. The ordered square clears the first two hurdles with flying colors but stumbles decisively on the third.

Therefore, the ordered square is famously ​​non-metrizable​​. Its intricate, dictionary-like topology of vertical slices is so alien to our everyday geometric intuition that no simple notion of "distance" can ever hope to describe it. It is a testament to the power of abstraction, a perfectly logical construction that defies simple visualization, serving as a vital landmark in vanishes vast and wondrous landscape of topology.

After our journey through the fundamental principles of the ordered square, you might be left with a sense of both familiarity and unease. It is, after all, just the unit square. We can draw it, point to its corners, and trace its edges. Yet, cloaked in the order topology, it becomes a world unto itself, a kind of "topologist's laboratory" where our intuitions about space, continuity, and connection are put to the ultimate test. It is in these applications—or, more accurately, in the surprising failures of our usual tools—that the true character and pedagogical power of the ordered square are revealed. It teaches us not by conforming to our expectations, but by brilliantly subverting them.

Let's start with a seemingly simple task. In any well-behaved space, we should be able to separate two distinct, closed objects. Consider the left edge of our square, $A = \{0\} \times [0,1]$, and the right edge, $B = \{1\} \times [0,1]$. How can we build a fence between them? In the familiar Euclidean plane, we'd just take a strip down the middle. In the ordered square, the same idea works, but for a much deeper reason. Because the lexicographical order prioritizes the $x$-coordinate, a set like $U = [0, c) \times [0,1]$ for some $c \in (0,1)$ is a genuine open set—it's just the interval of all points from the beginning, $(0,0)$, up to the point $(c,0)$. Similarly, $V = (c, 1] \times [0,1]$ is an open set. These two sets perfectly separate the left and right edges, demonstrating that the space is, in this regard, quite normal.

This "primacy of the x-coordinate" allows for the construction of beautifully simple continuous functions. If we want to build a continuous function that is $0$ on the left half of the square and $1$ on the right half, we can simply make it a function of $x$ alone, smoothly transitioning from $0$ to $1$ as $x$ increases. The powerful Tietze Extension Theorem, which guarantees we can extend a continuous function from a closed subset to the whole space, often yields this wonderfully intuitive result. If you define a function to be $0$ on the left edge and $1$ on the right, the most natural continuous extension to the entire square is simply the projection function, $F(x,y) = x$. The same principle holds even for more complex starting sets, like the anti-diagonal line, or for constructing the theoretical building blocks of the Tietze extension proof itself. It seems, for a moment, that the $y$-coordinate is just along for the ride.

But this is a trap! The simplicity is an illusion. What happens if we try to separate two disjoint sets on the same vertical line? Suppose we have a segment $A = \{c\} \times [0, y_A]$ and another segment $B = \{c\} \times [y_B, 1]$ on the same line $x=c$. Now, a function that only depends on $x$ is useless. The topology suddenly reveals its hidden, intricate structure. To build a continuous function $f$ that is $0$ on $A$ and $1$ on $B$, we must respect the strange "jump" that occurs when we cross from one $x$-value to the next. A continuous function on the ordered square must satisfy a bizarre condition: the limit as you approach a line $x=c$ from the left, along the top edge ($y=1$), must equal the function's value at the bottom of the line $x=c$ (at $y=0$). A similar condition holds approaching from the right. The only way to satisfy these jump conditions and separate our two sets is for the function to be constant (and zero) for all $x c$, and constant (and one) for all $x > c$. On the line $x=c$ itself, the function must continuously interpolate from $0$ to $1$ as $y$ travels from $y_A$ to $y_B$. This is a world away from the simple continuity of the Euclidean plane. It is a profound lesson: the ordered square is not merely a product of two intervals; it is a unified, indivisible entity with its own peculiar laws of connection.

The nature of connection in the ordered square is perhaps its most famous and startling feature. Let's start with a simple observation. In a simple line of points, if you remove any point except an endpoint, you split the line in two. The same is true here. Because the space is linearly ordered, removing any single point $p$ (that isn't the absolute minimum or maximum) disconnects the space into two non-empty open sets: the set of all points less than $p$, and the set of all points greater than $p$. Thus, every point except for $(0,0)$ and $(1,1)$ is a cut point.

This suggests a space that is connected, but perhaps fragilely so. The truth is far stranger. The ordered square is indeed a connected space—it cannot be broken into two disjoint open pieces. Yet, it is profoundly, fundamentally not path-connected. Imagine you are an ant at the bottom-left corner, $(0,0)$, and you wish to crawl to your friend at the top-right, $(1,1)$. In our world, this is a simple matter. In the ordered square, it is an impossible journey.

Why? Suppose such a path—a continuous function from the time interval $[0,1]$ into the square—could exist. The image of this path would have to be a connected and compact subset of the square. But because it starts at the minimum point and ends at the maximum, the image would have to be the entire square. Here lies the contradiction. For any $x$ between $0$ and $1$, the vertical line segment $\{x\} \times (0,1)$ is an open set in the ordered square. These segments are all disjoint from one another. This gives us an uncountable collection of disjoint, non-empty open sets. However, a fundamental result in topology states that the continuous image of a simple interval like $[0,1]$ (which is metrizable and "small" in a topological sense) cannot contain such an uncountable family of disjoint open sets. The very fabric of the ordered square is too "roomy" and "un-navigable" to be traced out by a simple path. You can move up and down a vertical line, but you can never make a continuous leap from one $x$-coordinate to another. The path components, the regions of mutual reachability, are just the individual vertical lines. It is a world of infinite, parallel, isolated universes.

We arrive now at the most profound lesson the ordered square has to teach us. There is a classic, beautiful theorem stating that any continuous function from a compact, connected, linearly ordered space to itself must have a fixed point (a point $p$ such that $f(p)=p$). The proof is as simple as it is elegant: if a function $f$ had no fixed points, then for every point $p$, either $f(p) > p$ or $f(p) p$. One can then show that the set of points where $f(p) > p$ and the set where $f(p) p$ are both non-empty and open. But this would mean the space is disconnected, a contradiction.

The ordered square is compact. It is connected. It is linearly ordered. Therefore, it must have the fixed-point property.

Except it doesn't. There are continuous functions on the ordered square with no fixed points.

Here we face a paradox. We have a seemingly flawless proof leading to a conclusion that is demonstrably false. This is the best kind of puzzle in science and mathematics, because the flaw is not a simple calculation error, but a deep misunderstanding of our own hidden assumptions. The "standard argument" for the fixed-point property relies on an unstated lemma, a consequence of the Intermediate Value Theorem: that in such a space, the image of any interval $[a,b]$ is also a connected interval. This is true for the real number line. It is not true for the ordered square. An "interval" in the ordered square, like the set of all points between $(0.2, 0.5)$ and $(0.8, 0.5)$, is not a connected set. It is a collection of uncountable, disconnected vertical slices. The continuity of a function is not strong enough to weld these pieces back together in its image. The proof fails because our intuition about what an "interval" is, an intuition forged on the real line, is wrong.

The ordered square, then, serves as a crucial counterexample. It is a sentinel that guards the frontiers of topology, warning us when our theorems rely on unspoken assumptions. It forces us to distinguish between connectivity and path-connectivity, to scrutinize the conditions for the Intermediate Value Theorem, and to appreciate that the properties of the familiar real line are not universal truths. By exploring its strange and beautiful landscape, we do not merely learn about one peculiar space; we sharpen our understanding of the entire mathematical universe.