Topological Structure of Metric Spaces

Basic Notions in Metric Spaces Metric spaces form the foundation for much of analysis. They give us a way to measure distances between objects and study how limits, continuity, and compactness work in abstract settings. This unit introduces the key concepts you'll need to understand how metric spaces behave. Open Balls and Neighborhoods The most fundamental concept in metric spaces is the open ball. Given a point $x$ in a metric space $M$ with distance function $d$, an open ball of radius $r > 0$ centered at $x$ is the set of all points whose distance from $x$ is strictly less than $r$: $$B(x, r) = \{y \in M \mid d(x, y) < r\}$$ Think of this intuitively: in the real numbers with the usual distance, an open ball around 5 with radius 2 is the interval $(3, 7)$—everything within distance 2 of 5, but not including the endpoints. A neighborhood of a point $x$ is any set that contains some open ball around $x$. This is a broader concept: a neighborhood doesn't have to be an open ball itself, it just needs to contain one. For example, the interval $[3, 7]$ is a neighborhood of 5 (even though it's closed) because it contains the open ball $(3, 7)$ around 5. Why care about neighborhoods? Because they let us describe "points being close to $x$" without specifying exactly which open ball we're talking about. Open Sets and the Induced Topology Now we move from local properties around individual points to global properties of sets. A set $U$ is open if it is a neighborhood of every one of its points. In other words, $U$ is open when every point in $U$ has an open ball around it that stays entirely within $U$. Here's the key insight: every open ball is itself an open set. If you take any open ball $B(x, r)$, then for any point $y$ inside it, you can draw a smaller ball around $y$ that still fits inside the original ball. This might seem obvious, but it's important: open balls form a basis for the topology induced by the metric. This means that the metric gives us a natural notion of which sets are "open," and this structure—the collection of all open sets—is called the topology induced by the metric. Let me give a concrete example. In $\mathbb{R}$ with the usual distance, the open interval $(0, 1)$ is open because for every point $x$ in it, you can find an open ball around $x$ that stays in $(0, 1)$. But the closed interval $[0, 1]$ is not open because the point 0 has no ball around it that stays entirely in $[0, 1]$ (any ball around 0 goes slightly negative). Convergence of Sequences In metric spaces, we can talk about whether sequences converge. A sequence $(xn)$ converges to a point $x$ if the terms eventually get arbitrarily close to $x$. More precisely: $$xn \to x \text{ if for every } \varepsilon > 0 \text{, there exists } N \text{ such that } d(xn, x) < \varepsilon \text{ for all } n \geq N$$ This is the $\varepsilon$-$N$ definition you may have seen in calculus. The idea is simple: no matter how small a tolerance $\varepsilon$ you set, eventually all terms of the sequence are within that tolerance of $x$. For example, the sequence $1/n$ converges to 0 in $\mathbb{R}$ because for any $\varepsilon > 0$, we can choose $N$ large enough that $1/N < \varepsilon$, and then all terms $1/n$ for $n \geq N$ are within $\varepsilon$ of 0. The beauty of this definition is that it works the same way in any metric space, not just the real numbers. Cauchy Sequences and Completeness There's another important type of sequence: a Cauchy sequence. A sequence $(xn)$ is Cauchy if its terms eventually get arbitrarily close to each other (rather than to a fixed limit): $$\text{Cauchy if for every } \varepsilon > 0 \text{, there exists } N \text{ such that } d(xm, xn) < \varepsilon \text{ for all } m, n \geq N$$ Here's a crucial observation: in $\mathbb{R}$, every Cauchy sequence converges to some real number. This is actually a deep fact—it's equivalent to the completeness of the real numbers. But not all metric spaces have this property! A metric space is complete if every Cauchy sequence converges to a point in that space. Completeness is a powerful property. It means that if you have a sequence whose terms are getting closer and closer together, you're guaranteed that it actually converges to something in your space. Why is this important? Many important theorems in analysis (like the Banach fixed-point theorem) require completeness to work. A space without completeness is "missing" some limit points, and this can cause theorems to fail. Examples of Incomplete Spaces To understand completeness better, let's look at spaces that aren't complete. The open interval $(0, 1)$ with the usual metric: Consider the sequence $xn = 1/n$. This is a Cauchy sequence: the terms $1/n$ get arbitrarily close together as $n$ grows. But what do they converge to? In $\mathbb{R}$, they converge to 0. However, 0 is not in the interval $(0, 1)$. So this Cauchy sequence has no limit in the space—the space is incomplete because it's "missing" the endpoint 0. The rational numbers $\mathbb{Q}$ with the usual metric: This is a more subtle example. Consider a sequence of rationals converging to $\sqrt{2}$, like the decimal approximations $1.4, 1.41, 1.414, 1.4142, \ldots$. This is a Cauchy sequence of rational numbers (the terms get arbitrarily close), but it converges to $\sqrt{2}$, which is irrational. Since $\sqrt{2} \notin \mathbb{Q}$, the space $\mathbb{Q}$ is incomplete—it's missing irrational limits. These examples show that completeness is a genuine restriction. Many naturally occurring spaces are incomplete, and we often need to "complete" them. Completion of a Metric Space Fortunately, there's a theorem that saves us: every metric space has a unique completion. The completion is a complete metric space that contains your original space as a dense subset (meaning the original space's closure is the whole completion). For instance, the completion of $\mathbb{Q}$ is $\mathbb{R}$. The completion of $(0, 1)$ is $[0, 1]$. In each case, we've added in exactly the limit points that were missing. While we won't need to construct completions in detail, understanding that completions exist helps us know that incompleteness isn't permanent—we can always "fix" it by adding missing limits. Boundedness and Total Boundedness Two related concepts help us understand the "size" of a metric space. A metric space $M$ is bounded if there's a single distance that separates any two points. More precisely, if there exists $R$ such that: $$d(x, y) \leq R \text{ for all } x, y \in M$$ For example, any open ball is bounded, but the entire real line is not. Total boundedness is subtly different and more interesting. A metric space is totally bounded if it can be covered by finitely many open balls of any given radius. That is, for every $r > 0$, we can find finitely many balls $B(x1, r), B(x2, r), \ldots, B(xk, r)$ that together cover the entire space. Here's the key difference: a totally bounded space must be bounded, but the converse isn't true. The interval $(0, 1)$ is bounded (all distances are less than 1), but consider what happens with total boundedness: if we want to cover it with balls of very small radius $r$, we need at least $\lceil 1/r \rceil$ balls—a finite number. So $(0, 1)$ is totally bounded. But an unbounded space like $\mathbb{R}$ is not totally bounded: no finite collection of balls of any fixed radius can cover the entire real line. The distinction matters because total boundedness is more restrictive, and spaces that are both complete and totally bounded have special properties (compactness, as we'll see next). Compactness Compactness is one of the most important concepts in metric spaces. A metric space is compact if every open cover has a finite subcover. Let me unpack this: An open cover is a collection of open sets whose union is the entire space A finite subcover is a finite subcollection of those sets that still cover the entire space The definition says: no matter how you try to cover the space with open sets, you can always find a finite subset of those sets that still covers everything. For metric spaces, we have several equivalent ways to characterize compactness, and these are extremely useful: Characterization 1: Sequential Compactness — A metric space is compact if and only if every sequence has a convergent subsequence. This is often called sequential compactness and is usually the easiest condition to verify in practice. Characterization 2: Complete and Totally Bounded — A metric space is compact if and only if it is both complete and totally bounded. Why are these equivalent? The intuition is compelling: if a space is totally bounded, we can cover it with finitely many small balls, and if it's complete, any sequence's "clustering behavior" (forced by total boundedness) leads to an actual limit. Conversely, any compact space must be complete (Cauchy sequences have convergent subsequences, hence converge) and totally bounded (covering with balls is the definition). Example: The closed interval $[0, 1]$ is compact. Every sequence in $[0, 1]$ has a convergent subsequence (Bolzano-Weierstrass theorem), and the space is both complete and totally bounded. The open interval $(0, 1)$ is not compact because it's not complete (as we saw, the sequence $1/n$ is Cauchy but doesn't converge in the space). Compactness matters because many key theorems require it: continuous functions on compact spaces are uniformly continuous, the extreme value theorem requires compactness, and many optimization problems require compactness to guarantee solutions.