Metric Spaces

Definition. A metric space is a pair $(M, d)$ where $M$ is a set and $d : M \times M \to [0, \infty)$ satisfies, for all $x, y, z \in M$:

1. Non-negativity: $d(x, y) \geq 0$, with equality iff $x = y$.
2. Symmetry: $d(x, y) = d(y, x)$.
3. Triangle inequality: $d(x, z) \leq d(x, y) + d(y, z)$.

Every metric induces a topology: the open sets are unions of open balls $B_r(p) = \{x : d(x, p) < r\}$.

Example 7.1 (Euclidean metric). On $\mathbb{R}^n$:

$d_2(\mathbf{x}, \mathbf{y}) = \sqrt{\sum_{i=1}^{n} (x_i - y_i)^2}.$

Example 7.2 ($p$-norm metrics). For $1 \leq p \leq \infty$:

$d_p(\mathbf{x}, \mathbf{y}) = \left(\sum_{i=1}^{n} |x_i - y_i|^p\right)^{1/p}, \qquad d_\infty(\mathbf{x}, \mathbf{y}) = \max_{1 \leq i \leq n} |x_i - y_i|.$

All of these induce the standard topology on $\mathbb{R}^n$.

Example 7.3 (Discrete metric). For any set $X$:

$d(x, y) = \begin{cases} 0 & \text{if } x = y, \\ 1 & \text{if } x \neq y. \end{cases}$

The discrete metric induces the discrete topology.

Definition. A sequence $(x_n)$ in $(M, d)$ converges to $x \in M$ (written $x_n \to x$) if for every $\varepsilon > 0$ there exists $N$ such that $d(x_n, x) < \varepsilon$ for all $n \geq N$.

Proposition 7.1. In a metric space, $x_n \to x$ if and only if for every open neighbourhood $U$ of $x$, there exists $N$ with $x_n \in U$ for all $n \geq N$.

Limits in metric spaces are unique (this follows from the Hausdorff property).

Definition. A sequence $(x_n)$ is Cauchy if for every $\varepsilon > 0$ there exists $N$ such that $d(x_m, x_n) < \varepsilon$ for all $m, n \geq N$.

Every convergent sequence is Cauchy. The converse need not hold.

Definition. A metric space $(M, d)$ is complete if every Cauchy sequence converges.

Example 7.4. $\mathbb{R}^n$ with the Euclidean metric is complete.

Example 7.5. $\mathbb{Q}$ with $d(x, y) = |x - y|$ is not complete: the Cauchy sequence $3, 3.1, 3.14, 3.141, 3.1415, \ldots$ does not converge in $\mathbb{Q}$.

Proposition 7.2. A closed subset of a complete metric space is complete.

Definition. A Banach space is a complete normed vector space.

Example 7.6. $(C([a, b]), \|\cdot\|_\infty)$ — the space of continuous functions on $[a, b]$ with the sup norm — is a Banach space.

Definition. A map $f : M \to M$ is a contraction if there exists $0 \leq c < 1$ such that $d(f(x), f(y)) \leq c \cdot d(x, y)$ for all $x, y \in M$.

Theorem 5.6 (Banach Fixed Point Theorem). If $(M, d)$ is a complete metric space and $f : M \to M$ is a contraction, then $f$ has a unique fixed point $x^*$, and for any $x_0 \in M$, the iteration $x_{n+1} = f(x_n)$ converges to $x^*$.

Proof. For any $x_0$, the sequence $x_n = f^n(x_0)$ is Cauchy (by repeated application of the contraction condition), hence converges to some $x^*$. By continuity of $f$, $x^* = f(x^*)$. Uniqueness follows from the contraction condition: if $x^* = f(x^*)$ and $y^* = f(y^*)$, then $d(x^*, y^*) = d(f(x^*), f(y^*)) \leq c \cdot d(x^*, y^*)$, so $d(x^*, y^*) = 0$. $\square$

Compact $\Rightarrow$ complete and totally bounded. Complete + totally bounded $\Rightarrow$ compact.

• Assuming all metrics come from a norm. The discrete metric does not come from any norm (norms are homogeneous: $\|\alpha x\| = |\alpha|\|x\|$, but $d(\alpha x, 0) = 1$ for $\alpha \neq 0$).
• Confusing completeness with closedness. A subset of a metric space can be closed but not complete (if the ambient space is not complete), and complete but not closed (if complete in a subspace topology).
• Thinking contractions require $d(f(x), f(y)) < d(x, y)$ for all $x \neq y$. This is weaker and does not guarantee a fixed point. The strict inequality $d(f(x), f(y)) \leq c d(x, y)$ with $c < 1$ is essential.
• Forgetting that convergence of sequences is not enough in general topology. In metric spaces sequential convergence characterises the topology, but in general topological spaces nets or filters are needed.

• Numerical analysis: The Banach fixed-point theorem justifies iterative methods like Newton’s method and Picard iterations for ODEs.
• Computer science: Metric spaces model edit distances (Levenshtein, Hamming) for string matching and bioinformatics sequence alignment.
• Image processing: The Hausdorff distance measures similarity between shapes and is used in computer vision for template matching.
• Functional analysis: $C([a,b])$ with the sup norm is a complete metric space, providing the setting for the Stone-Weierstrass approximation theorem.

Problem 1. Show that $\mathbb{R}^n$ with the Euclidean metric is complete.

Solution. Let $\{x_k\}$ be a Cauchy sequence in $\mathbb{R}^n$. Then each coordinate sequence $\{x_k^{(i)}\}$ is Cauchy in $\mathbb{R}$ (since $|x_k^{(i)} - x_\ell^{(i)}| \leq d_2(x_k, x_\ell)$). $\mathbb{R}$ is complete, so each coordinate converges to $x^{(i)}$. Then $x_k \to x$ componentwise, and by the triangle inequality $d_2(x_k, x) \to 0$. $\blacksquare$

Problem 2. Let $X = (0,1)$ with $d(x,y) = |x-y|$. Show $X$ is not complete.

Solution. The sequence $x_n = 1/n$ is Cauchy: for any $\varepsilon > 0$, choose $N > 2/\varepsilon$, then for $m,n \geq N$, $|1/m - 1/n| \leq 1/m + 1/n \leq 2/N < \varepsilon$. But $x_n \to 0 \notin (0,1)$, so the sequence does not converge in $X$. Thus $(0,1)$ with the Euclidean metric is not complete. However, $[0,1]$ is complete (closed subset of complete $\mathbb{R}$). $\blacksquare$