Metric Topology Review: Essential Concepts Made Simple

A metric space is a set X with a distance function d(x, y) that is non-negative, zero exactly when x equals y, symmetric, and satisfies the triangle inequality. The metric generates a topology: a set is open if around every point there is a small open ball contained in the set. This is the setting of most of classical analysis.

The review matters because nearly every definition in analysis is a statement about a metric topology: convergence, continuity, compactness, and completeness all start here. Results proved for general metric spaces apply instantly to the real line, Euclidean spaces, and function spaces. If the metric picture is sharp, the abstract versions become easy.

Key Concepts

• Open ball B(x, r) = {y : d(x, y) < r}; open balls form a base for the metric topology.
• Open and closed sets: open sets contain a ball around each point; closed sets contain all their limit points.
• Convergence: x_n converges to x when d(x_n, x) tends to 0.
• Continuity f is continuous at x if small d(x, y) forces small d(f(x), f(y)) (epsilon-delta).
• Completeness: every Cauchy sequence converges; R and R^n are complete.
• Bounded and totally bounded: a set is bounded if it fits in some ball; totally bounded if it can be covered by finitely many balls of any prescribed radius.

Formulas

• Euclidean metric d(x, y) = sqrt(sum (x_i - y_i)^2), in the same units as the coordinates.
• Convergence test: x_n -> x exactly when d(x_n, x) < epsilon eventually, for every epsilon > 0.
• Continuity modulus: uniform continuity on a set means delta depends only on epsilon, not on the point; every continuous function on a compact metric space is uniformly continuous (Heine-Cantor theorem).

Worked Example

In R^2 with the Euclidean metric, is p = (3, 3.9) in the open ball B((0, 0), 5)? Compute d(p, (0,0)) = sqrt(9 + 15.21) = sqrt(24.21), about 4.92. Since 4.92 is smaller than 5, p is inside the ball. Note q = (3, 4) has distance exactly 5, so q is on the boundary sphere and is not in the open ball. For convergence, x_n = (1/n, 2 + 1/n) has distance sqrt(2)/n to (0, 2), which tends to 0, so x_n converges to (0, 2).

Common Misconceptions

• "Open balls are always open in the topological sense." In a metric space this is true by definition, but for a general distance-like function it can fail.
• "Closed means not open." Sets can be both open and closed (like the whole space and the empty set), or neither; closed means containing all limit points.
• "Bounded implies totally bounded." In infinite dimensions bounded sets need not be totally bounded, so the unit ball need not be compact.

Real-World Applications

• Optimization: gradient descent convergence proofs live in metric spaces, with the metric measuring distance to the optimum.
• Numerical methods: error bounds are metric estimates, and completeness guarantees Cauchy approximations converge to a solution.
• Machine learning: clustering, nearest neighbors, and embeddings all depend on choosing a good metric on the data space.
• Differential equations: existence proofs apply fixed-point theorems in complete metric spaces of functions.

Metric topology is analysis with a ruler: open balls generate the topology, the triangle inequality does the heavy lifting, and completeness plus total boundedness characterize compactness. Keep the epsilon-delta definitions beside the ball picture and the abstract theory feels concrete.