Metric sequences, closed sets and continuity

Try

In a metric space, x_n→a means d(x_n,a)→0. This definition needs no coordinate subtraction. In the discrete metric, convergence is especially restrictive: using tolerance 1/2 forces every sufficiently late term to equal a. In Euclidean space, a sequence can approach without ever reaching its limit. The same definition produces different behavior because the neighborhoods differ.

Predict

Predict which familiar proofs survive unchanged. Limit uniqueness uses d(a,b)≤d(a,x_n)+d(x_n,b), so it needs only the triangle inequality. Continuity of a map between metric spaces compares an input distance with an output distance; these may be different metrics on completely different sets. A proof that subtracts two points is therefore not general enough unless those points belong to a vector space.

Experiment

Translate the concepts.

Real-line expression	Metric-space replacement
|x_n−a|<ε	d_X(x_n,a)<ε
|f(x)−f(a)|<ε	d_Y(f(x),f(a))<ε
An open interval around a	An open ball around a

Use the table to rewrite a continuity proof without algebraic coordinates. Keep source metric d_X and target metric d_Y labeled; confusing them can hide an invalid estimate when the map changes spaces.

Understand

A continuous map preserves sequence limits. Given ε in the target, continuity supplies δ in the source, and convergence eventually places inputs within that δ-ball. Conversely, failure of continuity at a gives a fixed ε₀>0 and, for every n, a point x_n within 1/n of a whose image stays at least ε₀ from f(a). This sequence disproves preservation. Thus sequential continuity and epsilon–delta continuity are equivalent for metric spaces.

There is also an open-set characterization. If f is continuous and V is open in the target, every a in f⁻¹(V) has an image neighborhood inside V; continuity supplies a source neighborhood inside the preimage. Conversely, if all open preimages are open, take the preimage of an ε-ball around f(a). It contains a and hence a source δ-ball, giving continuity. This proof works without numerical formulas for the map.

Apply

The distance-to-a-point function is continuous: the triangle inequality in both directions gives |d(x,a)−d(y,a)|≤d(x,y). Consequently closed balls are closed, since their complement can be described by distance greater than a radius. They need not be compact; the infinite discrete space’s radius-one closed ball is the entire noncompact space.

Repair “closed sets make their sequences converge.” Closedness says only that an already convergent sequence of set points has its limit in the set. Conversely, if a subset misses a closure point a, choose set points within 1/n of a to witness failure of that property. Completeness is not required for this characterization because the claimed limit is already a point of the ambient space.

Continue

Optional depth notes that sequential characterizations need not describe every possible topological space, but they do work throughout metric spaces. That boundary matters when later topology goes beyond distances.

Save one epsilon–delta proof, one inverse-image proof and one sequence-counterexample for maps between explicit metric spaces. Explain which formulation makes the task shortest without weakening its quantifiers. Next, completeness will ask whether Cauchy sequences have limits in the space itself, separating a distance-based tail property from the existence of missing points.