Metric Spaces: Basic Concepts and Examples

Motivation: Distance as an Abstraction

The concept of “distance” is ubiquitous: the distance between points on a plane, between functions (how similar are two curves?), between strings (how many substitutions are needed to turn one word into another?). A metric space is the minimal abstraction that encapsulates the axioms of distance and enables analytic methods to be transferred into various contexts. Functional analysis studies infinite-dimensional metric spaces, where the objects are functions themselves.

Definition of a Metric Space

Metric space (X, d): a set X with a function metric d: X×X → ℝ₊, satisfying:

Positivity: d(x,y) ≥ 0; d(x,y) = 0 ⟺ x = y.
Symmetry: d(x,y) = d(y,x).
Triangle inequality: d(x,z) ≤ d(x,y) + d(y,z).

The axioms are minimal: everything needed for analysis—convergence, continuity, compactness—follows from them. The concepts of “neighborhood”, “open set”, “closed set” are transferred literally.

Important Examples

Euclidean space ℝⁿ: d(x,y) = √(Σᵢ(xᵢ−yᵢ)²). The standard example.

Spaces lᵖ (1 ≤ p ≤ ∞): sequences (xₙ) with ‖x‖_p < ∞:

p < ∞: ‖x‖_p = (Σ|xₙ|ᵖ)^{1/p}.
p = ∞: ‖x‖_∞ = sup|xₙ|.

C[a,b]: continuous functions on [a,b]. Supremum metric: d(f,g) = max|f(x)−g(x)|.

Lᵖ[a,b]: measurable functions with ∫|f|ᵖ < ∞; functions coinciding almost everywhere are identified.

Edit metric (Levenshtein): on strings—the minimum number of insertions/deletions/substitutions. d(“cat”, “cod”) = 1. Used in spell-checkers and bioinformatics.

Completeness

Cauchy sequence: {xₙ} is a Cauchy sequence if ∀ε>0 ∃N: n,m>N ⟹ d(xₙ,xₘ) < ε. Any convergent sequence is Cauchy. The converse is not true in incomplete spaces.

Complete metric space: every Cauchy sequence converges. ℝⁿ, lᵖ, Lᵖ[a,b], C[a,b] with the supremum norm are complete. ℚ with the usual metric is incomplete (√2 ∉ ℚ, yet can be approximated by rational numbers).

Baire theorem (category): A complete metric space cannot be represented as a countable union of nowhere dense sets. A powerful tool: proof of existence of a continuous nowhere differentiable function.

Numerical Example

Problem: Show that the sequence xₙ = Σₖ₌₁ⁿ (−1)ᵏ/k in ℝ is a Cauchy sequence and find its limit.

Step 1. The series Σ (−1)ᵏ/k is alternating. By the Leibniz criterion: aₙ = 1/n → 0, and decreases monotonically → series converges. Therefore, {xₙ} converges, hence is a Cauchy sequence in ℝ.

Step 2. Estimate |xₙ − xₘ| for n < m. By the Leibniz criterion: |xₙ − S| ≤ 1/(n+1), where S is the sum of the series. Then |xₙ − xₘ| ≤ |xₙ − S| + |S − xₘ| ≤ 1/(n+1) + 1/(m+1) < 2/(n+1) → 0.

Step 3. Limit: Σₖ₌₁^∞ (−1)ᵏ/k = −1 + 1/2 − 1/3 + 1/4 − ... = −ln(2) ≈ −0.693. From the expansion of ln(1+x) at x=1: ln(2) = 1 − 1/2 + 1/3 − ..., hence Σ(−1)ᵏ/k = −ln(2).

Step 4. The same sequence in ℚ: {xₙ} ⊂ ℚ, limit ln(2) ∉ ℚ. In (ℚ,d) the Cauchy sequence does not converge → ℚ is incomplete ✓.

Step 5. L¹ metric on C[0,1]: fₙ(x) = xⁿ. d(fₙ,0) = ∫₀¹ xⁿ dx = 1/(n+1) → 0. But fₙ(1) = 1 ≠ 0(1)—the limit in L¹ is not a continuous function equal to zero. Hence (C[0,1], d_{L¹}) is incomplete ✓.

Real-World Application

Machine learning: the parameter space of a neural network is a metric space. Completeness guarantees that the gradient descent algorithm does not “fall outside” the space. Baire’s theorem is used in the proof of the universal approximation theorem for neural networks.

Additional Aspects

Metric and normed spaces form the language on which most of modern analysis is formulated. Completeness (every fundamental sequence converges) divides spaces into “decent” (where existence theorems and limit transitions work) and “wild” ones. Cantor completion gives a canonical procedure to finish a space to completeness—thus ℚ becomes ℝ. On Banach spaces, three fundamental theorems of functional analysis apply: Hahn–Banach (extension of functionals), Banach–Steinhaus (uniform boundedness of operator families), and the open mapping/closed graph theorem. These theorems are practical tools for the theory of differential equations, optimization, and numerical analysis.

Connection with Other Branches of Mathematics

Metric spaces permeate the theory of differential equations via the contraction mapping principle. In a complete metric space, Banach’s theorem states uniqueness of the fixed point of a contraction operator; it is in precisely this form that the Picard–Lindelöf theorem about existence and uniqueness of the solution to the Cauchy problem for ordinary differential equations is formulated. Similar ideas are used in the theory of Fredholm and Volterra integral equations, where the space of functions is endowed with an appropriate metric and the solution operator is viewed as a contraction.

Connection with algebra is manifest in the notion of a normed algebra and completions. Construction of the completion of a metric space is a particular case of universal constructions in the category of modules and rings: from rational numbers with the p-adic metric one obtains fields Q_p, which are foundational in algebraic number theory. The concept of a metric group, and subsequently a locally compact Haar group, links the metric with harmonic analysis and group representations.

Topology abstracts the metric structure, but many central results, such as metrizability by Urysohn and Nagata–Smirnov, describe when a purely topological space admits a metric compatible with its open sets. In probability theory, metrics on distribution spaces, such as the Prokhorov metric or Wasserstein distance, allow convergence in distribution to be formulated as convergence in appropriate complete separable metric spaces (Polish spaces). This underlies the theorems of functional limit theory, for instance, Donsker’s invariance principle. In numerical methods, the metric sets the criterion for error and algorithm stability: Lax’s theorem on the equivalence of stability and convergence of schemes for linear problems relies on the norm (metric) in spaces of grid functions.

Historical Reference and Development of the Idea

The origin of the concept of abstract distance is tied to work in geometry at the end of the 19th century. F. Riesz and M. Fréchet in the early 20th century consciously separated the notion of metric from specific geometric models. Fréchet in his 1906 dissertation at Toulouse University introduced the terms “metric space” and “functional space”, considering sets of functions with a prescribed distance between them; part of the results was published in Rendiconti del Circolo Matematico di Palermo. The idea of completeness traces back to Cauchy and Cantor, who constructed the real numbers as the completion of rationals via fundamental sequences. Hausdorff in “Grundzüge der Mengenlehre” (1914) formed the axiomatic framework for topological and metric spaces, while Baire in 1899–1909 formulated the categorical theorem, revealing the unexpected profundity of completeness.