1 Foundations: Metric spaces and distance
1.1 Metric space basics
A metric space is a set equipped with a distance function that measures how far two points are from each other. Formally, a metric space \((X,d)\) consists of a set \(X\) and a function \(d:X\times X\to[0,\infty)\) satisfying: \(d(x,y)=0\) iff \(x=y\), symmetry \(d(x,y)=d(y,x)\), and the triangle inequality \(d(x,z)\le d(x,y)+d(y,z)\). These axioms encode the intuitive behavior of distance without requiring coordinates.
1.2 Neighborhoods via open balls
The metric determines a natural notion of neighborhood. For a point \(a\in X\) and radius \(\varepsilon>0\), the open ball \(B(a,\varepsilon)=\{x\in X: d(x,a)<\varepsilon\}\) collects points that lie within distance \(\varepsilon\) of \(a\). Neighborhood statements about limits can then be translated into statements about whether sequence terms eventually enter every such ball.
1.3 Limits of functions versus limits of sequences
Convergence is often first encountered through sequences, but it also underlies limits of functions. If \(f:X\to Y\) maps between metric spaces, the limit of \(f(x)\) as \(x\) approaches a point \(a\) can be expressed using sequences: one tests \(f(x_n)\) along sequences \(x_n\to a\). This bridges pointwise behavior with sequence-based definitions, enabling many standard results in analysis to be proved in a unified manner.
2 Definition of convergence
2.1 Convergence of sequences
Let \((X,d)\) be a metric space and \((x_n)\) a sequence in \(X\). The sequence converges to \(x\in X\) if the terms eventually become arbitrarily close to \(x\).
2.1.1 Epsilon–ball formulation
The standard definition uses \(\varepsilon\)-balls: \((x_n)\) converges to \(x\) if for every \(\varepsilon>0\) there exists \(N\) such that for all \(n\ge N\), \(d(x_n,x)<\varepsilon\). Intuitively, no matter how tight a distance requirement is imposed, the sequence ultimately satisfies it forever.
2.1.2 Examples illustrating convergence
| In \(\mathbb{R}\) with the usual metric \(d(a,b)= | a-b | \), the sequence \(x_n=1/n\) converges to \(0\) because \( | 1/n-0 | <\varepsilon\) for all sufficiently large \(n\). In a discrete metric space where \(d(x,y)=0\) if \(x=y\) and \(1\) otherwise, convergence to \(x\) means the sequence is eventually constant equal to \(x\). In general metric spaces, sequences may approach points along complicated paths, but the criterion always reduces to distance shrinkage. |
|---|
2.2 Equivalent formulations
Many alternative characterizations are useful, especially when proving theorems.
2.2.1 Sequential characterization using balls
A common reformulation uses the ball language directly: \((x_n)\to x\) iff for every \(\varepsilon>0\), there exists \(N\) so that \(n\ge N\) implies \(x_n\in B(x,\varepsilon)\). This is essentially the same statement as the epsilon condition, expressed as membership in shrinking neighborhoods.
2.2.2 Convergence in terms of distances to the limit
Sometimes it is convenient to look at the real sequence \(d(x_n,x)\). Convergence \((x_n)\to x\) is equivalent to \(d(x_n,x)\to 0\) in the usual real-number sense. This emphasizes that convergence depends only on how the distance from \(x_n\) to the candidate limit behaves.
2.3 Nonconvergence and failure modes
A sequence might fail to converge for several distinct reasons. It can oscillate between separated regions (never settling within a small ball), or it can drift without approaching any specific point. In noncompact settings, boundedness alone does not guarantee convergence; the sequence may have accumulation points but still fail to settle on one limit. Metric structure makes such failures precise: for nonconvergence to \(x\), there must exist some \(\varepsilon>0\) such that infinitely many terms remain at distance at least \(\varepsilon\) from \(x\).
3 Uniqueness and elementary properties
3.1 Uniqueness of limits
Limits in metric spaces are unique. If \((x_n)\to x\) and \((x_n)\to y\), then \(x=y\). The argument relies on choosing \(\varepsilon\) smaller than half the distance between \(x\) and \(y\); eventually the sequence must lie in both corresponding balls, which is impossible unless the balls overlap—forcing \(d(x,y)=0\).
3.2 Convergent sequences are bounded
If \((x_n)\to x\), then \((x_n)\) is bounded. Indeed, choose \(\varepsilon=1\) (or any fixed positive radius). For large \(n\), \(x_n\in B(x,1)\), so distances from \(x\) are uniformly controlled; the finitely many initial terms cannot destroy boundedness. Boundedness is thus a necessary condition for convergence.
3.3 Subsequence behavior
Subsequences preserve the convergence behavior, with limits aligning when convergence occurs.
3.3.1 Every convergent sequence has convergent subsequences
Any subsequence of a convergent sequence converges to the same limit. This follows because the tail of the original sequence eventually falls inside every ball around the limit, and every subsequence eventually samples terms from that tail as well.
3.3.2 Limits of subsequences agree with the original limit
Conversely, if a sequence has a subsequence converging to \(x\), that does not automatically mean the whole sequence converges to \(x\). However, if the original sequence is known to converge, then any convergent subsequence must share that same limit by uniqueness.
4 Cauchy sequences and completeness
4.1 Definition of Cauchy sequence
A sequence \((x_n)\) is Cauchy if its terms become close to each other as the index grows, regardless of any specific candidate limit. Formally, \((x_n)\) is Cauchy if for every \(\varepsilon>0\) there exists \(N\) such that for all \(m,n\ge N\), \(d(x_m,x_n)<\varepsilon\).
4.2 Cauchy criterion for convergence
In a complete metric space, Cauchy sequences converge. This yields a powerful criterion: rather than guessing a limit, one can test whether the sequence is internally stable under the metric. In general metric spaces that are not complete, a sequence can be Cauchy without converging to a point inside the space.
4.3 Complete metric spaces
A metric space is complete if every Cauchy sequence converges to a point of the space. Completeness is the property that prevents “missing limit points.” Many classical spaces used in analysis are complete, including \(\mathbb{R}^n\) with the Euclidean metric and function spaces under suitable norms.
4.3.1 Convergence versus Cauchy in complete spaces
In complete metric spaces, convergence and the Cauchy property coincide: every convergent sequence is Cauchy in any metric space, and every Cauchy sequence converges precisely when the space is complete. This equivalence is central to constructing limits and proving existence results.
4.4 Construction and interpretation of limits
Cauchy sequences can be viewed as “approximations that refine themselves.” Each term is close to later terms, so the sequence behaves like a convergent approximation even when the limit is not explicitly known. Completeness guarantees that such approximations correspond to an actual point in the space, giving a conceptual basis for limit constructions in proofs.
5 Continuity and convergence
5.1 Continuity via sequential limits
Continuity of a function between metric spaces can be characterized using sequences. A map \(f:X\to Y\) is continuous at \(a\in X\) if whenever \(x_n\to a\), one has \(f(x_n)\to f(a)\). This criterion is often easier to verify in practice than working directly with \(\varepsilon\)-\(\delta\) definitions, because it leverages the metric’s sequential structure.
5.2 Convergence under continuous maps
Continuous functions respect convergence: they send limits of sequences to limits of the images.
5.2.1 Image of a convergent sequence
If \(x_n\to a\) and \(f\) is continuous at \(a\), then \(f(x_n)\) converges to \(f(a)\). The behavior is not merely pointwise; it preserves the “eventual closeness” encoded by the metric.
5.2.2 Limit preservation
When continuity holds not just at a point but on a set, limit preservation applies to sequences approaching points within that set. This allows many arguments in analysis to be conducted by tracking convergence through function composition and algebraic operations that are compatible with the metric.
5.3 Uniform continuity and stronger convergence statements
Uniform continuity strengthens the relationship between closeness in the domain and closeness in the codomain by using one control function for all points, not just near a single \(a\). Under uniform continuity, one can often pass convergence without restricting to a particular point of approach, and one can establish more robust statements when sequences are controlled globally. In many analytic settings, uniform continuity supports interchange of limits and estimates across varying inputs.
6 Convergence of series and norms (common analytic settings)
6.1 Series convergence in metric spaces from induced metrics
In a metric space, series \(\sum_{n=1}^\infty a_n\) are defined through the convergence of partial sums \(s_N=\sum_{n=1}^N a_n\). Thus, series convergence reduces to sequence convergence in the metric space hosting the terms. The metric provides the meaning of “distance between partial sums” and therefore the notion of whether the sequence of partial sums approaches a point.
6.2 Normed spaces as metric spaces
| A norm \(\|\cdot\|\) on a vector space induces a metric via \(d(x,y)=\|x-y\|\). In normed spaces, convergence \(x_n\to x\) translates to \(\|x_n-x\|\to 0\). This turns abstract convergence into a familiar analytic form and aligns with the way series, operators, and estimates are typically handled. |
|---|
6.3 Absolute versus unconditional convergence (where applicable)
| In many Banach spaces (complete normed spaces), distinctions between types of convergence become important. For scalar series in \(\mathbb{R}\) or \(\mathbb{C}\), absolute convergence implies convergence. In normed spaces, one can distinguish between convergence of \(\sum a_n\) and stronger conditions like \(\sum \|a_n\|\) being finite (absolute convergence), or conditions under which rearrangements of series do not change the sum (unconditional convergence). These distinctions depend on the structure of the space and the nature of the series. |
|---|
7 Topological connections
7.1 Induced topology from a metric
Every metric space determines a topology: open sets are those that can be expressed as unions of open balls. This means convergence, closure, and continuity can be formulated in topological language while still relying on metric definitions when convenient. The resulting topology is Hausdorff, ensuring limits—when they exist—are unique.
7.2 Closed sets and convergence
A key interaction between topology and convergence is that convergence can be tested against closed sets. If \(F\subseteq X\) is closed and \(x_n\in F\) for all \(n\), then any limit of \((x_n)\) must lie in \(F\). Conversely, if whenever a sequence in \(F\) converges its limit lies in \(F\), then \(F\) is closed.
7.3 Closure characterization via limits
The closure of a set \(A\), denoted \(\overline{A}\), consists precisely of points that can be approached by limits of sequences from \(A\). In metric spaces this can often be described explicitly: \(x\in\overline{A}\) iff there exists a sequence \((a_n)\subseteq A\) such that \(a_n\to x\). This provides a practical bridge between set-theoretic closure and constructive approximation.
8 Extended notions (when needed)
8.1 Convergence of filters (high-level viewpoint)
Filters generalize sequences by describing “eventual behavior” more flexibly. A filter on \(X\) is a collection of subsets closed under supersets and finite intersections, representing sets that are considered large or typical. Convergence of a filter to a point \(x\) means that every neighborhood of \(x\) is eventually present in the filter. This viewpoint is useful in settings where sequences are insufficient to capture all topological phenomena.
8.2 Nets and general spaces beyond sequences
Nets extend sequences by allowing directed index sets, enabling convergence in more general topological spaces. A net \((x_\alpha)\) is indexed by a directed set and “eventually” refers to indices beyond any given one in the direction. Nets recover sequence convergence when the index set is \(\mathbb{N}\) and can capture convergence in situations where sequences may not.
8.2.1 Relationship to metric-space sequential convergence
In metric spaces, the extra generality of nets is usually unnecessary for describing convergence because metric topologies are sufficiently well behaved: sequential information captures the topological notion of closure and continuity. Thus, while filters and nets are crucial in broader topology, metric spaces allow convergence to be handled entirely through sequences and their \(\varepsilon\)-ball behavior.