1 Definition and basic intuition
A Cauchy sequence is a sequence in which the terms become arbitrarily close to each other as the sequence progresses, without requiring that the sequence’s destination point be known in advance. Informally, the later terms form a “tight cluster”: once the index is large enough, any two sufficiently far-out terms are close.
1.1 Metric-space formulation
Let \((X,d)\) be a metric space and \((x_n)\) a sequence in \(X\). The sequence is called Cauchy if:
For every \(\varepsilon>0\) there exists \(N\) such that for all \(m,n\ge N\), \[ d(x_m,x_n)<\varepsilon. \]
This definition depends only on the metric \(d\), so it generalizes naturally across settings where “distance” is defined.
1.2 Epsilon–N characterization
The epsilon–\(N\) form is the standard way to express the idea of “eventually within tolerance.” The tolerance is \(\varepsilon\), and the point in the sequence after which the tolerance holds is \(N\). Once such an \(N\) exists for every \(\varepsilon\), the sequence is Cauchy.
1.3 Relation to pairwise closeness of terms
The Cauchy property is stronger than merely saying that \(x_n\) is close to some fixed element. It is instead a statement about pairwise distances between terms: not just \(d(x_n,x_{n+k})\) for one fixed shift \(k\), but \(d(x_m,x_n)\) uniformly over all pairs far enough along the sequence.
1.4 Examples and non-examples
- Example (convergent sequences): In any metric space, if \(x_n\to x\), then \((x_n)\) is Cauchy. (A proof appears in §2.1.)
- Example (Cauchy but not convergent): In an incomplete metric space, one can construct Cauchy sequences that fail to converge within the space. A classic source is \(\mathbb{Q}\) with the usual absolute-value metric: certain rational sequences “approach” an irrational limit but never converge in \(\mathbb{Q}\).
- Non-example (oscillation without tightening): If the sequence keeps jumping between two regions separated by a fixed positive distance, then it cannot be Cauchy. For instance, a sequence alternating between points at distance \(1\) will not satisfy the “eventually closer than \(\varepsilon\)” condition for \(\varepsilon<1\).
2 Properties of Cauchy sequences
Cauchy sequences behave predictably under basic operations and functorial constructions. They also connect tightly to the structure of the ambient space, especially regarding completeness and uniqueness.
2.1 Every convergent sequence is Cauchy
If \(x_n\to x\) in a metric space \((X,d)\), then \((x_n)\) is Cauchy.
2.1.1 Proof sketch using the triangle inequality
Given \(\varepsilon>0\), choose \(N\) such that \(d(x_n,x)<\varepsilon/2\) for all \(n\ge N\). Then for \(m,n\ge N\), \[ d(x_m,x_n)\le d(x_m,x)+d(x,x_n) < \varepsilon/2+\varepsilon/2=\varepsilon. \] Hence the Cauchy condition holds.
2.2 Subsequence behavior
Cauchy sequences retain their Cauchy character under passage to subsequences.
2.2.1 Subsequences of Cauchy sequences are Cauchy
Let \((x_{n_k})\) be a subsequence. If \((x_n)\) is Cauchy, then for any \(\varepsilon>0\), there exists \(N\) such that all \(x_m,x_n\) with indices \(\ge N\) are within \(\varepsilon\). Pick \(K\) so that \(n_k\ge N\) for all \(k\ge K\). Then for \(i,j\ge K\), \[ d(x_{n_i},x_{n_j})<\varepsilon, \] so the subsequence is Cauchy.
2.3 Boundedness consequences
In metric spaces, Cauchy sequences are typically bounded in the sense that all sufficiently far-out terms lie in some ball of finite radius. More precisely, choose \(\varepsilon=1\). Then there is \(N\) such that for all \(n\ge N\), \[ d(x_n,x_N)<1, \] so the tail \(\{x_n:n\ge N\}\) lies in the open ball centered at \(x_N\) with radius \(1\). The initial finitely many terms can be handled separately.
2.4 Uniqueness of limits in Hausdorff spaces
In a Hausdorff topological space, limits of sequences—when they exist—are unique. Since every Cauchy sequence in a complete metric space converges, and because limits in metric spaces are unique, a Cauchy sequence cannot “converge to two different points” in such settings. The Hausdorff property ensures that two distinct candidate limits can be separated by disjoint neighborhoods, contradicting convergence.
2.5 Compatibility with Lipschitz and uniform maps
Cauchy sequences interact well with certain maps:
- Lipschitz maps: If \(f:(X,d_X)\to (Y,d_Y)\) satisfies
\[ d_Y(f(x),f(x'))\le L\, d_X(x,x') \] for some \(L\ge 0\), then \(f\) sends Cauchy sequences in \(X\) to Cauchy sequences in \(Y\). The same \(\varepsilon\) criterion holds after scaling by \(L\).
- Uniformly continuous maps: A uniformly continuous function similarly preserves the Cauchy property. Uniform continuity ensures that small distances in \(X\) force small distances in \(Y\) uniformly over the whole space, allowing the “eventually close” behavior to transfer.
3 Cauchy sequences and completeness
Completeness is exactly the condition that turns the Cauchy notion into convergence.
3.1 Complete metric spaces
A metric space \((X,d)\) is complete if every Cauchy sequence in \(X\) converges to a point in \(X\). Completeness is the backbone of much of analysis: it guarantees that “approaches” within the space do not fall short of an actual limit.
3.2 Cauchy criterion for convergence
In a complete metric space, the following becomes equivalent:
- \((x_n)\) converges in \(X\)
- \((x_n)\) is Cauchy
Thus, the Cauchy criterion provides a practical method for proving convergence: rather than guessing a candidate limit, one verifies the internal closeness of the sequence’s tail.
3.3 Real and rational numbers
The real line and the rationals illustrate how completeness matters.
3.3.1 Why rationals are not complete
Consider a rational sequence that approximates \(\sqrt{2}\) increasingly well. Such a sequence can be made Cauchy in \(\mathbb{Q}\) under the usual absolute-value metric, because the approximations get closer and closer to each other. However, its limit in \(\mathbb{R}\) is irrational, so the sequence cannot converge within \(\mathbb{Q}\). Therefore \(\mathbb{Q}\) fails completeness.
3.3.2 Construction leading to real numbers
One standard approach to constructing \(\mathbb{R}\) from \(\mathbb{Q}\) is to “fill in” missing limits by identifying Cauchy sequences of rationals that converge to the same real quantity. This uses an equivalence relation on rational Cauchy sequences and defines operations termwise. The resulting structure yields a complete ordered field satisfying the expected properties of the real numbers.
3.4 Cauchy sequences in normed and Banach spaces
| In normed vector spaces, metric and norm formulations coincide in a natural way: the distance \(d(x,y)\) is typically \(\|x-y\|\). |
|---|
3.4.1 Banach space viewpoint
A Banach space is a normed vector space that is complete under its norm metric. Completeness in this setting ensures the existence of limits of iterative procedures and series expansions. Many foundational theorems in functional analysis rely on showing that a sequence (often built from approximations) is Cauchy, and then invoking completeness to obtain an actual element of the space.
4 Operations and transformations
Cauchy sequences form a stable framework under algebraic operations and under suitable transformations.
4.1 Sum, scalar multiple, and product rules
In normed (or more generally, metric) contexts where operations are continuous or Lipschitz-like:
- Sum: If \((x_n)\) and \((y_n)\) are Cauchy, then so is \((x_n+y_n)\), because
\[
| \| (x_m+y_m)-(x_n+y_n)\|\le \|x_m-x_n\|+\|y_m-y_n\|. |
|---|
\]
- Scalar multiple: For a scalar \(\alpha\), if \((x_n)\) is Cauchy then \((\alpha x_n)\) is Cauchy, since
\[
| \|\alpha x_m-\alpha x_n\|= | \alpha | \,\|x_m-x_n\|. |
|---|
\]
For products, one typically needs additional structure (e.g., working in a Banach algebra or using norm bounds ensuring multiplication is compatible with the metric). Under appropriate hypotheses, a product of Cauchy sequences is again Cauchy.
4.2 Quotients under non-vanishing conditions
Quotients behave well when denominators do not approach zero. For sequences \(x_n\) and \(y_n\), a typical condition is that \(y_n\) is bounded away from \(0\) in norm (for all sufficiently large \(n\)), and both numerator and denominator sequences are Cauchy. Then \(x_n/y_n\) becomes Cauchy due to estimates that control \[ \frac{x_m}{y_m}-\frac{x_n}{y_n} \] by a combination of differences \(x_m-x_n\) and \(y_m-y_n\).
4.3 Composition with continuous functions
Continuity alone does not always preserve the Cauchy property in arbitrary metric spaces: the needed control is often uniform rather than pointwise. However, in many common analytic settings, one uses uniform continuity (or continuity on compact sets) to ensure that images of Cauchy sequences remain Cauchy.
4.4 Uniform continuity and Cauchy sequences
If \(f\) is uniformly continuous and \((x_n)\) is Cauchy in \(X\), then \((f(x_n))\) is Cauchy in \(Y\). The argument uses the definition of uniform continuity to find, for a given tolerance in \(Y\), a corresponding tolerance in \(X\) that works simultaneously for all large indices.
5 Special forms of Cauchy sequences
Several important families of Cauchy sequences arise naturally in analysis and can be characterized more explicitly.
5.1 Cauchy sequences in \u211d (absolute value metric)
| In \(\mathbb{R}\) with the metric \(d(a,b)= | a-b | \), the Cauchy condition becomes: for every \(\varepsilon>0\), there is \(N\) such that for \(m,n\ge N\), |
|---|
\[
| x_m-x_n | <\varepsilon. |
|---|
\] This is the setting where many standard estimates are easiest to implement.
5.2 Cauchy sequences in \u211d^n and Euclidean spaces
| In Euclidean space with the norm induced by the dot product, \(\|u-v\|\) replaces absolute value. The Cauchy property takes the same form: |
|---|
\[
| \|x_m-x_n\|<\varepsilon\quad (m,n\ge N). |
|---|
\] Because all norms on finite-dimensional spaces are equivalent, one can often use whichever norm simplifies computations.
5.3 Monotone Cauchy sequences
A sequence that is both monotone (nondecreasing or nonincreasing) and Cauchy has particularly strong behavior. Intuitively, monotonicity rules out oscillation, while the Cauchy property forces tightening. In ordered settings like \(\mathbb{R}\), these conditions help identify the limit and provide alternative convergence criteria.
5.4 Cauchy sequences from series
A central construction is: partial sums of a series often form a Cauchy sequence. If \[ s_n=\sum_{k=1}^n a_k, \] then \((s_n)\) is Cauchy exactly when the series behaves like “the tails go to zero” in a quantitative sense. This links the Cauchy concept to convergence of infinite sums.
5.5 Relationship to convergence of series
If the space is complete and \((s_n)\) is Cauchy, then it converges, meaning the series \(\sum a_k\) converges. Conversely, if a series converges, its partial sums form a Cauchy sequence. This equivalence is widely used to prove convergence via tail estimates rather than by summing explicitly.
6 Common criteria and test techniques
In practice, verifying the Cauchy property often relies on inequalities that bound distances between distant terms by manageable expressions.
6.1 Using the tail definition efficiently
To use the tail definition, one aims to find a single \(N\) such that the inequality holds for all indices beyond it. This is typically achieved by bounding \(d(x_m,x_n)\) with a function that decreases as the indices increase.
6.2 Comparison estimates in proofs
A common strategy is to compare distances with a simpler quantity known to shrink. For instance, one may show \[ d(x_m,x_n)\le g(\min\{m,n\}) \] where \(g(k)\to 0\). Then choosing \(N\) such that \(g(N)<\varepsilon\) completes the Cauchy proof.
6.3 Telescoping and remainder bounds
Telescoping identities often convert complicated differences into a sum of manageable “increments.” If the increments can be bounded and the remainder term can be made small, one obtains the Cauchy property for sequences derived from iterative processes or partial sums.
6.4 Cauchy criterion via partial sums
For sequences defined by partial sums, the Cauchy property can be tested by controlling the tail: \[
| d(s_m,s_n)=\left\|\sum_{k=n+1}^m a_k\right\|. |
|---|
\] Bounding this tail uniformly in \(m\ge n\) yields a direct criterion for the series to converge in a complete space.
6.5 Cauchy criteria in different norms
In normed spaces, different norms may simplify verification. In finite-dimensional settings, norm equivalence allows switching between norms without changing whether a sequence is Cauchy or convergent. In infinite-dimensional spaces, however, one must be careful: distinct norms can lead to different notions of convergence and Cauchy behavior.
7 Applications in analysis
Cauchy sequences are not only a theoretical characterization of convergence; they also drive constructive proofs.
7.1 Constructing limits and proving existence
When explicit limits are difficult to guess, showing that an approximating sequence is Cauchy can establish existence of a limit in a complete space. This underpins many methods in which one builds approximations and then uses completeness to extract the limiting object.
7.2 Completeness-based theorems
Completeness is the engine behind a range of standard results. A typical pattern is:
1 Definition and basic intuition
2 Properties of Cauchy sequences
3 Cauchy sequences and completeness
4 Operations and transformations
7.3 Role in fixed-point principles (conceptual)
Fixed-point results often involve iterative sequences generated by an operator. Under contraction-type assumptions, the iterates form a Cauchy sequence. Completeness then ensures that the iteration converges to a fixed point. While the exact theorem varies, the conceptual role of Cauchy sequences is consistent: they translate repeated approximation into actual existence.
7.4 Foundational role in real analysis
In real analysis, the Cauchy framework provides rigorous definitions of limits and underlies proofs of convergence and existence theorems. It also supports the development of integration and differentiation by enabling careful control of approximating sequences.
8 Exercises and further reading
Exercises on Cauchy sequences cultivate technical facility with inequalities, quantifiers, and metric-norm reasoning.
8.1 Typical problem types
Common tasks include:
- proving a given sequence is Cauchy by producing explicit \(N(\varepsilon)\),
- showing that a sequence is not Cauchy by constructing a fixed \(\varepsilon\) and infinitely many counterexamples,
- translating convergence questions into Cauchy questions in complete spaces,
- analyzing partial sums and series using tail estimates,
- verifying preservation of the Cauchy property under sums, scalar multiples, or uniformly continuous functions.
8.2 Suggested reading paths
Useful learning paths typically proceed from:
1 Definition and basic intuition
2 Properties of Cauchy sequences
3 Cauchy sequences and completeness
4 Operations and transformations
5 Special forms of Cauchy sequences
Textbooks in real analysis and functional analysis often revisit Cauchy sequences repeatedly in different guises.
8.3 Proof strategies and common pitfalls
- Quantifier order mistakes: Cauchy proofs require: for each \(\varepsilon\) there exists \(N\) such that for all \(m,n\ge N\). Confusing “there exists” and “for all” can invalidate arguments.
- Using pointwise continuity instead of uniform continuity: When mapping Cauchy sequences, continuity may not be enough; uniform continuity (or compactness-based uniformity) is frequently required.
- Forgetting completeness: In an incomplete space, Cauchy does not imply convergence; one must check whether the limit exists in the space.