Skip to content
Tutorials Center
My learning

Topics

Lesson 9 of 36 ยท Metric Spaces

Convergence and Completeness

Understand convergent and Cauchy sequences in metric spaces, why the rationals have holes, and what completeness means, with exact-arithmetic code experiments.

Sequences are the first place where a metric does real work. With distance we can say what it means for a list of points to approach a target, and we can ask whether a list that looks like it is settling down actually has something to settle on. That second question leads to completeness, one of the most important properties a metric space can have.

Convergence

A sequence xโ‚, xโ‚‚, xโ‚ƒ, โ€ฆ in a metric space (X, d) converges to L โˆˆ X if for every ฮต > 0 there is an index N such that d(xโ‚™, L) < ฮต for all n โ‰ฅ N. Equivalently, every open ball around L contains all the terms of the sequence from some point on. We write xโ‚™ โ†’ L.

Two quick facts:

  • Limits are unique in a metric space. If xโ‚™ โ†’ L and xโ‚™ โ†’ M, then d(L, M) โ‰ค d(L, xโ‚™) + d(xโ‚™, M), which can be made smaller than any positive number, so d(L, M) = 0 and L = M.
  • The metric matters. The sequence 1, 1/2, 1/3, โ€ฆ converges to 0 with the usual distance. In the discrete metric a sequence converges only if it is eventually constant, because the open ball B(L, 1/2) is just .
Python How far must you go? Finding N for each epsilon

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Cauchy sequences

Sometimes you want to know a sequence is settling down without knowing the limit in advance. A sequence is Cauchy if for every ฮต > 0 there is N with d(xโ‚˜, xโ‚™) < ฮต for all m, n โ‰ฅ N: eventually all terms are close to each other.

Every convergent sequence is Cauchy: if both xโ‚˜ and xโ‚™ are within ฮต/2 of L, they are within ฮต of each other. The reverse is the interesting question. Does every Cauchy sequence converge?

Completeness

A metric space is complete if every Cauchy sequence in it converges to a point of the space.

  • โ„ is complete. This is essentially the completeness axiom of the real numbers.
  • โ„โฟ with the Euclidean metric is complete.
  • The rational numbers โ„š are not complete.
  • The open interval (0, 1) with the usual distance is not complete: 1/2, 1/3, 1/4, โ€ฆ is Cauchy, but its limit 0 is not in (0, 1).
  • A closed subset of a complete space is complete; a complete subset of any metric space is closed.

Completeness is a property of the metric, not of the topology alone: (0, 1) and โ„ have the same topology (they are homeomorphic) but only one of them is complete. This is a good example of something that is not a topological invariant.

Why โ„š has holes

Start with xโ‚ = 1 and apply Newton's method for xยฒ = 2: x_{n+1} = (xโ‚™ + 2/xโ‚™) / 2. Every term is a fraction, each step lands closer, and the terms are Cauchy. But the only candidate limit is โˆš2, which is irrational. So the Cauchy sequence has no limit in โ„š.

The code uses exact fractions so there is no hiding rounding error. It also shows that the squares approach 2 while the terms never reach it.

Python A Cauchy sequence of fractions with no limit in Q

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The final line prints False. We cannot test every possible rational limit by running code, but a classical proof shows that no fraction squares to 2.

A Cauchy sequence that does not converge

Take the metric space X = (0, 1] with the usual distance, and the sequence xโ‚™ = 1/n. The terms get arbitrarily close to each other, so it is Cauchy. In โ„ it converges to 0, but 0 โˆ‰ X, so inside X it has no limit. The space X has a missing point, and that is exactly what incompleteness means: a hole where a limit should be.

Python Cauchy but not convergent in (0, 1]

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Completeness and the next steps

In later chapters, completeness reappears in analysis (the Banach fixed-point theorem) and in the contrast with compactness, a stronger property that implies completeness for metric spaces. Another useful fact: every metric space can be completed by adding the missing limits. โ„ is the completion of โ„š, and that is one way real numbers are built.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. What does it mean for a sequence in a metric space to be Cauchy?

    Cauchy means that for every epsilon the terms beyond some index are within epsilon of each other; no limit is mentioned.
  2. 2. Which space is NOT complete?

    Newton's iteration for sqrt(2) gives a Cauchy sequence of rationals with no rational limit.
  3. 3. In a set with the discrete metric, when does a sequence converge?

    The ball of radius 1/2 around the limit is a single point, so all but finitely many terms must equal the limit.
  4. 4. Every convergent sequence in a metric space is...

    If both terms are within epsilon/2 of the limit, they are within epsilon of each other.

Summary

  • xโ‚™ โ†’ L means every ball around L eventually contains the whole tail of the sequence.
  • Limits in metric spaces are unique; whether a sequence converges depends on the metric.
  • Cauchy sequences have terms that become close to each other; every convergent sequence is Cauchy.
  • A space is complete when every Cauchy sequence has a limit inside it; โ„ is complete, โ„š and (0, 1) are not.
  • Completeness is not preserved by homeomorphisms, so it is a metric property, not a purely topological one.

Next: Definition of a Topology