Metric Spaces and Contraction Mappings

Imagine you have an equation you want to solve, and you try a guess‑and‑improve method. You make a guess, see how far off you are, and then take a step that brings you closer to the real answer. To know if this process will actually work, we need a precise way to talk about “closeness” and “shrinking.” That is exactly what metric spaces and contraction mappings give us—a clear way to measure distance and guarantee that certain methods settle on a unique solution.

The Big Picture#

In numerical analysis we repeatedly build sequences of approximations that we hope will settle down to the correct solution. But what does “settle down” mean? And what conditions do we need to be sure a given problem has exactly one solution and that our method will find it? This chapter develops the essential mathematical tools: metric spaces extend the everyday idea of distance; completeness makes sure that sequences that “should” converge actually stay inside the space we are working in; contraction mappings are functions that always shrink distances; and the Banach fixed point theorem brings them together by proving that contractions on complete metric spaces have a unique fixed point and that simple iteration always finds it. These ideas support existence and uniqueness proofs for differential equations and are the theory behind many numerical solvers.

Distance and Metric Spaces#

We start with the intuitive idea of distance. On the number line, the distance between two numbers and is . In the plane, the straight‑line distance from to is . These familiar formulas all obey three common‑sense rules:

1. Distance is never negative, and the distance between two points is zero only when they are the same point.
2. The distance from to is the same as from to .
3. Going directly from to is never longer than going from to and then from to (the triangle inequality).

A metric space takes these rules and uses them as a definition, without tying them to numbers. We can have any set of objects—functions, sequences, vectors, even colours—as long as we have a distance function that satisfies the three properties.

Metric space: A set together with a function is called a metric space if for all :

1. , and if and only if .
2. .
3. (triangle inequality).
   The function is called the metric. We often write .

Example (The usual plane). Take and define . This is the Euclidean metric. It clearly satisfies the first two rules; the triangle inequality follows from the geometric fact that a straight line is the shortest path between two points.

Example (Manhattan metric). On the same set , we can define a different metric: . This is the distance a taxi would travel in a city with a perfect grid of streets. The triangle inequality still holds because the sum of absolute differences obeys the ordinary triangle inequality for real numbers.

Example (Discrete metric). On any non‑empty set , define if , and if . This is indeed a metric (verify the rules), though it captures a very coarse notion of “distance” — everything is either itself or one unit away.

These examples show that a metric space gives a clear way to talk about closeness without needing coordinates. An open ball (all points with distance less than from a centre) generalises the idea of a neighbourhood, but in this chapter we will focus on sequences and convergence.

📝 Section Recap: A metric space is a set plus a way to measure distance that follows the four rules: distance is never negative, zero only for identical points, symmetric, and obeys the triangle inequality. This lets us talk about closeness in many different contexts.

Sequences, Cauchy Sequences, and Completeness#

Once we have a distance, we can say when a sequence of points settles down to a limit. Suppose is a metric space and is a sequence in . We say converges to a point (written ) if the distances become arbitrarily small as grows:

For every there exists such that .

In plain words, after some index all terms of the sequence lie inside any tiny neighbourhood of .

Convergence is a useful idea, but to prove it we usually need to guess the limit . What if we do not know the limit yet? The notion of a Cauchy sequence helps: it only demands that the terms eventually get close to each other.

Cauchy sequence: A sequence in a metric space is called Cauchy if for every there exists such that for all , .

Every convergent sequence is automatically Cauchy (by the triangle inequality, bring both terms close to the limit). But the opposite is not always true! Think of the rational numbers with the usual distance . The sequence generated by the decimal expansion of is a Cauchy sequence in — the terms squeeze together — yet it does not converge to any rational number, because is not in . The limit is “missing” from the space.

To avoid this problem, we introduce the idea of completeness:

Complete metric space: A metric space is complete if every Cauchy sequence in converges to a point that belongs to .

The real numbers with are complete — this is essentially the defining property that separates from . The Euclidean plane with is also complete. Any closed subset of (with the inherited metric) is complete, while an open interval such as is not complete: the Cauchy sequence tends to , which is not in the set.

Completeness is important for fixed‑point theorems: when we repeatedly apply a contraction, we get a Cauchy sequence, and we need to know its limit actually belongs to the space.

📝 Section Recap: A Cauchy sequence is one where the terms eventually cluster together. A space is complete if every Cauchy sequence converges to a point inside the space. Completeness ensures that a sequence that “should” converge really does have a limit in the space.

Lipschitz Functions and Contraction Mappings#

Now we look at functions that respect distance. In calculus, a function is continuous if small changes in input give small changes in output. A Lipschitz continuous function gives a number that limits how much the output can stretch compared to the input.

Let and be two metric spaces. A function is called Lipschitz continuous if there exists a constant such that for all ,

The smallest such is the Lipschitz constant. If , we give a special name:

Contraction mapping: A function from a metric space to itself is a contraction if there exists a constant with such that for all .

Think of a contraction as a function that shrinks all distances by at least a factor (strictly less than ). The simple function on is a contraction with . The function on the interval is also a contraction. Why? By the Mean Value Theorem, for any there is some between them such that . Since , we get . Moreover, maps into (because , ), so it is a contraction on that complete space.

A function need not be a contraction everywhere; often we restrict the domain to a closed bounded region where the Lipschitz constant drops below . This is a common trick when applying the fixed‑point theorem.

📝 Section Recap: A Lipschitz function does not stretch distances more than a constant factor . When and the function maps a space into itself, it is a contraction mapping — it pulls points closer together.

The Banach Fixed Point Theorem#

The main result of this chapter is the Banach fixed point theorem (also called the contraction mapping principle). It tells us when a function has exactly one fixed point, and how to find it by simply iterating from any starting point.

Banach fixed point theorem (contraction mapping principle): Let be a non‑empty, complete metric space, and let be a contraction mapping. Then has a unique fixed point (i.e. ). Moreover, for any initial guess , the sequence defined by converges to , and the error after steps satisfies

The proof is constructive and shows exactly what to do.

Proof sketch.

1. Pick any and set .
2. Apply the contraction property repeatedly:
   
3. For , use the triangle inequality: Because , the right‑hand side goes to as , so is Cauchy.
4. Since is complete, the sequence converges to some .
5. Contractions are continuous, so take the limit on both sides of to get .
6. Uniqueness: if and were both fixed, then . Because , this forces , so .
7. The error bound follows by letting in the inequality from step 3. ∎

Example: Solving . Take with the usual metric, and . As we saw, maps into itself and is a contraction with . The theorem guarantees there is one and only one solution in , and starting from any guess, say , the iteration will converge to the fixed point (approximately ). The error bound says after steps the distance to the true answer is at most , which is a safe overestimate; in practice convergence is faster.

This theorem is very handy: if we can rewrite a problem as finding a fixed point of a contraction on a complete space, we automatically get a unique solution and a simple method to find it.

📝 Section Recap: The Banach fixed point theorem says that a contraction mapping on a non‑empty complete metric space has a unique fixed point, and the simple iterative sequence converges to it from any starting point. The proof gives an explicit error bound.

Why This Matters for Differential Equations#

The contraction mapping principle is more than just theory—it drives the proof that many ordinary differential equations (ODEs) have unique solutions. Consider an initial value problem

If is continuous and satisfies a Lipschitz condition in its second argument, i.e.

for all near and all , then we can turn the ODE into an integral equation:

Now define an operator that acts on continuous functions by

A solution of the ODE is exactly a fixed point of : a function with . Pick a short time interval and look at the space of continuous functions on that interval, using the distance

This is a complete metric space. Because is Lipschitz, one can show that if we choose small enough so that , then is a contraction. The Banach fixed point theorem then guarantees a unique fixed point—a unique solution of the initial value problem on that interval. Starting from a constant function and iterating is called Picard iteration. So the abstract ideas of completeness and contractions directly give us a guarantee that many ODEs have unique solutions and that simple iteration works. Later, we will use the same fixed‑point logic on computer approximations to make sure they are reliable.

📝 Section Recap: When an ODE satisfies a Lipschitz condition, we can turn it into a fixed‑point problem on a complete function space. The contraction mapping principle then guarantees a unique solution, and Picard iteration gives a way to find it. This is the basis for many numerical methods.

Summary#

You now understand the main ideas that answer “Does a solution exist, and is it unique?” for many problems. We started by extending the idea of distance so we can talk about convergence in all sorts of settings. Then we made sure sequences that “should” converge actually do, by requiring completeness. A contraction gave us a clear way to guarantee that repeating a function pulls points together. Finally, the Banach fixed point theorem brought everything together, giving us a proof that a unique solution exists and a method to find it, plus an error estimate. This theorem will show up again and again as the foundation for solving differential equations numerically.