Banach Fixed-Point Theorem

The Banach fixed-point theorem states that every contraction mapping from a nonempty complete metric space into itself has exactly one fixed point. Repeatedly applying the mapping, starting from any point in the space, produces a sequence converging to that fixed point. The theorem therefore combines existence, uniqueness, and a method of approximation. It is a fundamental tool in mathematical analysis, particularly in proofs concerning equations and iterative procedures. (math.ucla.edu)

Statement and hypotheses

Let $(X,d)$ be a nonempty complete metric space, and let $T:X\to X$ satisfy

where $q$ is a fixed constant with $0\le q<1$. Then:

1. There exists a unique $x^*\in X$ such that $Tx^*=x^*$.

2. For every $x_0\in X$, the sequence defined by

   $$x_{n+1}=Tx_n$$

   converges to $x^*$. (math.ucla.edu)

The contraction condition is a form of Lipschitz continuity: distances are reduced by a uniform factor strictly less than one. Completeness means that every Cauchy sequence in $X$ converges to a point belonging to $X$. The requirement $T:X\to X$ ensures that every successive iterate remains in the space. (math.ucla.edu)

No vector-space structure, differentiability, or compactness is required. In particular, the theorem is not restricted to a Banach space, despite its name. A useful application setting is a nonempty closed subset of a complete metric space, equipped with the inherited metric, provided that the mapping preserves that subset. (faculty.etsu.edu)

Proof by successive approximation

Choose $x_0\in X$ and set $x_{n+1}=Tx_n$. Repeated use of the contraction inequality gives

For $m>n$, the triangle inequality and the formula for a geometric series imply

The right-hand side tends to zero, so $(x_n)$ is Cauchy. Completeness supplies a limit $x^*\in X$. (math.ucla.edu)

A contraction is a continuous mapping, hence

Thus the limit is a fixed point. If $y^*$ is another fixed point, then

Since $q<1$, this forces $d(x^*,y^*)=0$, establishing uniqueness. (cs.cornell.edu)

The proof yields an algorithm—fixed-point iteration, also called successive approximation—rather than merely asserting existence. The fixed point is obtained as a limit, not necessarily after finitely many iterations. (faculty.etsu.edu)

Error estimates and convergence rate

For $0<q<1$, the iteration satisfies the a priori estimate

This bounds the error using the contraction constant, the first step, and the iteration count. If $q=0$, the mapping is constant and its value is the fixed point, reached after one application. (faculty.etsu.edu)

An a posteriori estimate, using quantities available during iteration, is

More generally, for any $x\in X$,

The latter follows by applying the triangle inequality to $x,Tx,x^*$ and rearranging. It relates the fixed-point residual $d(x,Tx)$ to the actual error. These bounds provide mathematical stopping criteria for approximations. (math.umd.edu)

The direct estimate

shows geometric convergence, commonly termed linear convergence in numerical analysis. This is a guaranteed upper bound, not necessarily the exact asymptotic rate: a particular iteration may converge faster. When $q$ is close to one, the guarantee can be slow and the residual-based error bound correspondingly large. (cs.cornell.edu)

An elementary example

As an illustration, consider

on the real numbers with their usual distance. Then

so $T$ is a contraction with $q=1/3$. Its fixed-point equation gives $x^*=1$, and direct calculation yields

Thus every starting value converges to the same fixed point.

This example illustrates the theorem’s uniform convergence guarantee across starting points; it does not depend on choosing an initial value near the solution. (math.ucla.edu)

Applications to equations

Differential equations

A central application is the Picard–Lindelöf theorem for local existence and uniqueness of solutions to a differential equation. The initial-value problem

can be rewritten as

The right-hand side defines an operator $T$ on a suitable space of continuous functions. (faculty.etsu.edu)

If $f$ is continuous and Lipschitz in its second variable with constant $L$ on the relevant region, then, on an interval satisfying $|t-t_0|\le h$,

Choosing $h$ so that $Lh<1$, and ensuring that $T$ preserves the chosen closed set of functions, makes the operator a contraction. Its fixed point is the desired solution, and the successive approximations are Picard iterates. Completeness of the function space under the supremum norm is essential to this argument. (math.northwestern.edu)

Nonlinear equations and numerical methods

An equation $F(x)=0$ can sometimes be reformulated as $x=T(x)$. If $T$ is contractive on a complete invariant domain, the theorem establishes a unique solution within that domain and convergence of the corresponding iteration. Different reformulations of the same equation need not have the same convergence properties. (cs.cornell.edu)

For a differentiable real-valued mapping on an interval, a uniform derivative bound

is a sufficient contraction criterion, by the mean value theorem. In higher dimensions, an analogous bound uses the operator norm of the Jacobian matrix on a convex domain. Such derivative criteria verify contractivity; they do not replace the separate requirement that the mapping preserve the domain. (cs.cornell.edu)

Scope and limitations

The hypotheses distinguish the theorem from more general fixed-point existence results.

• Completeness cannot simply be omitted. On $(0,\infty)$, the map $T(x)=x/2$ is a contraction but has no fixed point in its domain. Its iterates approach zero, which lies outside the space. (cs.cornell.edu)
• The uniform factor must be strictly less than one. Allowing $q=1$ permits the identity map, with every point fixed, or a translation on $\mathbb R$, with no fixed point.
• Contractivity is sufficient, not necessary. A mapping may have a unique fixed point without satisfying a global contraction estimate.
• The guarantee is domain-specific. A contraction on an invariant subset gives uniqueness there, not automatically throughout a larger ambient space. (faculty.etsu.edu)

The Brouwer fixed-point theorem, by comparison, guarantees a fixed point for a continuous self-map of a nonempty compact convex subset of finite-dimensional Euclidean space. It does not generally guarantee uniqueness or convergence of successive iteration. Banach’s theorem imposes the stronger contraction condition on the mapping, but requires neither compactness nor convexity of the complete metric space. (faculty.etsu.edu)

Historical origin

The theorem is named after Stefan Banach. His 1922 paper Sur les opérations dans les ensembles abstraits et leur application aux équations intégrales, published in Fundamenta Mathematicae, volume 3, pages 133–181, developed abstract results for function spaces and their application to integral equations. The contraction principle belongs to this foundational work in functional analysis. (impan.pl)

References

1. 3. The Banach Contraction Principlefaculty.etsu.edu
2. CS 4220: Numerical Analysiscs.cornell.edu
3. AMSC/CMSC 466: Fixed Point Iteration and Contraction Mapping Theoremmath.umd.edu
4. Math 320-2: Real Analysismath.northwestern.edu
5. Sur les opérations dans les ensembles abstraits et leur application aux équations intégralesimpan.pl