Contraction mappings sit at the point where [metric spaces](/page/Metric%20Space), [continuity](/page/Continuity), and iteration meet. A map may be continuous without making distances more predictable, and it may send a space into itself without moving points toward any stable location. A contraction mapping imposes a uniform shrinking rule: every pair of points becomes closer by a fixed factor strictly below $1$. That single inequality powers fixed point theorems, existence and uniqueness arguments for [ordinary differential equations](/page/Ordinary%20Differential%20Equation), iterative numerical schemes, and parts of [Banach space](/page/Banach%20Space) theory. The guiding question is not just whether repeated application of a map converges. It is whether convergence is forced by geometry rather than by a special formula. If $f:X \to X$ is a contraction, then the orbit \begin{align*} x_0,\ f(x_0),\ f(f(x_0)),\ \ldots \end{align*} shrinks its successive errors at a geometric rate. The [Banach fixed point theorem](/theorems/270), often called the [contraction mapping principle](/page/Contraction%20Mapping%20Principle), turns this geometric control into an existence and uniqueness result once the ambient metric space is complete. ## Definition A contraction is a strengthening of Lipschitz continuity. A Lipschitz map controls how much distances can expand; a contraction requires that all distances shrink by the same factor strictly less than $1$. The strict inequality is the feature that makes iteration effective. [definition: Contraction Mapping] Let $(X,d)$ be a metric space. A function \begin{align*} f:X &\to X \end{align*} is a contraction mapping on $X$ if there exists a constant $c \in [0,1)$ such that \begin{align*} d(f(x),f(y)) \le c\,d(x,y) \end{align*} for all $x,y \in X$. [/definition] The definition allows many possible constants, and in applications the choice of constant affects quantitative convergence estimates. To compare two contraction arguments, or to decide whether an estimate is strong enough for a fixed point theorem, it is useful to record the best global distance-expansion bound attached to the map. [definition: Lipschitz Constant of a Self-Map] Let $(X,d)$ be a metric space and let $f:X \to X$ be a function. The Lipschitz constant of $f$ is \begin{align*} \operatorname{Lip}(f) := \inf\{L \ge 0 : d(f(x),f(y)) \le L\,d(x,y) \text{ for all } x,y \in X\}. \end{align*} [/definition] With this notation, a self-map is a contraction precisely when its Lipschitz constant is strictly less than $1$. Once a map has this shrinking property, the next natural operation is to apply it repeatedly and watch whether the resulting sequence settles. [definition: Iterates of a Self-Map] Let $X$ be a set and let $f:X \to X$ be a function. Define $f^{0}:X \to X$ by $f^0=\operatorname{id}_X$, and for each $n \in \mathbb{N}$ define \begin{align*} f^n := f \circ f^{n-1}. \end{align*} [/definition] Here the superscript denotes repeated composition, not exponentiation of function values. Iteration raises the central question of what it would mean for the process to stop changing: a limiting endpoint should be a point left unchanged by one more application of the map. [definition: Fixed Point] Let $X$ be a set and let $f:X \to X$ be a function. A point $x^* \in X$ is a fixed point of $f$ if \begin{align*} f(x^*)=x^*. \end{align*} [/definition] The notation $x^*$ is common in fixed point theory and dynamical systems. It is not a dual-space symbol here; it labels a distinguished point stabilized by the map. ## Equivalent Characterisations The contraction condition can be read as a geometric statement about diameters. Instead of testing pairs point by point, one can ask how the map changes the size of every subset. [quotetheorem:8339] The diameter formulation is useful when an argument tracks whole sets rather than individual pairs of points. For example, it says at once that the image of a ball, interval, or candidate set of iterates has controlled size. It is not a new fixed point principle, and it does not by itself give convergence; it is a repackaging of the same pairwise contraction estimate in language suited to set estimates. The quantification over every nonempty subset is what makes the formulation equivalent to the original definition rather than merely a one-directional consequence. Limit arguments for iterates require more than diameter language; they require the map to respect convergence. Without continuity, even a convergent sequence of iterates might have an image sequence whose limit cannot be identified by applying the map to the limit. A global distance estimate removes this obstruction by forcing convergence to be preserved. [quotetheorem:8340] [Uniform continuity](/page/Uniform%20Continuity) is the first limit-theoretic payoff of the contraction inequality. The point is not merely that $f$ is continuous at each input, but that one distance threshold works uniformly across the whole metric space. This is what later permits an iteration limit to be tested by passing $f$ through the limit: if $x_n \to x$, then $f(x_n) \to f(x)$ without choosing new local estimates around each $x_n$. The theorem is also deliberately limited. Uniform continuity does not by itself say that distances shrink, that a fixed point exists, or that iterates converge at a geometric rate. Those stronger conclusions come from the strict constant $c<1$ together with completeness hypotheses. Here the result records the baseline regularity forced by any Lipschitz estimate before the page turns to settings where such estimates are obtained in practice. Many applications do not begin with an abstract metric formula; they begin with estimates in a normed space. A normed-space formulation lets the contraction condition speak directly to differences of vectors, functions, or sequences. This motivates recording the same concept in norm notation before discussing derivative-based tests. [definition: Contraction on a Normed Space] Let $(V,\|\cdot\|_V)$ be a [normed vector space](/page/Normed%20Vector%20Space). A function $f:V \to V$ is a contraction on $V$ if there exists $c \in [0,1)$ such that \begin{align*} \|f(x)-f(y)\|_V \le c\,\|x-y\|_V \end{align*} for all $x,y \in V$. [/definition] Differentiable maps on Euclidean domains are often easier to estimate through their derivatives than by comparing all pairs of points directly. The next criterion turns a uniform derivative bound into a contraction estimate when the domain contains the relevant line segments. [quotetheorem:8341] The convexity hypothesis lets line segments stay inside $U$, so the derivative bound can be integrated along the segment from $x$ to $y$. Without some path control, a derivative estimate on disconnected or badly shaped domains need not compare arbitrary pairs as directly. ## Standard Examples Linear maps give the cleanest test cases. They show that contraction is not about being small at one point; it is about the map shrinking all differences uniformly. [example: Linear Scaling on the Real Line] Let $X=\mathbb{R}$ with the usual metric $d(x,y)=|x-y|$, and define $f:\mathbb{R}\to\mathbb{R}$ by $f(x)=ax+b$, where $a,b\in\mathbb{R}$. For any $x,y\in\mathbb{R}$, \begin{align*} f(x)-f(y)=(ax+b)-(ay+b)=a(x-y). \end{align*} Taking absolute values gives \begin{align*} |f(x)-f(y)|=|a(x-y)|=|a|\,|x-y|. \end{align*} Therefore $f$ satisfies the contraction inequality with constant $c=|a|$ whenever $|a|<1$. Conversely, suppose $f$ is a contraction with some constant $c<1$. If $x\ne y$, then $|x-y|>0$, so the identity above gives \begin{align*} |a|=\frac{|f(x)-f(y)|}{|x-y|}\le c<1. \end{align*} Thus $f$ is a contraction exactly when $|a|<1$. When $|a|<1$, in particular $a\ne 1$, so the fixed point equation can be solved without division by zero: \begin{align*} x=ax+b. \end{align*} Subtracting $ax$ from both sides gives \begin{align*} (1-a)x=b. \end{align*} Dividing by $1-a$ gives \begin{align*} x^*=\frac{b}{1-a}. \end{align*} The coefficient $a$ controls the uniform shrinking factor, while the constant term $b$ changes the location of the fixed point. [/example] This example captures the main pattern: the contraction estimate controls differences, while fixed point location depends on the full map. The slope decides convergence speed; the intercept decides where the iteration settles. Not every map that moves points inward is a contraction. The next example separates bounded image from uniform distance shrinking. [example: Squaring on the Unit Interval] Let $X=[0,1]$ with the usual metric, and define $f:X \to X$ by \begin{align*} f(x)=x^2. \end{align*} For $x,y \in [0,1]$, the difference of squares factors as \begin{align*} f(x)-f(y)=x^2-y^2=(x-y)(x+y). \end{align*} Taking absolute values gives \begin{align*} |f(x)-f(y)|=|(x-y)(x+y)|=|x-y|\,|x+y|. \end{align*} Since $0 \le x \le 1$ and $0 \le y \le 1$, we have $0 \le x+y \le 2$, hence $|x+y|=x+y \le 2$. Therefore \begin{align*} |f(x)-f(y)| \le 2|x-y|. \end{align*} Thus $f$ is Lipschitz on $[0,1]$ with Lipschitz constant at most $2$. We now show that no contraction constant $c<1$ can work on all of $[0,1]$. Take $x=1$ and choose any $y \in [0,1)$, so $|1-y|>0$. Then \begin{align*} \frac{|f(1)-f(y)|}{|1-y|}=\frac{|1-y^2|}{|1-y|}. \end{align*} Since $0 \le y<1$, both $1-y$ and $1+y$ are nonnegative, and \begin{align*} 1-y^2=(1-y)(1+y). \end{align*} Hence \begin{align*} \frac{|1-y^2|}{|1-y|}=\frac{(1-y)(1+y)}{1-y}=1+y. \end{align*} If a contraction constant $c<1$ existed, then for every $y \in [0,1)$ we would have $1+y \le c$. Taking $y=0$ already gives $1 \le c$, contradicting $c<1$. Therefore $f$ is not a contraction on $[0,1]$. The fixed point equation is \begin{align*} x^2=x. \end{align*} Subtracting $x$ from both sides gives \begin{align*} x^2-x=0. \end{align*} Factoring gives \begin{align*} x(x-1)=0. \end{align*} Thus the fixed points in $[0,1]$ are $0$ and $1$. This example shows that a Lipschitz self-map, even one with bounded image in a compact interval, can fail to be a contraction and can have more than one fixed point. [/example] The failure at the endpoint explains why local behavior matters. On a smaller interval $[0,a]$ with $0<a<1/2$, the same map is a contraction because $x+y \le 2a<1$. The strictness of the constant also matters. Maps with Lipschitz constant $1$ can behave like contractions in some examples but do not satisfy the fixed point theorem in general. [example: Translation as a Nonexpansive Map] Let $X=\mathbb{R}$ with the usual metric $d(x,y)=|x-y|$, and define $f:\mathbb{R}\to\mathbb{R}$ by \begin{align*} f(x)=x+1. \end{align*} For any $x,y\in\mathbb{R}$, \begin{align*} f(x)-f(y)=(x+1)-(y+1). \end{align*} Cancelling the constant terms gives \begin{align*} f(x)-f(y)=x-y. \end{align*} Taking absolute values gives \begin{align*} |f(x)-f(y)|=|x-y|. \end{align*} Thus $f$ satisfies the nonexpansive estimate \begin{align*} d(f(x),f(y))\le 1\cdot d(x,y) \end{align*} for all $x,y\in\mathbb{R}$. No contraction constant $c<1$ can work. Taking $x=1$ and $y=0$ gives \begin{align*} d(f(1),f(0))=|(1+1)-(0+1)|. \end{align*} The right-hand side is \begin{align*} |(1+1)-(0+1)|=|2-1|=1. \end{align*} Also, \begin{align*} d(1,0)=|1-0|=1. \end{align*} If $f$ were a contraction with constant $c<1$, then the contraction inequality applied to $1$ and $0$ would give \begin{align*} 1=d(f(1),f(0))\le c\,d(1,0)=c. \end{align*} This contradicts $c<1$, so $f$ is not a contraction. Finally, a fixed point would have to satisfy \begin{align*} f(x)=x. \end{align*} Substituting the formula for $f$ gives \begin{align*} x+1=x. \end{align*} Subtracting $x$ from both sides gives \begin{align*} 1=0, \end{align*} which is impossible in $\mathbb{R}$. Hence this nonexpansive self-map has no fixed point, showing why the strict inequality in the contraction constant is essential. [/example] This example is the basic warning behind the condition $c<1$. Preserving distances is too weak for fixed point existence, even in a complete space. Completeness is a separate hypothesis in the fixed point theorem. A contraction may behave perfectly inside a larger complete space but fail to land on a fixed point in an incomplete subspace. [example: Contraction on an Incomplete Space Without a Fixed Point] Let $X=(0,1)$ with the usual metric $d(x,y)=|x-y|$, and define $f:X\to X$ by \begin{align*} f(x)=\frac{x}{2}. \end{align*} If $x\in(0,1)$, then $0<x<1$, so dividing by $2$ gives \begin{align*} 0<\frac{x}{2}<\frac{1}{2}<1. \end{align*} Thus $f(x)\in(0,1)$, so $f$ is indeed a self-map of $X$. For any $x,y\in X$, \begin{align*} f(x)-f(y)=\frac{x}{2}-\frac{y}{2}. \end{align*} Factoring out $\frac{1}{2}$ gives \begin{align*} f(x)-f(y)=\frac{1}{2}(x-y). \end{align*} Taking absolute values and using $|\lambda z|=|\lambda|\,|z|$ for [real numbers](/page/Real%20Numbers) gives \begin{align*} |f(x)-f(y)|=\left|\frac{1}{2}(x-y)\right|=\frac{1}{2}|x-y|. \end{align*} Therefore \begin{align*} d(f(x),f(y))=\frac{1}{2}d(x,y). \end{align*} Since $\frac{1}{2}\in[0,1)$, the map $f$ is a contraction on $X$ with contraction constant $c=\frac{1}{2}$. Now suppose $x\in X$ were a fixed point. Then $f(x)=x$, so \begin{align*} \frac{x}{2}=x. \end{align*} Subtracting $\frac{x}{2}$ from both sides gives \begin{align*} 0=x-\frac{x}{2}. \end{align*} Combining the terms on the right gives \begin{align*} 0=\frac{x}{2}. \end{align*} Multiplying by $2$ gives \begin{align*} x=0. \end{align*} But $0\notin(0,1)$, so no point of $X$ satisfies the fixed point equation. This contraction has no fixed point because the point toward which the iteration is driven lies outside the incomplete space $X$. [/example] The example shows why the [Banach fixed point theorem](/theorems/289) is a theorem about complete metric spaces, not merely about contractions. Before existence can be discussed, however, the contraction inequality already rules out the possibility of two distinct fixed points. ## Properties The central property of a contraction is uniqueness of fixed points. This part does not require completeness; it follows directly from the distance-shrinking inequality. [quotetheorem:8342] A fixed point theorem needs both uniqueness and existence. The contraction inequality gives uniqueness, but existence requires the iterative sequence to converge inside the space. The obstruction is that the distances between successive iterates may shrink while the limiting point still lies outside the metric space; completeness is precisely the hypothesis that prevents this failure. [quotetheorem:71] The fixed point theorem gives an algorithm as well as an existence result. To use the algorithm responsibly, the next estimate measures how quickly the iterates approach the fixed point. [quotetheorem:8343] This estimate is called a priori because it involves the unknown fixed point. Numerical methods also need bounds that can be evaluated from the computed iterates themselves. Comparing successive iterates gives a practical stopping rule and measures the remaining error without already knowing $x^*$. [quotetheorem:8344] This estimate is a posteriori because its right-hand side is computable after the iterates have been produced. In practice it turns a small step size $d(x_n,x_{n-1})$ into a certified bound on the remaining distance to the fixed point, so it can justify a stopping rule. Its limitation is just as important: the certificate is valid only after the contraction constant and the complete invariant space have already been established. A small numerical step alone is not evidence of correctness without those hypotheses. Some constructions build a complicated update map by composing simpler maps, each with its own Lipschitz estimate. To prove that the final composite is still contractive, one must control how much shrinking or expansion accumulates at each stage. This reduces the problem to tracking the product of the individual constants. [quotetheorem:8309] This property lets contraction estimates be assembled from smaller steps. It is especially useful when an operator is built from an integral transform, a nonlinear substitution, and a projection or restriction. ## Relationship to Other Concepts Contraction mappings refine [Lipschitz continuity](/page/Lipschitz%20Function). The difference is not cosmetic: Lipschitz constants greater than $1$ permit expansion, constants equal to $1$ permit isometries and translations, while constants below $1$ force repeated distances to decay geometrically. They also connect directly to [complete metric spaces](/page/Complete%20Metric%20Space). Completeness is the condition that prevents the limiting point of an iteration from falling outside the space. Many applications first choose a function space and a closed subset of it precisely so that the contraction mapping principle applies. In differential equations, a common strategy rewrites an initial value problem as a fixed point problem. Let $I=[t_0,t_0+h]$, let $u_0 \in \mathbb{R}^m$, and let $F:I \times \mathbb{R}^m \to \mathbb{R}^m$ be continuous. Write $C(I;\mathbb{R}^m)$ for the space of continuous functions from $I$ to $\mathbb{R}^m$, equipped for this discussion with the supremum metric. A solution of \begin{align*} u'(t)=F(t,u(t)), \qquad u(t_0)=u_0 \end{align*} can be sought as a fixed point of the integral operator $T:C(I;\mathbb{R}^m) \to C(I;\mathbb{R}^m)$ defined by \begin{align*} (Tu)(t)=u_0+\int_{t_0}^{t} F(s,u(s))\,ds. \end{align*} On a sufficiently short time interval, a Lipschitz condition on $F$ in the $u$ variable can make $T$ a contraction on a complete function space. [remark: Local Time Intervals in ODE Applications] The contraction constant for the integral operator usually contains the length of the time interval. Shrinking the interval can turn a merely Lipschitz nonlinear term into a contraction estimate for the fixed point operator. [/remark] Contraction methods also appear in [numerical analysis](/page/Numerical%20Analysis). Fixed point iteration, Picard iteration, and certain implicit schemes are justified by showing that the update map is a contraction in a carefully chosen metric. The concept is related but not identical to stability in dynamical systems. A map with an attracting fixed point may draw nearby orbits inward without being a contraction on the whole space. Contraction is a global, uniform metric condition, while attraction can be local and asymptotic. [remark: Local Versus Global Contraction] A function can fail to be a contraction on all of $X$ but become a contraction after restricting to a smaller subset. This distinction is central in local existence theorems, where the space and radius are chosen so the relevant operator maps a complete closed subset into itself. [/remark] ## References [Metric Space](/page/Metric%20Space). [Continuity](/page/Continuity). [Complete Metric Space](/page/Complete%20Metric%20Space). [Banach Space](/page/Banach%20Space). Walter Rudin, *Principles of Mathematical Analysis* (1976). Erwin Kreyszig, *Introductory Functional Analysis with Applications* (1978). Lawrence C. Evans, *Partial Differential Equations* (2010).