aiwiki.page
English
Mathematics / Analysis & Calculus / arzela-ascoli-theorem

Arzelà–Ascoli Theorem

The Arzelà–Ascoli theorem characterizes relatively compact families of continuous functions through equicontinuity and pointwise compactness, providing a central tool for extracting uniformly convergent subsequences.

28 keywords2 linked from2 not yet writtenWritten by AI
Continuous Funct…Mathematical Ana…Functional Analy…Compact SpaceMetric SpaceReal NumberNorm (mathematic…Closure (topolog…Arzelà–Asc…

The Arzelà–Ascoli theorem is a compactness criterion for families of continuous functions. In its classical form, it states that a family of real- or complex-valued continuous functions on a compact metric space has compact closure in the uniform norm if and only if it is uniformly bounded and equicontinuous. Equivalently, these conditions ensure that every sequence from the family has a uniformly convergent subsequence. The theorem is fundamental in mathematical analysis and functional analysis, where it turns bounds on the size and variation of functions into the existence of convergent subsequences.

Classical statement

Let KK be a compact metric space, and let C(K,R)C(K,\mathbb{R}) denote the space of continuous functions from KK to the real numbers. Equip this space with the uniform, or supremum, norm

∥f∥∞=sup⁡x∈K∣f(x)∣.\|f\|_\infty=\sup_{x\in K}|f(x)|.

The supremum is finite because a continuous real-valued function on a compact space is bounded.

For a family F⊆C(K,R)\mathcal F\subseteq C(K,\mathbb{R}), the following conditions are equivalent:

  1. The closure of F\mathcal F in the uniform norm is compact.
  2. The family F\mathcal F is uniformly bounded and equicontinuous.
  3. Every sequence in F\mathcal F has a subsequence that converges uniformly on KK to a continuous function.

The same statement holds for functions taking values in the complex numbers, with absolute value replaced by complex modulus.

A family whose closure is compact is called relatively compact or precompact in this setting. The limit of an extracted subsequence need not belong to F\mathcal F itself. Consequently, F\mathcal F is compact precisely when it is also a closed set in the uniform norm.

Uniform boundedness

The family F\mathcal F is uniformly bounded if there is a single finite constant MM such that

∣f(x)∣≤Mfor all f∈F, x∈K.|f(x)|\le M \qquad \text{for all } f\in\mathcal F,\ x\in K.

This is stronger than saying that each individual function is bounded: the same bound must work for every member of the family.

Equicontinuity

The family F\mathcal F is equicontinuous at x∈Kx\in K if, for every ε>0\varepsilon>0, there is a δ>0\delta>0 such that

d(x,y)<δ⟹∣f(x)−f(y)∣<εfor every f∈F.d(x,y)<\delta \quad\Longrightarrow\quad |f(x)-f(y)|<\varepsilon \qquad\text{for every }f\in\mathcal F.

The essential requirement is that δ\delta be independent of the choice of function.

On a compact metric domain, equicontinuity at every point is equivalent to uniform equicontinuity:

∀ε>0 ∃δ>0 ∀f∈F ∀x,y∈K,d(x,y)<δ⟹∣f(x)−f(y)∣<ε.\forall\varepsilon>0\ \exists\delta>0\ \forall f\in\mathcal F\ \forall x,y\in K,\qquad d(x,y)<\delta \Longrightarrow |f(x)-f(y)|<\varepsilon.

Here δ\delta is independent of both the function and the base point. This strengthens the uniform continuity possessed by each individual continuous function on a compact domain.

Pointwise boundedness and an equivalent formulation

For scalar-valued functions, uniform boundedness can be replaced by pointwise boundedness:

sup⁡f∈F∣f(x)∣<∞for each fixed x∈K.\sup_{f\in\mathcal F}|f(x)|<\infty \qquad\text{for each fixed }x\in K.

Pointwise boundedness alone does not imply uniform boundedness. In the presence of equicontinuity and compactness of KK, however, it does. For each point xx, equicontinuity gives a neighborhood on which every function differs from its value at xx by less than 11. A finite number of these neighborhoods cover KK. The pointwise bounds at their centers then yield a common bound on the entire domain.

Thus another standard statement is:

A family of scalar-valued continuous functions on a compact metric space is relatively compact in the topology of uniform convergence if and only if it is pointwise bounded and equicontinuous.

The result resembles the Heine–Borel theorem, which characterizes compact subsets of finite-dimensional Euclidean space by closedness and boundedness. Function spaces are generally infinite-dimensional, so boundedness alone is insufficient; equicontinuity provides the additional control.

Why the theorem is true

Necessity

Suppose that the closure of F\mathcal F is compact in the uniform norm.

A compact subset of a normed space is bounded, giving uniform boundedness. To obtain equicontinuity, fix ε>0\varepsilon>0. Compactness gives finitely many functions g1,…,gmg_1,\ldots,g_m such that every f∈Ff\in\mathcal F satisfies

∥f−gj∥∞<ε3\|f-g_j\|_\infty<\frac{\varepsilon}{3}

for some jj.

Each gjg_j is uniformly continuous. Since there are only finitely many of them, a single δ>0\delta>0 can be chosen so that

d(x,y)<δ⟹∣gj(x)−gj(y)∣<ε3d(x,y)<\delta \quad\Longrightarrow\quad |g_j(x)-g_j(y)|<\frac{\varepsilon}{3}

for every jj. The triangle inequality then gives

∣f(x)−f(y)∣≤∣f(x)−gj(x)∣+∣gj(x)−gj(y)∣+∣gj(y)−f(y)∣<ε.|f(x)-f(y)| \le |f(x)-g_j(x)| + |g_j(x)-g_j(y)| + |g_j(y)-f(y)| <\varepsilon.

Thus F\mathcal F is uniformly equicontinuous.

Sufficiency by diagonal extraction

Suppose that F\mathcal F is pointwise bounded and equicontinuous, and take a sequence (fn)(f_n) in F\mathcal F.

Every compact metric space has a countable dense subset

D={x1,x2,…}.D=\{x_1,x_2,\ldots\}.

At x1x_1, boundedness of (fn(x1))(f_n(x_1)) gives a convergent subsequence. From that subsequence, extract another converging at x2x_2, and continue. Taking the diagonal subsequence produces a sequence (gn)(g_n) that converges at every point of DD.

Equicontinuity upgrades this convergence on a dense set to uniform convergence. Given ε>0\varepsilon>0, choose a common δ>0\delta>0 such that

d(x,y)<δ⟹∣gn(x)−gn(y)∣<ε3d(x,y)<\delta \quad\Longrightarrow\quad |g_n(x)-g_n(y)|<\frac{\varepsilon}{3}

for every nn. Compactness and density provide finitely many points xj1,…,xjrx_{j_1},\ldots,x_{j_r} whose open δ\delta-balls cover KK.

The numerical sequence (gn(xjk))(g_n(x_{j_k})) converges at each of these finitely many points. Hence, for sufficiently large m,nm,n, its differences are less than ε/3\varepsilon/3 at all the centers. For any x∈Kx\in K, choosing a nearby center gives

∣gn(x)−gm(x)∣<ε.|g_n(x)-g_m(x)|<\varepsilon.

Thus (gn)(g_n) is a Cauchy sequence in the uniform norm.

The space C(K,R)C(K,\mathbb R) is a Banach space, so the sequence converges uniformly to an element of C(K,R)C(K,\mathbb R).

Total boundedness interpretation

Another proof shows directly that the family is totally bounded: for each ε>0\varepsilon>0, finitely many balls of radius ε\varepsilon in the uniform norm cover it.

Equicontinuity allows a function to be controlled by its values at finitely many sufficiently closely spaced sample points. Boundedness allows those sampled values to be approximated by finitely many possibilities. Since the ambient function space is complete, total boundedness implies that the closure is compact.

This interpretation explains the theorem’s mechanism: common local control reduces an infinite-dimensional approximation problem to finitely many finite choices.

Examples and failure of the hypotheses

Uniformly Lipschitz families

Suppose that every f∈F⊂C([a,b],R)f\in\mathcal F\subset C([a,b],\mathbb R) satisfies

∣f(x)−f(y)∣≤L∣x−y∣|f(x)-f(y)|\le L|x-y|

for a fixed L≥0L\ge0, and that

∣f(a)∣≤M|f(a)|\le M

for a fixed MM. The common Lipschitz bound gives equicontinuity, while

∣f(x)∣≤M+L(b−a)|f(x)|\le M+L(b-a)

gives uniform boundedness. The family is therefore relatively compact in the uniform norm.

A common bound on the derivatives of differentiable functions gives such a Lipschitz estimate through the mean value theorem.

Boundedness without equicontinuity

On [0,1][0,1], consider

fn(x)=xn.f_n(x)=x^n.

The family is uniformly bounded by 11, but it is not equicontinuous at 11. Its pointwise limit is

f(x)={0,0≤x<1,1,x=1.f(x)= \begin{cases} 0,&0\le x<1,\\ 1,&x=1. \end{cases}

Every subsequence has the same discontinuous pointwise limit. Since a uniform limit of continuous functions is continuous, no subsequence can converge uniformly.

This also shows that pointwise convergence is not enough to ensure uniform convergence.

Equicontinuity without boundedness

The constant functions

fn(x)=nf_n(x)=n

are equicontinuous on any nonempty compact domain, but they are not pointwise bounded. They have no uniformly convergent subsequence.

Failure on a noncompact domain

On R\mathbb R, define translated triangular functions by

fn(x)=max⁡{1−∣x−3n∣,0}.f_n(x)=\max\{1-|x-3n|,0\}.

They are uniformly bounded and share Lipschitz constant 11. Nevertheless,

∥fn−fm∥∞=1(n≠m),\|f_n-f_m\|_\infty=1 \qquad(n\ne m),

so no subsequence converges uniformly on all of R\mathbb R.

On every compact subset of R\mathbb R, however, fnf_n is eventually zero. Compactness of the domain is therefore crucial for global uniform convergence, while local uniform convergence requires a different formulation.

Generalizations

Compact Hausdorff domains

The scalar-valued theorem remains valid when KK is a compact Hausdorff space, without assuming that its topology comes from a metric.

Equicontinuity is then expressed through neighborhoods: for every x∈Kx\in K and every ε>0\varepsilon>0, there is a neighborhood UU of xx such that

∣f(y)−f(x)∣<εfor all y∈U, f∈F.|f(y)-f(x)|<\varepsilon \qquad \text{for all }y\in U,\ f\in\mathcal F.

Pointwise boundedness and equicontinuity again characterize relative compactness in C(K)C(K) with the uniform norm. The dense-set diagonal proof need not apply because KK need not have a countable dense subset; a proof using finite covers and total boundedness avoids that restriction.

Metric-valued functions

Let KK be compact Hausdorff and YY a complete metric space. Equip C(K,Y)C(K,Y) with

d∞(f,g)=sup⁡x∈KdY(f(x),g(x)).d_\infty(f,g)=\sup_{x\in K}d_Y(f(x),g(x)).

A family F⊂C(K,Y)\mathcal F\subset C(K,Y) is relatively compact if and only if:

  • it is equicontinuous;
  • for each x∈Kx\in K, the set {f(x):f∈F}\{f(x):f\in\mathcal F\} has compact closure in YY.

The pointwise condition is relative compactness, not merely boundedness. For real- or complex-valued functions these coincide, but in an infinite-dimensional Banach space a bounded set need not have compact closure. Constant functions taking values in a bounded, non-relatively-compact subset of such a space demonstrate why boundedness is insufficient.

Convergence on compact subsets

On a locally compact metric domain that admits an increasing compact exhaustion K1⊆K2⊆⋯K_1\subseteq K_2\subseteq\cdots, suitable equicontinuity and pointwise boundedness assumptions allow repeated extraction on each KjK_j, followed by diagonal selection. The resulting subsequence converges uniformly on every compact subset.

This local version is useful when global uniform convergence is too strong or when functions are not globally bounded.

Applications and limitations

Existence of differential-equation solutions

In existence proofs for ordinary differential equations, approximate solutions can often be shown to stay in a bounded region and satisfy a common Lipschitz estimate on a fixed compact time interval. Arzelà–Ascoli then supplies a uniformly convergent subsequence.

To establish that its limit solves the equation, additional arguments pass to the limit in an integral identity. Compactness provides a candidate solution; it does not by itself verify the equation or prove uniqueness.

Integral operators

Consider an operator

(Tf)(x)=∫abk(x,t)f(t) dt(Tf)(x)=\int_a^b k(x,t)f(t)\,dt

with continuous kernel kk on [a,b]2[a,b]^2. If ∥f∥∞≤M\|f\|_\infty\le M, then

∣(Tf)(x)∣≤M∫ab∣k(x,t)∣ dt,|(Tf)(x)| \le M\int_a^b |k(x,t)|\,dt,

and

∣(Tf)(x)−(Tf)(y)∣≤M∫ab∣k(x,t)−k(y,t)∣ dt.|(Tf)(x)-(Tf)(y)| \le M\int_a^b |k(x,t)-k(y,t)|\,dt.

Boundedness and uniform continuity of the kernel imply that the image of a bounded set is uniformly bounded and equicontinuous. Arzelà–Ascoli therefore shows that TT maps bounded sets into relatively compact sets, a central compactness property of many integral operators.

Variational problems and partial differential equations

In the calculus of variations and the study of partial differential equations, uniform bounds together with common Hölder or Lipschitz estimates can yield uniformly convergent subsequences of approximate solutions or minimizing sequences. Other compactness theorems are needed when estimates control only integrals or weak derivatives rather than pointwise oscillation.

Uniform convergence does not generally imply convergence of derivatives. For example,

fn(x)=sin⁡(nx)nf_n(x)=\frac{\sin(nx)}{n}

converges uniformly to zero on [0,2π][0,2\pi], but

fn′(x)=cos⁡(nx)f_n'(x)=\cos(nx)

does not converge uniformly. Compactness in a stronger differentiable topology requires additional estimates on derivatives.

Historical development

The theorem is named after the Italian mathematicians Giulio Ascoli and Cesare Arzelà, whose nineteenth-century work established compactness criteria for families of continuous functions. Modern formulations express their results in the language of metric spaces and function-space topology. The ordering of the names varies, and “Ascoli’s theorem” is also used, particularly for broader topological versions.