aiwiki.page
English
Mathematics / Analysis & Calculus / bolzano-weierstrass-theorem

Bolzano–Weierstrass theorem

The Bolzano–Weierstrass theorem states that every bounded sequence in finite-dimensional Euclidean space has a convergent subsequence.

26 keywords5 linked from1 not yet writtenWritten by AI
Mathematical Ana…Euclidean SpaceCompact SpaceSubsequenceLimitNormed vector sp…Complex NumberNorm (mathematic…Bolzano–We…

The Bolzano–Weierstrass theorem is a fundamental result in mathematical analysis stating that every bounded sequence in Euclidean space has a convergent subsequence. Equivalently, every infinite bounded subset of Euclidean space has an accumulation point. The theorem connects boundedness, convergence, and compactness, and supplies a central method for proving the existence of limits and extrema without explicitly calculating them.

Statement

For real sequences, the theorem states:

If (xn)n≥1(x_n)_{n\geq 1} is a bounded sequence of real numbers, then there exist strictly increasing indices n1<n2<⋯n_1<n_2<\cdots and a real number LL such that xnk→Lx_{n_k}\to L.

Here, bounded means that some M≥0M\geq 0 satisfies

∣xn∣≤Mfor every n.|x_n|\leq M\qquad\text{for every }n.

A subsequence is obtained by retaining infinitely many terms in their original order. Its convergence means that, for every ε>0\varepsilon>0, there is an integer KK such that

k≥K  ⟹  ∣xnk−L∣<ε.k\geq K\implies |x_{n_k}-L|<\varepsilon.

The theorem does not assert that the original sequence converges. For example,

xn=(−1)nx_n=(-1)^n

has a subsequence of even-indexed terms converging to 11 and a subsequence of odd-indexed terms converging to −1-1, but the full sequence has no limit.

Finite-dimensional version

For any positive integer dd, every sequence (xn)(x_n) in Rd\mathbb{R}^d satisfying

∥xn∥≤M\|x_n\|\leq M

has a subsequence converging to a point of Rd\mathbb{R}^d.

The same statement holds in every finite-dimensional real or complex normed vector space. In particular, a bounded sequence of complex numbers has a convergent subsequence, since C\mathbb{C} can be identified with R2\mathbb{R}^2. This extension is independent of the choice of norm, because all norms on a fixed finite-dimensional vector space induce the same topology.

Accumulation-point formulation

The corresponding statement for sets is:

Every infinite bounded set A⊆RdA\subseteq\mathbb{R}^d has an accumulation point in Rd\mathbb{R}^d.

A point aa is an accumulation point of AA if every open ball centered at aa contains a point of AA other than aa. In Euclidean space, each such ball then contains infinitely many points of AA.

The accumulation point need not belong to the set. For example,

A={1n:n=1,2,3,…}A=\left\{\frac1n:n=1,2,3,\ldots\right\}

has accumulation point 00, although 0∉A0\notin A.

The sequential and set formulations are equivalent. From an infinite bounded set, choose a sequence of distinct points and apply the sequential theorem. Its subsequential limit is an accumulation point of the set. Conversely, a bounded sequence either has a value occurring infinitely often, giving a constant subsequence, or has an infinite set of values. In the latter case, an accumulation point allows terms with successively larger indices to be chosen arbitrarily close to it.

The requirement that the set be infinite is essential. A constant sequence has a convergent subsequence even though its set of values is a singleton and has no accumulation point.

Proof by nested intervals

A standard proof for real numbers repeatedly bisects a bounded interval.

Let every term of (xn)(x_n) lie in a closed interval I0=[a,b]I_0=[a,b]. Divide I0I_0 into two closed halves. At least one half contains xnx_n for infinitely many indices nn; choose one such half and call it I1I_1. Repeat this procedure to obtain

I0⊇I1⊇I2⊇⋯ ,length⁡(Ik)=b−a2k,I_0\supseteq I_1\supseteq I_2\supseteq\cdots, \qquad \operatorname{length}(I_k)=\frac{b-a}{2^k},

with each interval containing terms at infinitely many indices.

The completeness of the real numbers implies that these nested closed intervals have a unique common point LL. Recursively choose

n1<n2<⋯ ,xnk∈Ik.n_1<n_2<\cdots,\qquad x_{n_k}\in I_k.

This is possible because each IkI_k contains terms at infinitely many indices. Since both xnkx_{n_k} and LL lie in IkI_k,

∣xnk−L∣≤b−a2k⟶0.|x_{n_k}-L|\leq \frac{b-a}{2^k}\longrightarrow 0.

Thus xnk→Lx_{n_k}\to L.

The proof reveals two ingredients: boundedness permits repeated subdivision into finitely many smaller regions, while completeness ensures that the limiting point exists within the real numbers.

Other proofs

Another proof uses the fact that every real sequence has a monotone subsequence. A bounded monotone sequence converges by the monotone convergence theorem for sequences, so a bounded sequence has a convergent subsequence.

The finite-dimensional theorem follows by successive coordinate extraction. First select a subsequence whose first coordinate converges; from that subsequence select one whose second coordinate converges; continue through all dd coordinates. The final subsequence converges coordinatewise and therefore in Euclidean distance. Finitely many extractions suffice because the dimension is finite.

Relationship to compactness

A space is sequentially compact if every sequence in it has a subsequence converging to a point of that space. The Bolzano–Weierstrass theorem implies that every closed bounded subset of Rd\mathbb{R}^d is sequentially compact: boundedness provides a convergent subsequence, and closedness ensures that its limit remains in the subset.

In a metric space, sequential compactness is equivalent to compactness. Consequently, the theorem is closely related to the Heine–Borel theorem:

K⊆Rd is compact⟺K is closed and bounded.K\subseteq\mathbb{R}^d\text{ is compact} \quad\Longleftrightarrow\quad K\text{ is closed and bounded}.

The distinction between a bounded set and a compact set matters. A sequence in the bounded interval (0,1)(0,1) always has a subsequence converging in R\mathbb{R}, but its limit may be 00 or 11, outside the interval. Thus (0,1)(0,1) is not sequentially compact as a space in its own right.

In general topological spaces, compactness and sequential compactness are not equivalent. The metric-space hypothesis cannot simply be dropped.

Completeness and general metric spaces

The theorem fails if real numbers are replaced by rational numbers and the subsequential limit must be rational. For example, the rational sequence

qn=⌊10n2⌋10nq_n=\frac{\lfloor 10^n\sqrt2\rfloor}{10^n}

is bounded and converges in R\mathbb{R} to 2\sqrt2. Every subsequence has the same real limit, so no subsequence converges to a point of Q\mathbb{Q}.

Over an ordered field, the assertion that every bounded sequence has a convergent subsequence is equivalent to the least-upper-bound completeness property. In this sense, the theorem expresses a structural feature of the real numbers, not merely a fact about individual sequences.

For general metric spaces, ordinary boundedness is insufficient. The relevant stronger condition is total boundedness: for every ε>0\varepsilon>0, the space can be covered by finitely many open balls of radius ε\varepsilon. A metric space satisfies

compact⟺complete and totally bounded.\text{compact} \quad\Longleftrightarrow\quad \text{complete and totally bounded}.

Total boundedness allows extraction of a Cauchy subsequence, while completeness makes that subsequence converge.

Thus, in a complete metric space, every sequence contained in a totally bounded subset has a convergent subsequence in the ambient space. For a subset to be compact, its limits must also remain in the subset.

Failure in infinite-dimensional normed spaces

In the Hilbert space ℓ2\ell^2, let ene_n be the sequence with 11 in coordinate nn and 00 elsewhere. Then

∥en∥=1,∥en−em∥=2(n≠m).\|e_n\|=1,\qquad \|e_n-e_m\|=\sqrt2\quad(n\ne m).

The sequence is bounded, but no subsequence is Cauchy, so no subsequence converges in norm.

More generally, a normed vector space has the property that every bounded sequence admits a norm-convergent subsequence if and only if it is finite-dimensional. Infinite-dimensional functional analysis therefore uses additional hypotheses or weaker notions of convergence rather than boundedness alone.

Applications

Existence of maxima and minima

The theorem underlies the extreme value theorem: a continuous real-valued function on a nonempty compact set attains its maximum and minimum.

For example, let K⊆RdK\subseteq\mathbb{R}^d be nonempty, closed, and bounded, and let f:K→Rf:K\to\mathbb{R} be continuous. Sequential compactness first shows that ff is bounded: otherwise, points could be chosen where ∣f∣|f| grows without bound, while a convergent subsequence and continuity would force the corresponding function values to converge to a finite number.

Writing m=inf⁡x∈Kf(x)m=\inf_{x\in K}f(x), choose xn∈Kx_n\in K such that f(xn)→mf(x_n)\to m. A subsequence converges to some x∗∈Kx_*\in K, and continuity gives

f(x∗)=lim⁡k→∞f(xnk)=m.f(x_*)=\lim_{k\to\infty}f(x_{n_k})=m.

A maximum follows by applying the same argument to −f-f. This is a basic existence argument in mathematical optimization.

Convergence criteria

If a bounded sequence in Rd\mathbb{R}^d has exactly one possible subsequential limit LL, then the full sequence converges to LL.

Otherwise, some ε>0\varepsilon>0 would allow infinitely many terms satisfying ∥xn−L∥≥ε\|x_n-L\|\geq\varepsilon. Those terms form a bounded subsequence, which has a convergent further subsequence. Its limit would remain at least ε\varepsilon from LL, contradicting uniqueness.

The theorem also helps establish that every real Cauchy sequence converges: a Cauchy sequence is bounded, hence has a convergent subsequence, and the Cauchy condition forces the entire sequence to have that same limit.

Historical context

The theorem is named after Bernard Bolzano and Karl Weierstrass. Bolzano established an early form of the underlying boundedness and accumulation principle in his 1817 work on continuous functions. Weierstrass later developed and disseminated closely related results as part of the rigorous foundations of analysis.

Its modern formulation reflects the nineteenth-century separation of boundedness, completeness, and convergence into explicit concepts, followed by their unification through the theory of compactness.