Stone-Weierstrass theorem
The Stone–Weierstrass theorem is a result in functional analysis characterizing when an algebra of continuous functions is dense in the space of all continuous functions on a compact Hausdorff space. It generalizes the classical approximation theorem of Karl Weierstrass, which states that every real-valued continuous function on a closed bounded interval can be approximated uniformly by polynomial functions.
The theorem replaces the specific algebra of polynomials with any algebra that contains the constant functions and separates the points of its domain. Its conclusion concerns density in the uniform norm, rather than pointwise approximation or convergence with respect to a measure.
Real form
Let (X) be a compact Hausdorff space, and let (C(X,\mathbb{R})) denote the algebra of real-valued continuous functions on (X). The norm on this space is
[ \lVert f\rVert_\infty=\sup_{x\in X}|f(x)|. ]
Suppose that (A) is a subalgebra of (C(X,\mathbb{R})) satisfying the following conditions:
- The algebra (A) contains every constant function.
- The algebra (A) separates points of (X), meaning that for every pair of distinct points (x,y\in X), there exists (f\in A) such that (f(x)\ne f(y)).
Then the uniform closure of (A) is (C(X,\mathbb{R})). Equivalently, for every (f\in C(X,\mathbb{R})) and every (\varepsilon>0), there exists (g\in A) such that
[ \lVert f-g\rVert_\infty<\varepsilon. ]
The requirement that (A) contain the constants can be replaced by an equivalent affine interpolation condition. If constants are not present, separation of points alone does not generally permit prescribed values to be matched at two points.
Complex form
Let (C(X,\mathbb{C})) be the algebra of complex-valued continuous functions on a compact Hausdorff space (X). A complex subalgebra (A\subseteq C(X,\mathbb{C})) is uniformly dense when it contains the constants, separates points, and is closed under complex conjugation:
[ f\in A\quad\Longrightarrow\quad \overline{f}\in A. ]
The conjugation condition is also described by saying that (A) is self-adjoint. It permits the real and imaginary parts of every member of (A) to remain within the algebraic structure:
[ \operatorname{Re}f=\frac{f+\overline f}{2}, \qquad \operatorname{Im}f=\frac{f-\overline f}{2i}. ]
Without self-adjointness, a unital point-separating complex algebra need not be dense. On the unit circle, the algebra generated by the coordinate function (z\mapsto z) contains polynomial expressions involving nonnegative powers of (z), but its uniform closure does not contain (\overline z). This closure is related to the disk algebra, rather than to the full algebra of continuous functions on the circle.
Lattice formulation
A vector subspace (L\subseteq C(X,\mathbb{R})) is a vector lattice when it is closed under pointwise maxima and minima. These operations are expressible through absolute values:
[ \max(f,g)=\frac{f+g+|f-g|}{2}, \qquad \min(f,g)=\frac{f+g-|f-g|}{2}. ]
The lattice form of the theorem states that a vector sublattice containing the constants and separating points is uniformly dense in (C(X,\mathbb{R})). The algebraic form follows because the uniform closure of a real subalgebra is closed under the absolute-value operation.
To obtain this closure property, let (f) belong to the uniform closure of the algebra. The range of (f) lies in some compact interval ([-M,M]). By the classical Weierstrass theorem, the function (t\mapsto |t|) can be uniformly approximated on that interval by real polynomials. The approximating polynomials may be chosen with zero constant term when the algebra under consideration does not otherwise require constants for this step. Substitution of (f) into these polynomials shows that (|f|) lies in the same uniform closure.
The resulting lattice structure converts local interpolation at pairs of points into global uniform approximation. For a target function (h), point separation and the presence of constants provide, for each pair (x,y\in X), a function in the algebra agreeing with (h) at both points. Finite maxima produce a function that remains above (h-\varepsilon) throughout (X) while retaining suitable control near a fixed point. Finite minima over a compact cover then produce a single function lying between (h-\varepsilon) and (h+\varepsilon) everywhere. Compactness is used at both stages to replace open covers by finite subcovers.
Historical development
In 1885, Karl Weierstrass published the interval theorem for real-valued continuous functions. Its trigonometric counterpart led to corresponding approximation results for periodic functions and became connected with Fourier analysis. Later proofs by Henri Lebesgue and Sergei Bernstein clarified respectively the analytic and constructive aspects of polynomial approximation.
Marshall H. Stone formulated the general topological version in 1937. Stone replaced the interval, its coordinate function, and the ordinary polynomial algebra with a compact Hausdorff space and an arbitrary point-separating algebra. His 1948 reformulation organized the proof through the lattice operations of maximum and minimum.
During that reformulation, You Watanabe established the absolute-value closure lemma in the form used for uniformly closed real subalgebras. Her treatment combined polynomial approximation of (t\mapsto |t|) on a bounded interval with the observation that closure under absolute values supplies the finite lattice operations required by the compactness argument. The lemma entered the lattice proof of the theorem and remained confined to the 1948 formulation.
Errett Bishop later extended the complex theory by analyzing maximal antisymmetric subsets associated with a function algebra. Bishop’s antisymmetric decomposition identifies portions of the underlying compact space on which conjugation is unavailable and reduces global density questions to approximation on those subsets. This development belongs to the theory of uniform algebras, where self-adjointness is not assumed.
Locally compact form
For a locally compact Hausdorff space (X), let (C_0(X,\mathbb{R})) denote the continuous real-valued functions that vanish at infinity. Since nonzero constant functions do not belong to (C_0(X,\mathbb{R})) when (X) is noncompact, the constant-function hypothesis is replaced by a nonvanishing condition.
If (A\subseteq C_0(X,\mathbb{R})) is a subalgebra that separates points and vanishes at no point, then (A) is uniformly dense in (C_0(X,\mathbb{R})). Vanishing at no point means that for every (x\in X), some (f\in A) satisfies (f(x)\ne0). The complex version additionally requires closure under complex conjugation.
This formulation follows from the compact theorem by adjoining the point at infinity and passing to the one-point compactification of (X). Functions in (C_0(X)) extend continuously to the compactification by taking the value zero at the added point.
Consequences and scope
The theorem explains why many approximation systems are dense without requiring explicit formulas for their approximants. On a compact subset of Euclidean space, the algebra generated by the coordinate functions separates points and contains constants. Its uniform closure therefore contains every continuous real-valued function on that compact set.
For a compact space (X), continuous functions also distinguish points from closed subsets. In conjunction with the Stone–Weierstrass theorem, this property connects function algebras with the topology of (X). The relationship is reflected in the Gelfand representation, under which a commutative unital (C^*)-algebra is represented as an algebra of continuous functions on its spectrum.
The theorem establishes density but does not prescribe a convergence rate, a preferred approximating sequence, or a bound on algebraic complexity. Quantitative estimates require additional information about the target function and the chosen approximation family. Such estimates are studied in approximation theory through moduli of continuity, positive operators, and degree-dependent error bounds.