Fact-checked by Grok 2 weeks ago

Riesz representation theorem

The Riesz representation theorem refers to a family of fundamental results in functional analysis that characterize the dual spaces of specific normed spaces of functions, establishing isomorphisms between bounded linear functionals and more concrete objects such as inner products or integrals against measures.^[1]^[2] In the context of Hilbert spaces, the theorem states that for any complex Hilbert space H and every bounded linear functional \phi: H \to \mathbb{C}, there exists a unique vector u \in H such that \phi(v) = \langle u, v \rangle for all v \in H, where \langle \cdot, \cdot \rangle denotes the inner product, and the operator norm of \phi equals the norm of u.^[1] This version, originally established by Frigyes Riesz around 1907 in his work on linear functional equations, identifies the dual space H^* with H itself, providing a cornerstone for the theory of self-dual spaces and enabling the spectral theorem for self-adjoint operators.^[1] A related and equally significant formulation, often called the Riesz–Markov–Kakutani representation theorem, applies to spaces of continuous functions. For a locally compact Hausdorff space X, every positive linear functional T on the space C_c(X) of complex-valued continuous functions with compact support can be represented uniquely as T(f) = \int_X f \, d\mu for some Radon measure \mu on X, where the integral is understood in the sense of Lebesgue integration.^[2] This result originated with Riesz's 1909 proof for the unit interval [0,1], was extended by Andrey Markov in 1938 to certain non-compact spaces, and generalized by Shizuo Kakutani in 1941 to compact Hausdorff spaces, with the full version encompassing signed and complex measures.^[2] These theorems underpin key developments in modern analysis, including the construction of Haar measures on locally compact groups, the duality theory for C^*-algebras, and the foundations of distribution theory, while their proofs typically rely on tools like the Hahn–Banach theorem, Urysohn's lemma, and the Carathéodory extension procedure.^[1]^[2]

Preliminaries

Hilbert spaces

A Hilbert space is a complete inner product space over the real or complex numbers. More precisely, it is a vector space H equipped with an inner product \langle \cdot, \cdot \rangle: H \times H \to \mathbb{F} (where \mathbb{F} = \mathbb{R} or \mathbb{C}) that induces a norm \|x\| = \sqrt{\langle x, x \rangle} for all x \in H, such that the metric space (H, d) with d(x, y) = \|x - y\| is complete—meaning every Cauchy sequence in H converges to an element in H.^[3]^[4]^[1] The inner product must satisfy linearity in the second argument, conjugate symmetry \langle x, y \rangle = \overline{\langle y, x \rangle}, and positive-definiteness \langle x, x \rangle \geq 0 with equality if and only if x = 0. This structure ensures the norm obeys the parallelogram law \|x + y\|^2 + \|x - y\|^2 = 2(\|x\|^2 + \|y\|^2), distinguishing Hilbert spaces from general Banach spaces.^[1]^[4] The concept of Hilbert space originated in the early 20th century through the work of David Hilbert on integral equations, where he considered infinite-dimensional analogs of Euclidean spaces to solve problems in quadratic forms and spectral theory. Hilbert's foundational contributions around 1904–1910 laid the groundwork, with further developments by Erhard Schmidt and Frigyes Riesz formalizing the abstract framework by the 1910s.^[5] Key properties include the Cauchy-Schwarz inequality |\langle x, y \rangle| \leq \|x\| \|y\|, orthogonality (vectors x, y are orthogonal if \langle x, y \rangle = 0), and the existence of orthonormal bases: every Hilbert space has an orthonormal set that is maximal and complete, allowing Parseval's identity \|x\|^2 = \sum_{n} |\langle x, e_n \rangle|^2 for any orthonormal basis \{e_n\}. These features enable projections onto closed subspaces and underpin operator theory in functional analysis.^[1]^[4] Examples abound in both finite and infinite dimensions. Finite-dimensional Hilbert spaces include \mathbb{R}^n and \mathbb{C}^n with the standard dot product \langle x, y \rangle = \sum_{i=1}^n x_i \overline{y_i}. Infinite-dimensional instances are the sequence space \ell^2 of square-summable sequences (a_n) with \langle a, b \rangle = \sum_{n=1}^\infty a_n \overline{b_n} < \infty, and the function space L^2(\mu) of square-integrable functions on a measure space (X, \mu) with \langle f, g \rangle = \int_X f \overline{g} \, d\mu. All separable Hilbert spaces are isometric to \ell^2.^[3]^[4]

Inner products and dual spaces

An inner product space, also known as a pre-Hilbert space, is a vector space equipped with an inner product, which is a bilinear form that generalizes the dot product from Euclidean spaces. In the real case, the inner product \langle \cdot, \cdot \rangle: V \times V \to \mathbb{R} is bilinear (linear in both arguments) and symmetric \langle x, y \rangle = \langle y, x \rangle, with positive-definiteness \langle x, x \rangle \geq 0 with equality if and only if x = 0.^[4] For complex vector spaces, the inner product \langle \cdot, \cdot \rangle: V \times V \to \mathbb{C} is sesquilinear, meaning conjugate-linear in the first argument and linear in the second (\langle \lambda x, y \rangle = \overline{\lambda} \langle x, y \rangle), Hermitian symmetric (\langle x, y \rangle = \overline{\langle y, x \rangle}), and positive-definite.^[4] These properties ensure the inner product induces a norm \|x\| = \sqrt{\langle x, x \rangle}, which in turn defines a metric and topology on the space.^[4] The Cauchy-Schwarz inequality, |\langle x, y \rangle| \leq \|x\| \|y\|, follows directly from the positive-definiteness of the inner product and bounds the "angle" between vectors, enabling concepts like orthogonality (\langle x, y \rangle = 0) and projections.^[4] A Hilbert space is a complete inner product space, meaning every Cauchy sequence converges with respect to the induced norm; this completeness distinguishes Hilbert spaces from general inner product spaces and makes them suitable for analysis akin to Euclidean spaces but in infinite dimensions.^[4] Canonical examples include \ell^2(\mathbb{N}), the space of square-summable sequences with inner product \langle (a_n), (b_n) \rangle = \sum_n a_n \overline{b_n}, and L^2(\mathbb{R}), the space of square-integrable functions with \langle f, g \rangle = \int_{-\infty}^{\infty} f(x) \overline{g(x)} \, dx.^[4] The dual space H^* of a normed space H, such as a Hilbert space, consists of all continuous linear functionals \phi: H \to \mathbb{K} (where \mathbb{K} = \mathbb{R} or \mathbb{C}), equipped with the operator norm \|\phi\| = \sup_{\|x\| \leq 1} |\phi(x)|.^[4] In a Hilbert space, the inner product naturally identifies vectors with functionals via the map x \mapsto \phi_x, where \phi_x(y) = \langle x, y \rangle; this map is antilinear in the complex case (due to conjugate-linearity in the first slot) and isometric, providing an isometric isomorphism between H and H^*.^[4] Unlike general Banach spaces, where the dual may not be isomorphic to the space itself, the structure of Hilbert spaces allows this canonical pairing to characterize the entire dual, providing a concrete representation for linear functionals that underpins applications in quantum mechanics, signal processing, and approximation theory.^[4]

Linear and antilinear functionals

In the context of a complex Hilbert space H, a linear functional is a map \phi: H \to \mathbb{C} satisfying \phi(\alpha x + \beta y) = \alpha \phi(x) + \beta \phi(y) for all \alpha, \beta \in \mathbb{C} and x, y \in H.^[6] Such functionals are of central importance in functional analysis, as they form the algebraic dual space H^*, though in Hilbert spaces, attention is typically restricted to the continuous (or bounded) ones, where |\phi(x)| \leq M \|x\| for some M > 0 and all x \in H.^[7] The norm of a bounded linear functional is defined as \|\phi\| = \sup_{\|x\|=1} |\phi(x)|, which coincides with the operator norm induced by the Hilbert space norm.^[7] For real Hilbert spaces, the definition simplifies analogously, with linearity over \mathbb{R}, and all continuous linear functionals are representable via the inner product in a manner that aligns directly with the Riesz representation theorem. In the complex case, however, the sesquilinear nature of the inner product—conjugate-linear in the first argument and linear in the second—plays a key role in characterizing these functionals.^[6] An antilinear functional, also called conjugate-linear, is a map \psi: H \to \mathbb{C} satisfying \psi(\alpha x + \beta y) = \overline{\alpha} \psi(x) + \overline{\beta} \psi(y) for all \alpha, \beta \in \mathbb{C} and x, y \in H.^[6] Bounded antilinear functionals are defined similarly, with continuity equivalent to |\psi(x)| \leq M \|x\|. Unlike linear functionals, antilinear ones do not form the standard dual space but arise naturally in the structure of Hilbert spaces; for instance, the map z \mapsto \langle z, x \rangle is antilinear in z for fixed x \in H, reflecting the conjugate symmetry \langle x, y \rangle = \overline{\langle y, x \rangle}.^[6] The distinction between linear and antilinear functionals is crucial for the Riesz representation theorem in complex spaces, where the theorem identifies the dual H^* (of bounded linear functionals) with H via an antilinear isomorphism J: H \to H^* given by J(y)(x) = \langle y, x \rangle, satisfying J(\alpha y) = \overline{\alpha} J(y) and preserving norms \|J(y)\| = \|y\|.^[7] This antilinearity ensures compatibility with the inner product convention, distinguishing the complex case from the real one, where the isomorphism is linear. In applications, such as quantum mechanics, antilinear functionals appear in anti-unitary operators, but in pure functional analysis, they underscore the reflexive structure of Hilbert spaces.^[7]

Statement of the theorem

Version for complex Hilbert spaces

In a complex Hilbert space H, the Riesz representation theorem asserts that every continuous linear functional \phi: H \to \mathbb{C} can be uniquely expressed in terms of the inner product. Specifically, there exists a unique vector y \in H such that

\phi(x) = \langle x, y \rangle

for all x \in H, where \langle \cdot, \cdot \rangle: H \times H \to \mathbb{C} denotes the inner product, which is linear in the first argument, conjugate-linear (antilinear) in the second argument, and satisfies \langle x, x \rangle \geq 0 with equality if and only if x = 0.^[8]^[1] The uniqueness of y follows from the positive definiteness of the inner product: if \langle x, y_1 \rangle = \langle x, y_2 \rangle for all x \in H, then \langle y_1 - y_2, y_1 - y_2 \rangle = 0, implying y_1 = y_2. Moreover, the operator norm of \phi equals the Hilbert space norm of y, i.e., \|\phi\| = \|y\|, which is established via the Cauchy-Schwarz inequality: |\phi(x)| = |\langle x, y \rangle| \leq \|x\| \|y\| for all x, with equality achieved when x is a scalar multiple of y.^[1]^[9] This representation identifies the continuous dual space H^* with H itself via the antilinear isomorphism y \mapsto \phi_y, where \phi_y(x) = \langle x, y \rangle, preserving the norm and thus making H isometrically isomorphic to its dual. The theorem relies on the completeness of H and the sesquilinear nature of the inner product, distinguishing the complex case from the real case where the inner product is bilinear.^[8]^[1]

Version for real Hilbert spaces

In real Hilbert spaces, the inner product is a symmetric bilinear form, taking values in the real numbers \mathbb{R}, and satisfying \langle x, y \rangle = \langle y, x \rangle for all x, y \in H.^[10] A continuous linear functional on such a space H is a bounded linear map f: H \to \mathbb{R}, meaning there exists M > 0 such that |f(x)| \leq M \|x\| for all x \in H, where \|\cdot\| denotes the norm induced by the inner product, \|x\| = \sqrt{\langle x, x \rangle}.^[10] The Riesz representation theorem for real Hilbert spaces asserts that every continuous linear functional arises uniquely from the inner product with a fixed vector in the space. Specifically, let H be a real Hilbert space. For any continuous linear functional f: H \to \mathbb{R}, there exists a unique y \in H such that

f(x) = \langle x, y \rangle

for all x \in H. Moreover, the operator norm of f, defined as \|f\| = \sup_{\|x\| \leq 1} |f(x)|, equals the Hilbert space norm of y, i.e., \|f\| = \|y\|.^[11]^[12] This representation identifies the continuous dual space H^* of H isometrically with H itself via the map y \mapsto (x \mapsto \langle x, y \rangle).^[10] Uniqueness follows from the positive definiteness of the inner product: if \langle x, y_1 \rangle = \langle x, y_2 \rangle for all x \in H, then \langle y_1 - y_2, y_1 - y_2 \rangle = 0, implying y_1 = y_2. The norm equality \|f\| = \|y\| is established by applying the Cauchy-Schwarz inequality, which yields |f(x)| = |\langle x, y \rangle| \leq \|x\| \|y\|, so \|f\| \leq \|y\|, and conversely, \|y\|^2 = \langle y, y \rangle = f(y / \|y\|) \cdot \|y\| \leq \|f\| \|y\| if y \neq 0, implying \|y\| \leq \|f\|.^[11]^[12] This version differs from the complex case primarily in the absence of complex conjugation; in real spaces, the inner product is not sesquilinear but bilinear, and the representing vector y satisfies f(x) = \langle x, y \rangle without adjustment for antilinearity. The theorem, originally due to Frigyes Riesz in his 1907 work on quadratic mean spaces, underpins the reflexivity of real Hilbert spaces and extends to finite-dimensional Euclidean spaces, where it recovers the standard dot product representation of linear functionals.^[10]^[11]

Antilinear case

In complex Hilbert spaces, the Riesz representation theorem admits a natural extension to continuous antilinear functionals. An antilinear functional \phi: H \to \mathbb{C} on a complex Hilbert space H satisfies \phi(\alpha x + \beta y) = \overline{\alpha} \phi(x) + \overline{\beta} \phi(y) for all \alpha, \beta \in \mathbb{C} and x, y \in H. The theorem asserts that every such continuous antilinear functional \phi can be uniquely represented as

\phi(x) = \langle y, x \rangle

for some y \in H, where the inner product \langle \cdot, \cdot \rangle is linear in its first argument and conjugate-linear (antilinear) in its second argument.^[11] This representation preserves the norm: \|\phi\| = \sup_{\|x\| \leq 1} |\phi(x)| = \|y\|. The uniqueness of y follows from the positive-definiteness of the inner product: if \langle y_1, x \rangle = \langle y_2, x \rangle for all x \in H, then \langle y_1 - y_2, y_1 - y_2 \rangle = 0, implying y_1 = y_2.^[11] To establish existence, define the associated continuous linear functional \lambda: H \to \mathbb{C} by \lambda(x) = \overline{\phi(x)}. By the standard (linear) Riesz representation theorem, there exists a unique z \in H such that \lambda(x) = \langle x, z \rangle for all x \in H. Then,

\phi(x) = \overline{\langle x, z \rangle} = \langle z, x \rangle,

using the conjugate-symmetry of the inner product \langle u, v \rangle = \overline{\langle v, u \rangle}. Thus, y = z provides the desired representation. This construction also confirms continuity, as |\phi(x)| = |\langle y, x \rangle| \leq \|y\| \|x\|.^[11] This antilinear version is particularly relevant in applications involving symmetry operations, such as time-reversal in quantum mechanics, where antilinear operators map states to their conjugates while preserving norms. The theorem underscores the self-duality of Hilbert spaces, extending the identification of the dual space to the space of continuous antilinear functionals as well.^[11]

Proofs and constructions

General proof outline

The proof of the Riesz representation theorem relies on the structure of Hilbert spaces, particularly the existence of orthogonal complements and projections onto closed subspaces. For a continuous linear functional f: H \to \mathbb{K} on a Hilbert space H over \mathbb{K} = \mathbb{R} or \mathbb{C}, where \|f\| < \infty, the kernel \ker f = \{x \in H : f(x) = 0\} is a closed hyperplane (proper subspace if f \neq 0).^[1] If f = 0, set g = 0, satisfying f(x) = \langle g, x \rangle for all x \in H. Otherwise, the orthogonal complement M = (\ker f)^\perp is one-dimensional, as \ker f has codimension one and the decomposition H = \ker f \oplus M holds by the properties of Hilbert spaces. Select a unit vector y \in M (so \|y\| = 1), noting that f(y) \neq 0 since y \notin \ker f. Define g = f(y) y. For arbitrary x \in H, decompose x = z + \alpha y where z \in \ker f and \alpha = \langle y, x \rangle (in the real case; for complex, the inner product is sesquilinear, linear in the second argument). Then f(x) = f(z) + \alpha f(y) = \alpha f(y) = f(y) \langle y, x \rangle = \langle g, x \rangle, using the linearity of f and the reproducing property of the inner product. The norm equality \|f\| = \|g\| follows from the boundedness of f and Cauchy-Schwarz inequality, as \|f\| = \sup_{\|x\|=1} |f(x)| = |f(y)| = \|g\|. Uniqueness of g arises from the positive-definiteness of the inner product: if \langle g_1, x \rangle = \langle g_2, x \rangle for all x, then \langle g_1 - g_2, g_1 - g_2 \rangle = 0 implies g_1 = g_2. This construction establishes an isometric isomorphism between H and its dual H^*, mapping g \mapsto (x \mapsto \langle g, x \rangle), which is linear (antilinear in the complex case depending on convention). For the separable case, an alternative approach uses an orthonormal basis \{\phi_j\} to express f(\phi_j) = a_j, form g = \sum \overline{a_j} \phi_j (square-summable by boundedness), and extend by continuity, yielding the same representation.^[12]

Finite-dimensional example

In finite-dimensional Hilbert spaces, the Riesz representation theorem establishes that every linear functional is continuous and admits a unique representation via the inner product, reflecting the canonical isomorphism between the space and its dual. Specifically, let V be a finite-dimensional inner product space over \mathbb{R} or \mathbb{C} with inner product \langle \cdot, \cdot \rangle. For any linear functional \phi: V \to \mathbb{F} (where \mathbb{F} is the underlying field), there exists a unique vector u \in V such that \phi(v) = \langle u, v \rangle for all v \in V.^[13]^[14] This holds because, in finite dimensions, all linear functionals are automatically bounded, eliminating the need for separate continuity assumptions present in the infinite-dimensional case.^[7] To construct the representing vector u, select an orthonormal basis \{e_1, \dots, e_n\} for V. Define

u = \sum_{i=1}^n \overline{\phi(e_i)} e_i

(for real spaces, omit the conjugate). For any v \in V, express v = \sum_{i=1}^n \langle e_i, v \rangle e_i. Then,

\phi(v) = \sum_{i=1}^n \langle e_i, v \rangle \phi(e_i),

by linearity of \phi. On the other hand,

\langle u, v \rangle = \left\langle \sum_{i=1}^n \overline{\phi(e_i)} e_i, \sum_{j=1}^n \langle e_j, v \rangle e_j \right\rangle = \sum_{i=1}^n \overline{\phi(e_i)} \langle e_i, v \rangle,

using the sesquilinearity and orthonormality of the basis (noting that \langle e_i, v \rangle = \overline{\langle v, e_i \rangle} in the complex case, but the representation aligns with the conjugated coefficients). Thus, \phi(v) = \langle u, v \rangle. Uniqueness follows from the non-degeneracy of the inner product: if \langle u_1, v \rangle = \langle u_2, v \rangle for all v, then u_1 = u_2.^[13]^[14] A concrete illustration arises in V = \mathbb{R}^2 equipped with the Euclidean inner product \langle x, y \rangle = x_1 y_1 + x_2 y_2. Consider the linear functional \phi: \mathbb{R}^2 \to \mathbb{R} defined by \phi(x_1, x_2) = 3x_1 + 4x_2. Using the standard orthonormal basis \{e_1 = (1,0), e_2 = (0,1)\}, compute \phi(e_1) = 3 and \phi(e_2) = 4, yielding u = 3e_1 + 4e_2 = (3,4). Indeed, \phi(x) = \langle (3,4), x \rangle = 3x_1 + 4x_2 for all x = (x_1, x_2), confirming the representation. This example underscores how the theorem reduces linear functionals to simple projections onto a unique direction in the space.^[7]^[13]

Explicit representing vector

The explicit construction of the representing vector in the Riesz representation theorem relies on the structure of the Hilbert space, particularly when an orthonormal basis is available. For a Hilbert space H over \mathbb{C} or \mathbb{R}, given a continuous linear functional \phi: H \to \mathbb{K} (where \mathbb{K} is the scalar field), the theorem guarantees a unique v \in H such that \phi(x) = \langle v, x \rangle for all x \in H, with \|\phi\| = \|v\|.^[1] In separable Hilbert spaces, which admit a countable orthonormal basis \{e_n\}_{n=1}^\infty, the representing vector v can be constructed directly from the values of \phi on the basis elements:

v = \sum_{n=1}^\infty \overline{\phi(e_n)} e_n.

This infinite series converges in the norm topology of H, as the coefficients satisfy \sum_{n=1}^\infty |\phi(e_n)|^2 \leq \|\phi\|^2 < \infty by the boundedness of \phi and Bessel's inequality applied to the partial sums.^[1] To verify the representation, expand any x \in H in the basis: x = \sum_{n=1}^\infty \langle e_n, x \rangle e_n. Then, by linearity and continuity of \phi,

\phi(x) = \sum_{n=1}^\infty \langle e_n, x \rangle \phi(e_n) = \left\langle \sum_{n=1}^\infty \overline{\phi(e_n)} e_n, x \right\rangle = \langle v, x \rangle,

where the inner product series converges due to the Cauchy-Schwarz inequality and the square-summability of the coefficients. The norm equality \|v\|^2 = \sum_{n=1}^\infty |\phi(e_n)|^2 = \|\phi\|^2 follows from Parseval's identity. This construction is particularly useful in applications like L^2 spaces, where standard bases (e.g., Fourier basis) allow computational evaluation of v.^[1]^[14] For finite-dimensional Hilbert spaces, the construction simplifies to a finite sum over an orthonormal basis \{e_1, \dots, e_n\}: v = \sum_{k=1}^n \overline{\phi(e_k)} e_k. This directly extends the Gram-Schmidt process for finding coordinates. In non-separable spaces, an analogous construction uses any orthonormal basis (whose existence follows from the axiom of choice), but explicit computation requires specifying the basis, which may not be constructive without additional structure.^[14]^[1] An alternative basis-free construction proceeds via the kernel of \phi. If \phi \neq 0, let M = \ker \phi, a closed hyperplane. The orthogonal complement M^\perp is one-dimensional. Select any nonzero w \in M^\perp; then v = \frac{\phi(w)}{\|w\|^2} w satisfies the representation, as for x = y + c \frac{w}{\|w\|} with y \in M and c = \langle w/\|w\|, x \rangle, we have \phi(x) = c \phi(w) / \|w\| = \langle v, x \rangle. For complex spaces, the formula adjusts for the sesquilinear inner product, ensuring linearity in the first argument. This method highlights the geometric role of v as lying in the direction maximizing \phi.^[15]

Reflexivity of Hilbert spaces

The Riesz representation theorem establishes that every continuous linear functional on a Hilbert space H can be uniquely represented as an inner product with a fixed vector in H, thereby creating an isometric isomorphism between H and its continuous dual H^*. Specifically, for any f \in H^*, there exists a unique y \in H such that f(x) = \langle y, x \rangle for all x \in H, and \|f\| = \|y\|.^[1]^[16] A Banach space is reflexive if the canonical embedding J: H \to H^{**}, defined by J(x)(\phi) = \phi(x) for \phi \in H^*, is an isometric isomorphism onto the bidual H^{**}. In the case of Hilbert spaces, the Riesz representation theorem ensures reflexivity via the isomorphism \Phi: H \to H^* given by \Phi(y)(x) = \langle y, x \rangle. To see surjectivity, consider any \Lambda \in H^{**}. The composition \Lambda \circ \Phi: H \to \mathbb{C} is a continuous linear functional on H, so by the Riesz theorem, there exists a unique z \in H such that (\Lambda \circ \Phi)(x) = \langle z, x \rangle for all x \in H. Thus, for any \phi \in H^*, write \phi = \Phi(w) for unique w \in H; then \Lambda(\phi) = \Lambda(\Phi(w)) = \langle z, w \rangle = J(z)(\phi). Hence, \Lambda = J(z), showing J is surjective. The map J is clearly linear and isometric (as \|J(x)\| = \sup_{\|\phi\| \leq 1} |\phi(x)| = \|x\| by the Riesz theorem), and injective (if J(x) = 0, then \langle x, x \rangle = 0 implies x = 0). Therefore, J is an isometric isomorphism, confirming that Hilbert spaces are reflexive.^[16]^[17] This reflexivity property distinguishes Hilbert spaces from more general Banach spaces and has significant implications in functional analysis, such as ensuring the weak closure of convex sets coincides with their strong closure under certain conditions. The proof relies fundamentally on the inner product structure, which enables the explicit representation absent in non-Hilbert settings.^[1]

Adjoint operators

In Hilbert spaces, the Riesz representation theorem plays a fundamental role in establishing the existence and uniqueness of adjoint operators for bounded linear maps. Consider two complex Hilbert spaces H_1 and H_2, and let T: H_1 \to H_2 be a bounded linear operator. The adjoint operator T^*: H_2 \to H_1 is defined by the relation

\langle y, T x \rangle_{H_2} = \langle T^* y, x \rangle_{H_1}

for all x \in H_1 and y \in H_2, where \langle \cdot, \cdot \rangle_{H_i} denotes the inner product on H_i. This definition ensures that T^* preserves the sesquilinear form induced by the inner products.^[18]^[19] To prove the existence of T^*, fix y \in H_2 and define the functional \phi_y: H_1 \to \mathbb{C} by \phi_y(x) = \langle y, T x \rangle_{H_2}. This functional is linear in x and continuous, since |\phi_y(x)| \leq \|y\| \cdot \|T x\| \leq \|T\| \|y\| \|x\|, where \|T\| is the operator norm of T. By the Riesz representation theorem applied to H_1, there exists a unique z_y \in H_1 such that \phi_y(x) = \langle z_y, x \rangle_{H_1} for all x \in H_1. Setting T^* y = z_y yields the desired relation, and the map y \mapsto T^* y is well-defined.^[18]^[19] The operator T^* is linear, as linearity of the inner product implies T^*(\alpha y_1 + y_2) = \alpha T^* y_1 + T^* y_2 for scalars \alpha and vectors y_1, y_2 \in H_2. Moreover, T^* is bounded with \|T^*\| \leq \|T\|, obtained via the Cauchy-Schwarz inequality: \|T^* y\|^2 = \langle T^* y, T^* y \rangle_{H_1} = \langle y, T (T^* y) \rangle_{H_2} \leq \|y\| \|T (T^* y)\| \leq \|T\| \|y\| \|T^* y\|, which simplifies to the bound. Uniqueness follows from the uniqueness in the Riesz theorem; if another operator S^* satisfies the relation, then \langle (T^* - S^*) y, x \rangle_{H_1} = 0 for all x, y, implying T^* y = S^* y. In the case of real Hilbert spaces, the argument is analogous, with the inner product bilinear instead of sesquilinear.^[18]^[19] Several key properties of adjoints derive directly from this construction. The double adjoint satisfies (T^*)^* = T, since applying the process twice recovers the original operator via the defining relation. Composition reverses under adjoints: (S T)^* = T^* S^* for bounded linear S: H_2 \to H_3 and T: H_1 \to H_2. The norms are equal, \|T^*\| = \|T\|, and \|T T^*\| = \|T^* T\| = \|T\|^2, reflecting the isometry-like behavior induced by the inner product structure. Additionally, the kernel and range satisfy N(T) = R(T^*)^\perp and N(T)^\perp = \overline{R(T^*)}, linking null spaces to orthogonal complements of ranges. These relations underscore how the Riesz theorem equips Hilbert spaces with a rich duality that facilitates operator theory.^[18]^[19] For finite-dimensional examples, such as T represented by a matrix A on \mathbb{C}^n with the standard inner product, the adjoint corresponds to the conjugate transpose A^*, where \langle y, A x \rangle = \langle A^* y, x \rangle (with x^* the conjugate transpose of x). This illustrates the theorem's role in concrete settings, extending seamlessly to infinite dimensions via Riesz.^[18]

Self-adjoint and unitary operators

A bounded operator A \in B(H) on a complex Hilbert space H is self-adjoint if A = A^*. An equivalent characterization, derived via the Riesz theorem, is that A is self-adjoint if and only if \langle A f, f \rangle is real for every f \in H. Self-adjoint operators inherit key spectral properties from this structure; for instance, their spectrum lies on the real line, and the norm equals the spectral radius: \|A\| = r(A) = \sup \{ |\lambda| : \lambda \in \sigma(A) \}. Moreover, the Riesz theorem facilitates the spectral theorem for self-adjoint operators, which states that every such A is unitarily equivalent to a multiplication operator M_h on L^2(X, \mu) for some measure space (X, \mu) and real-valued function h, with \sigma(A) = h(X). This decomposition underscores the role of self-adjoint operators in representing observables in quantum mechanics.^[20] A bounded operator U \in B(H) is unitary if U^* U = U U^* = I, or equivalently, if U is a surjective isometry, preserving the inner product: \langle U x, U y \rangle = \langle x, y \rangle for all x, y \in H. The Riesz theorem ensures that unitary operators are normal (U U^* = U^* U) and isometric (\|U x\| = \|x\|), and are self-adjoint if and only if U^2 = I (e.g., unitary involutions). The spectral theorem for unitary operators, building on Riesz via adjoint constructions, asserts that every unitary U is unitarily equivalent to a multiplication operator on L^2(Y, \nu) by a function with values on the unit circle, so \sigma(U) \subseteq \{ z \in \mathbb{C} : |z| = 1 \}. This property highlights unitary operators' role in implementing symmetries and time evolutions in Hilbert space settings.^[20]

Applications and extensions

In functional analysis

The Riesz representation theorem establishes that every continuous linear functional on a Hilbert space H can be uniquely expressed as an inner product with a fixed vector in H, providing an explicit isomorphism between H and its continuous dual H^*. This self-duality simplifies many constructions in functional analysis, enabling concrete computations of dual elements without abstract machinery. For instance, it allows direct verification of boundedness and continuity properties for functionals derived from inner products, which is essential for analyzing convergence and approximation in infinite-dimensional settings.^[12] A prominent application arises in the theory of reproducing kernel Hilbert spaces (RKHS), where the theorem underpins the reproducing property. In an RKHS of functions on a set X, the evaluation functional \mathrm{ev}_x: f \mapsto f(x) is continuous for each x \in X. By Riesz representation, there exists a unique kernel function k_x \in H such that f(x) = \langle f, k_x \rangle_H for all f \in H, with the reproducing kernel K(x,y) = \langle k_y, k_x \rangle_H. This structure, formalized by Aronszajn, classifies Hilbert spaces of functions admitting continuous point evaluations and facilitates interpolation and regularization problems in functional analysis.^[21] In harmonic analysis, the theorem is instrumental in proving the Plancherel theorem, which asserts that the Fourier transform \mathcal{F}: L^2(\mathbb{R}^n) \to L^2(\mathbb{R}^n) extends to a unitary operator preserving the L^2-norm. The proof typically proceeds by establishing Parseval's identity on a dense subspace like Schwartz functions, then using Riesz representation to extend the inner product preservation to the full space: for f, g \in L^2, the functional \phi(h) = \langle \mathcal{F} f, \mathcal{F} g \rangle is represented by a unique element ensuring \|\mathcal{F} f\|_2 = \|f\|_2. This isometry is foundational for decompositions on locally compact groups and underpins applications in partial differential equations and signal processing within functional analysis.^[22]^[23] The theorem also supports developments in operator theory, particularly by enabling explicit forms for adjoints and spectral measures in the context of the spectral theorem for self-adjoint operators on Hilbert spaces. For a densely defined symmetric operator, Riesz representation identifies the adjoint via \langle Tf, g \rangle = \langle f, T^* g \rangle, facilitating the construction of spectral resolutions that diagonalize operators through multiplication by bounded functions on the spectrum. This duality is crucial for functional calculus and resolvent estimates in advanced functional analysis.^[24]

In quantum mechanics

In quantum mechanics, the state space of a physical system is modeled as a complex separable Hilbert space \mathcal{H}, where pure states correspond to unit vectors up to phase. The Riesz representation theorem plays a foundational role by establishing that the continuous dual space \mathcal{H}^* is isometrically isomorphic to \mathcal{H} itself via the inner product: for every bounded linear functional f \in \mathcal{H}^*, there exists a unique vector \phi \in \mathcal{H} such that f(\psi) = \langle \phi | \psi \rangle for all \psi \in \mathcal{H}, with \|f\| = \|\phi\|.^[25] This identification ensures that linear functionals on states, such as those computing transition amplitudes or expectation values, can be represented directly within the Hilbert space without recourse to an abstract dual.^[26] The theorem provides the mathematical justification for Dirac's bra-ket notation, introduced in the 1930s to formalize quantum computations. In this notation, kets |\psi\rangle denote vectors in \mathcal{H}, while bras \langle \phi| represent elements of \mathcal{H}^*, with the inner product \langle \phi | \psi \rangle yielding a complex scalar that encodes probability amplitudes.^[27] For instance, the probability of measuring state |\phi\rangle when prepared in |\psi\rangle is |\langle \phi | \psi \rangle|^2, and the theorem guarantees the uniqueness of the representing vector, avoiding ambiguities in infinite-dimensional spaces.^[28] This isomorphism also underpins the reflexivity of Hilbert spaces, allowing operators and their adjoints to be treated symmetrically, which is essential for defining observables as self-adjoint operators.^[26] A key application arises in computing expectation values of observables. For a self-adjoint operator \hat{A} representing an observable (e.g., position or momentum), the expectation in state |\psi\rangle is \langle \hat{A} \rangle = \langle \psi | \hat{A} \psi \rangle, where the functional \psi \mapsto \langle \psi | \hat{A} \psi \rangle leverages the Riesz representation to remain within \mathcal{H}.^[25] In the position representation, where states are wave functions \psi(x), the inner product becomes \langle \phi | \psi \rangle = \int_{-\infty}^{\infty} \phi^*(x) \psi(x) \, dx, and the theorem ensures that functionals like the position operator \hat{x} \psi(x) = x \psi(x) correspond to multiplication by the coordinate function.^[27] This framework extends to mixed states via density operators, where the Riesz theorem aids in representing statistical mixtures as trace-class operators with unit trace, facilitating the Born rule for probabilities.^[25]

Generalizations to other spaces

The Riesz representation theorem, in its classical form for Hilbert spaces, identifies continuous linear functionals with inner products against fixed vectors. A significant generalization extends this idea to spaces of continuous functions on compact Hausdorff spaces. Specifically, for the Banach space C(K) of continuous real- or complex-valued functions on a compact Hausdorff space K equipped with the supremum norm, every positive linear functional \Lambda: C(K) \to \mathbb{R} (or \mathbb{C}) corresponds uniquely to a regular Borel measure \mu on K such that \Lambda(f) = \int_K f \, d\mu for all f \in C(K). This result, known as the Riesz–Markov representation theorem, was first established by Frigyes Riesz in 1909 for the real case on locally compact spaces and later extended by Andrei Markov in 1938 and Shizuo Kakutani in 1941 to include complex measures and the compact case. Further generalizations address vector-valued functions and more general target spaces. For instance, Singer's representation theorem provides a Banach space analog for the dual of C(\Omega; X), where \Omega is a compact Hausdorff space and X is a Banach space. It states that the continuous dual of C(\Omega; X) is isometrically isomorphic to the space of X-valued countably additive measures of bounded variation on the Borel \sigma-algebra of \Omega, via the pairing \langle f, m \rangle = \int_\Omega f(\omega) \, dm(\omega). This theorem, originally proved by I. Singer in 1957, relies on the Hahn–Banach theorem and scalar Riesz–Markov representations to decompose vector measures.^[29] In the broader context of locally convex topological vector spaces, representations can be formulated for functionals on spaces of continuous functions taking values in such spaces. A key result by Diestel and Uhl (1968) establishes that for a compact Hausdorff space H, locally convex spaces E and F (with F Hausdorff), and the space C(H, E) of continuous E-valued functions on H with the topology of uniform convergence on compact subsets, every continuous linear functional T: C(H, E) \to F admits a representation T(f) = \int_H f(h) \, dK(h) involving a F-valued kernel K of bounded variation, under suitable regularity conditions. This extends the scalar and Banach-valued cases by leveraging the Mackey topology and Pettis integrability for vector measures. Such theorems underpin duality theory in more abstract settings, including non-normable spaces.^[30]

References

[1]
[PDF] hilbert spaces and the riesz representation theorem - UChicago Math
Abstract. The Riesz representation theorem is a powerful result in the theory of Hilbert spaces which classifies continuous linear functionals in terms of ...Missing: authoritative sources
[2]
[PDF] the riesz-markov-kakutani representation theorem - UChicago Math
Sep 23, 2023 · The RMK Theorem was first proved for the special case where X = [0, 1] in 1909 by Frigyes Riesz; cf. [12]. In 1938, Andrey Markov published work ...Missing: authoritative sources
[3]
Hilbert Space -- from Wolfram MathWorld
A Hilbert space is a vector space H with an inner product such that the norm defined by |f|=sqrt( ) turns H into a complete metric space.Missing: properties | Show results with:properties
[4]
[PDF] Chapter 6: Hilbert Spaces - UC Davis Math
Definition 6.2 A Hilbert space is a complete inner product space. In ... properties of infinite- dimensional Banach spaces in comparison with Hilbert spaces.
[5]
David Hilbert (1862 - 1943) - Biography - MacTutor
David Hilbert ... This work also established the basis for his work on infinite-dimensional space, later called Hilbert space, a concept that is useful in ...
[6]
[PDF] an introduction to functional analysis - UChicago Math
Aug 7, 2010 · A Hilbert Space is a complete inner product space. We now state ... Dual Spaces and the Riesz Representation Theorem. In this section we ...
[7]
[PDF] Analysis Preliminary Exam Workshop: Hilbert Spaces - UC Davis Math
The dual space of every Hilbert space is isomorphic (real case) or anti- isomorphic (complex case) to the Hilbert space. Theorem 11 (Riesz representation).
[8]
[PDF] Hilbert-Spaces.pdf - UCSD Math
Hilbert Spaces Basics. Definition 12.1. Let H be a complex vector space. An inner product on H is a function, h·,·i : H × H → C, such that (1) hax + by,zi = ...
[9]
[PDF] Bounded Linear Operators on a Hilbert Space - UC Davis Math
8.2 The dual of a Hilbert space. A linear functional on a complex Hilbert space H is a linear map from H to C. A. linear functional ϕ is bounded, or continuous ...
[10]
[PDF] brief note on complex hilbert spaces with hermitian inner-products
Theorem 3.1 (Riesz Representation Theorem). Let ` : H → C be a continuous linear functional. There exists a unique g ∈ H so that `(f) = hf,gi ...
[11]
[PDF] 1. Hilbert spaces A complex Hilbert space H is a complete normed ...
Theorem 1.6 (Riesz representation theorem). The space of linear functionals is isomorphic, as a normed. Page 4. space, to H itself, i.e. every continuous linear ...
[12]
https://math.jhu.edu/~lindblad/632/riesz.pdf
[13]
245B, notes 5: Hilbert spaces | What's new - Terry Tao
Jan 17, 2009 · The Riesz representation theorem for Hilbert spaces gives a converse: Theorem 1. (Riesz representation theorem for Hilbert spaces) Let H be ...
[14]
[PDF] The Riesz Representation Theorem
The following is called the Riesz Representation Theorem: Theorem 1 If T is a bounded linear functional on a Hilbert space H then there exists some g ∈ H ...Missing: paper | Show results with:paper
[15]
[PDF] Frames for Undergraduates Deguang Han Keri Kornelson David ...
Theorem 2.6 (Riesz Representation Theorem). Let Λ be a lin- ear functional on a finite-dimensional Hilbert space H. (In other words, Λ is a linear operator ...
[16]
[PDF] Math2040 Tutorial 9
• Riesz representation theorem: if ϕ ∈ L(V,F), there exists a unique u ∈ V such that ϕ(v) = hv, ui, where u is given by u = ϕ(e1)e1 + ··· + ϕ(en)en with ...
[17]
7.2 The Riesz representation theorem - Ryan Tully-Doyle
The Riesz representation theorem tells us essentially everything about the continuous linear functionals on a Hilbert space.Missing: original statement
[18]
[PDF] FUNCTIONAL ANALYSIS | Second Edition Walter Rudin
Rudin, Walter, (date). Functional analysis/Walter Rudin.-2nd ed. p. em. -(international series in pure and applied mathematics).
[19]
[PDF] Functional Analysis Lecture Notes - Michigan State University
Theorem 17.3. Every Hilbert space is reflexive. Proof. In this case X. 0 = X by Riesz-Fréchet. D. For Lp spaces we have. Theorem 17.4. Let X, µ be a finite ...
[20]
[PDF] Adjoint operators - MTL 411: Functional Analysis
Recall, Riesz representation theorem: Suppose f is a continuous linear functional on a. Hilbert space H. Then there exists a unique z in H such that f(x) ...
[21]
[PDF] ADJOINT OPERATORS Consider a Hilbert space X over a field F ...
We will see that the existence of so-called adjoints is guaranteed by Riesz' representation theorem. Theorem 1 (Adjoint operator). ... Let T be a bounded linear ...
[22]
[PDF] functional analysis lecture notes: adjoints in hilbert spaces
By the Riesz. Representation Theorem, there exists a unique vector h ∈ H such that. (Ax, y) = Lx = (x, h). Define A∗y = h. Verify that this map A∗ is ...
[23]
[PDF] Theory of Reproducing Kernels Author(s): N. Aronszajn Source
Theory of Reproducing Kernels. Author(s): N. Aronszajn. Source: Transactions of the American Mathematical Society, Vol. 68, No. 3 (May, 1950), pp. 337-404.
[24]
[PDF] LECTURE 4 — 09/28/2020 THE FOURIER TRANSFORM 1. The ...
Sep 28, 2020 · By Riesz representation theorem,. Fψ is represented by an L2 function. Moreover, the same formula above also implies that as an L2-function ...
[25]
[PDF] harmonic analysis on lca groups - UChicago Math
Aug 28, 2017 · functional, recovering the measure via the Riesz Representation Theorem. ... We shall conclude with a proof of the Plancherel Theorem. Theorem ...
[26]
The spectral theorem and its converses for unbounded symmetric ...
Dec 20, 2011 · ... spectral theorem for finite-dimensional self ... Riesz representation theorem for Hilbert spaces and locate a unique vector ...
[27]
[PDF] Representation Theory and Quantum Mechanics
Theorem 1.12 (The Riesz Representation Theorem). If ϕ. ∗ is a bounded linear functional of H, then there is a unique ϕ ∈ H such that. ϕ. ∗. (ψ) = ⟨ϕ|ψ⟩ for ...
[28]
[PDF] Chapter 3 Mathematical Formalism of Quantum Mechanics
up Riesz's representation theorem, which gives the mathematically exact justification. 7The last property is called reflexivity of the Hilbert space. Page 5 ...
[29]
[PDF] Mathematical surprises and Dirac's formalism in quantum mechanics
In section 3, we will define Dirac's notation more precisely while recalling some mathe- matical facts and evoking the historical development of quantum ...<|separator|>
[30]
[PDF] The Dirac Approach to Quantum Theory - San Bernardino
The Riesz theorem works for all finite dimensional and some infinite dimensional vector spaces. Note that the space V must be an inner product space, because ...
[31]
a simple proof of singer's representation theorem
Let Ω be a compact Hausdorff space and X a Banach space. Singer's theorem states that under the dual pairing (f, m) 7→. R hf, dmi, the dual space of C(Ω; X) is ...
[32]
A RIESZ REPRESENTATION THEOREM IN THE SETTING OF ...
Let H be a compact Hausdorff space, let E and F be locally convex topological vector spaces over the real or complex field where Fis Hausdorff. Let C(77, E) be ...