1 Definition and basic properties
A Hilbert space is a vector space equipped with an inner product and made complete by the norm induced from that inner product. This combination gives the space both algebraic structure and geometric meaning. In practice, Hilbert spaces extend familiar Euclidean ideas such as distance, angle, and orthogonality to settings with infinitely many dimensions.
The subject begins with the interplay between vector addition, scalar multiplication, and inner products. Once a suitable norm is defined, completeness becomes the crucial extra requirement. Many naturally occurring spaces fail to be Hilbert spaces because they are not complete, while others become especially useful precisely because all Cauchy sequences have limits inside the space.
1.1 Vector spaces and inner products
A Hilbert space is first a vector space over the real or complex numbers. The inner product assigns to each pair of vectors a scalar and is required to be linear in one argument, conjugate symmetric in the complex case, and positive definite. These axioms encode geometric information in algebraic form.
The inner product allows one to measure how closely two vectors align and to define orthogonality. It also determines the angle between vectors in a way that generalizes the dot product from elementary geometry. Because of this structure, many calculations in Hilbert spaces resemble computations in ordinary finite-dimensional geometry.
1.2 Norms induced by inner products
Every inner product defines a norm by taking the square root of the inner product of a vector with itself. This norm measures length and leads to a distance function between vectors. The resulting metric makes it possible to discuss convergence, continuity, and completeness.
Not every norm comes from an inner product. In Hilbert spaces, however, the norm and inner product are tightly linked, and the norm satisfies the parallelogram law. This connection is one reason Hilbert spaces have exceptionally rich geometric behavior.
1.3 Completeness
Completeness means that every Cauchy sequence converges to a limit in the space. In a Hilbert space, this requirement prevents “missing points” that would otherwise lie at the ends of convergent processes. It ensures that limiting arguments used in analysis remain valid within the same space.
Completeness is essential for many fundamental results, including projection theorems and representation theorems. Without it, orthogonal approximations may exist only as limiting objects outside the space. The completeness condition is what separates Hilbert spaces from more general inner product spaces.
1.4 Examples of Hilbert spaces
Hilbert spaces appear in many standard settings, including finite-dimensional geometry, sequence spaces, and function spaces. These examples show that the abstract definition captures both familiar and highly sophisticated situations. Despite their diversity, all share the same basic inner-product geometry.
1.4.1 Finite-dimensional Euclidean spaces
The most familiar examples are the Euclidean spaces \(\mathbb{R}^n\) and \(\mathbb{C}^n\) with the standard dot product or Hermitian inner product. In finite dimensions, all norms are equivalent, and completeness is automatic. These spaces provide the model for much of Hilbert space geometry.
1.4.2 Sequence spaces
A central infinite-dimensional example is \(\ell^2\), the space of square-summable sequences. Its inner product is defined by summing products of corresponding coordinates, and its norm measures the square root of the total energy in the sequence. This space is complete and plays a major role in analysis and applied mathematics.
1.4.3 Function spaces
Function spaces can also be Hilbert spaces when equipped with an appropriate inner product. The space \(L^2\) of square-integrable functions is the most important example. It is widely used because many physical and analytic quantities are naturally expressed as integrals of squares.
2 Geometry of Hilbert spaces
The geometry of a Hilbert space generalizes the geometry of ordinary Euclidean space. Concepts such as orthogonality, projections, and right triangles remain meaningful and powerful even in infinite dimensions. This geometric viewpoint is one of the main reasons Hilbert spaces are so useful.
2.1 Orthogonality
Two vectors are orthogonal if their inner product is zero. This notion extends the idea of perpendicularity from plane geometry. Orthogonality is fundamental in decomposition, approximation, and series expansions.
Orthogonal vectors interact in especially simple ways. Their contributions to lengths add independently, and their geometric independence makes them ideal building blocks for analysis. Many constructions in Hilbert spaces aim to convert complicated vectors into sums of orthogonal pieces.
2.2 Orthogonal complements
Given a subspace, its orthogonal complement consists of all vectors orthogonal to every vector in that subspace. This set is itself a subspace and captures the directions completely independent of the original space. Orthogonal complements are central to understanding how a Hilbert space splits into parts.
The orthogonal complement of a closed subspace has particularly strong properties. It often provides the natural remainder term in decomposition formulas. In many problems, identifying the orthogonal complement is the key to finding the best approximation.
2.3 Projections onto subspaces
A projection sends a vector to its nearest point in a subspace. In Hilbert spaces, projections onto closed subspaces exist and are unique. This fact gives a precise notion of “best approximation” in infinite-dimensional settings.
Orthogonal projections are especially important because the error vector is orthogonal to the subspace. This orthogonality characterizes the projection and simplifies many computations. Projections are used in approximation theory, operator theory, and numerical methods.
2.4 Pythagorean theorem
If two vectors are orthogonal, the square of the norm of their sum equals the sum of the squares of their norms. This is the Pythagorean theorem in Hilbert space form. It expresses the independence of orthogonal components.
The theorem extends to orthogonal decompositions involving more than two terms. It underlies many estimates and stability arguments in analysis. In a Hilbert space, this familiar geometric law becomes a central algebraic tool.
2.5 Parallelogram law
The parallelogram law relates the norms of two vectors, their sum, and their difference. It is a characteristic identity of inner product spaces. In fact, a norm satisfying this law can be used to recover an inner product.
This identity shows that Hilbert space geometry is highly rigid. It distinguishes inner-product norms from more general norms and explains why many geometric arguments work so smoothly. The parallelogram law also links length measurements with the underlying bilinear structure.
3 Bases and expansions
One of the most useful features of Hilbert spaces is the availability of orthonormal systems and expansion formulas. These tools generalize coordinate systems from finite-dimensional linear algebra. They allow vectors to be studied through coefficients relative to a well-chosen family of basis vectors.
3.1 Orthonormal sets
An orthonormal set is a collection of vectors that are mutually orthogonal and each have norm one. Such sets are the Hilbert space analogue of perpendicular unit vectors. They provide the simplest possible coordinate directions.
Orthonormal sets are valuable because coefficients in expansions along them are easy to compute. They also minimize interference between components. In infinite-dimensional spaces, orthonormal sets can be finite or countably infinite.
3.2 Orthonormal bases
An orthonormal basis is an orthonormal set whose linear span is dense in the space. Every vector can then be approximated arbitrarily well by finite linear combinations of basis elements. In separable Hilbert spaces, orthonormal bases are often countable.
Unlike bases in finite-dimensional linear algebra, an orthonormal basis in a Hilbert space may require limits for full expansion. Still, the coefficients behave much like coordinates. This makes orthonormal bases one of the most powerful organizing principles in the subject.
3.3 Fourier series in Hilbert spaces
Fourier series are expansions of vectors in terms of an orthonormal basis, often consisting of trigonometric functions. In Hilbert space language, they are coordinate expansions relative to a chosen orthonormal system. The abstract theory generalizes classical Fourier analysis beyond periodic functions.
Such expansions can represent functions, signals, or other objects as sums of modes. The coefficients capture the contribution of each mode. This viewpoint is foundational in analysis, especially when solving differential equations or studying oscillatory behavior.
3.4 Parseval's identity
Parseval's identity states that, for an orthonormal basis, the squared norm of a vector equals the sum of the squares of its expansion coefficients. This is an exact energy conservation law in Hilbert space form. It ties together geometry and series expansion.
The identity shows that no information is lost when passing from a vector to its coordinates in an orthonormal basis. It is especially important in Fourier analysis, where it expresses equality between the total energy of a function and the energy of its spectral components. Parseval's identity is one of the clearest manifestations of Hilbert space structure.
3.5 Bessel's inequality
Bessel's inequality states that for any orthonormal set, the sum of the squares of the coefficients of a vector is bounded above by the square of its norm. This gives a universal estimate even when the orthonormal set is not a basis. It guarantees that coefficient sequences remain controlled.
The inequality is often used to prove convergence and to show that certain series cannot grow too large. When the orthonormal set is a basis, Bessel's inequality becomes an equality and turns into Parseval's identity. Thus it serves as a preliminary form of the full expansion theorem.
4 Subspaces and decomposition
Subspaces in a Hilbert space inherit much of the ambient structure, but only closed subspaces behave well with respect to limits and projections. Decomposition theorems describe how a space can be broken into complementary pieces. These results are essential in both abstract theory and practical computation.
4.1 Closed subspaces
A subspace is closed if it contains the limits of all convergent sequences taken within it. Closedness is necessary for many geometric constructions, including orthogonal projection. Without it, best approximations may fail to exist.
Closed subspaces are the natural setting for decomposition results. They ensure that the internal geometry of the subspace matches the topology of the whole space. Many important invariant subspaces in analysis are studied through this property.
4.2 Direct sums
A direct sum decomposes a space into parts that intersect trivially and combine to form the whole. In Hilbert spaces, the most important direct sums are orthogonal direct sums. These preserve geometry and make coordinate-like descriptions possible.
Direct sum decompositions simplify problems by separating independent components. They are used in operator theory, spectral analysis, and the study of closed subspaces. When the summands are orthogonal, norm computations become especially transparent.
4.3 Hilbert space decomposition theorem
The decomposition theorem states that a Hilbert space can be written as the orthogonal direct sum of a closed subspace and its orthogonal complement. This is a cornerstone of Hilbert space theory. It formalizes the idea that every vector has a unique projection onto a closed subspace plus a perpendicular remainder.
This theorem is the abstract form of geometric decomposition into parallel and perpendicular parts. It underlies least-squares approximation and many existence proofs. In applications, it provides the conceptual framework for solving equations by splitting them into simpler pieces.
4.4 Dense subspaces
A subspace is dense if its closure is the whole space. Dense subspaces are important because vectors in the entire space can be approximated arbitrarily well by elements of the subspace. Many useful function classes are dense in larger Hilbert spaces.
Dense subspaces often serve as manageable domains for operators or as starting points for constructing bases. They are especially helpful when exact formulas are available only on a simpler class of vectors. Completeness then allows the formulas to extend by continuity.
5 Linear operators on Hilbert spaces
Linear operators are the main objects studied on Hilbert spaces. They describe transformations of vectors and encode differential, integral, and algebraic processes. The inner-product structure allows operators to be analyzed through adjoints, symmetry, and spectral behavior.
5.1 Bounded linear operators
A bounded linear operator is one that does not enlarge vectors without limit relative to their norm. Boundedness is equivalent to continuity in normed spaces. Such operators are especially well behaved because they extend limits and preserve convergence.
In Hilbert spaces, bounded linear operators form a rich algebra with many structural properties. They are the natural class for much of operator theory. Boundedness is often the minimum requirement for a linear transformation to be analytically tractable.
5.2 Adjoint operators
The adjoint of an operator is defined by moving the operator from one side of the inner product to the other. It generalizes the transpose of a matrix and is uniquely determined when the operator is bounded. Adjoint operators reveal how a transformation interacts with the geometry of the space.
Adjoints are central to self-adjointness, unitarity, and spectral theory. They also make it possible to define normality and to study symmetry in an abstract setting. Many identities in Hilbert spaces are best expressed using adjoints.
5.3 Self-adjoint operators
An operator is self-adjoint if it equals its adjoint. This is the infinite-dimensional analogue of a symmetric matrix. Self-adjoint operators have especially important spectral properties and arise naturally in physics and differential equations.
Their eigenvalues are real when they exist, and their behavior is closely tied to measurable observables and quadratic forms. Self-adjointness often signals that an operator is compatible with the inner-product geometry. For this reason, it is one of the most studied properties in the subject.
5.4 Unitary operators
A unitary operator preserves inner products, and therefore lengths and angles. It is the Hilbert space analogue of a rotation or rigid motion. Unitary operators are reversible and have inverses equal to their adjoints.
Because they preserve geometry exactly, unitary operators are central in quantum theory and Fourier analysis. They transport orthonormal bases to orthonormal bases and maintain norms of vectors. This makes them ideal for describing changes of coordinates and symmetry transformations.
5.5 Compact operators
Compact operators send bounded sets to relatively compact sets, making them resemble finite-rank operators in many ways. They often arise from integral operators and approximation processes. Compactness introduces a discrete flavor into infinite-dimensional analysis.
These operators frequently have spectral properties that are easier to describe than those of general bounded operators. Their nonzero spectrum typically consists of isolated eigenvalues with finite multiplicity, accumulating only at zero. This behavior makes compact operators a key bridge between finite- and infinite-dimensional theory.
6 Spectral theory
Spectral theory studies the values and structures associated with operators, especially self-adjoint and compact ones. It generalizes the role of eigenvalues in matrix theory. In Hilbert spaces, spectral methods provide deep insight into operator behavior and function expansion.
6.1 Spectrum of an operator
The spectrum of an operator is the set of scalar values for which the operator fails to have a bounded inverse in the appropriate sense. It extends the notion of eigenvalues beyond finite-dimensional settings. The spectrum may contain continuous parts as well as isolated points.
Spectral information reveals how an operator behaves under perturbation and iteration. It also provides a natural classification of operators. In many applications, knowing the spectrum is more informative than listing eigenvalues alone.
6.2 Spectral theorem
The spectral theorem gives a powerful structural description of self-adjoint and some related operators. It says that such operators can be represented through multiplication operators or via a decomposition over spectral measures. This is one of the deepest results in Hilbert space theory.
The theorem generalizes diagonalization of symmetric matrices. It allows complicated operators to be studied through simpler building blocks indexed by spectral values. Many applications in quantum mechanics and differential equations rely on this representation.
6.3 Eigenvalues and eigenvectors
An eigenvector of an operator is a nonzero vector that is scaled, not rotated, by the operator. The corresponding scalar is the eigenvalue. These concepts capture invariant directions and are often the simplest way to understand an operator.
In Hilbert spaces, eigenvalues may form only part of the operator’s spectral picture. Nonetheless, when they exist, they often provide the most accessible information. Eigenvectors also furnish orthogonal expansions in favorable cases, linking operator theory with basis methods.
6.4 Functional calculus
Functional calculus allows one to apply functions to operators, especially self-adjoint ones, in a mathematically controlled way. It extends polynomial substitution to much broader classes of functions. This makes it possible to define expressions such as exponentials, square roots, and projections of operators.
The theory depends heavily on spectral decomposition. Once an operator has been understood spectrally, applying a function to it becomes natural. Functional calculus is indispensable in evolution equations, quantum theory, and analysis of semigroups.
7 Important theorems
Several foundational theorems in functional analysis are closely associated with Hilbert spaces and Banach spaces more generally. They provide extension, representation, and regularity results that support much of the theory. Together, they explain why linear methods are so effective in complete normed settings.
7.1 Riesz representation theorem
The Riesz representation theorem characterizes continuous linear functionals on a Hilbert space. It states that each such functional can be represented uniquely by taking an inner product with a fixed vector. This gives a concrete description of the dual space.
The theorem shows that Hilbert spaces are self-contained in a strong sense. Linear functionals become geometric objects rather than abstract maps. This is one of the defining features that makes Hilbert spaces so elegant and useful.
7.2 Hahn-Banach theorem
The Hahn-Banach theorem allows the extension of bounded linear functionals from a subspace to the whole space without increasing their norm. Although not specific to Hilbert spaces, it is fundamental in their broader analysis. It supports separation arguments and duality theory.
In Hilbert space contexts, the theorem is often used to prove existence results and to study convex sets. It complements the Riesz representation theorem by showing how functionals can be extended before being identified with vectors. Its influence reaches across much of functional analysis.
7.3 Open mapping theorem
The open mapping theorem states that a surjective bounded linear operator between Banach spaces maps open sets to open sets. In Hilbert spaces, this provides strong control over solvability of linear equations. It implies that bounded inverse behavior can be inferred from surjectivity.
This result is part of the foundational toolkit of modern analysis. It ensures that algebraic solvability has topological consequences. In practice, it helps justify the stability of solutions under perturbation.
7.4 Closed graph theorem
The closed graph theorem says that a linear operator between Banach spaces is bounded if its graph is closed. This offers a powerful criterion for continuity. In Hilbert spaces, it is frequently used to verify that naturally defined operators behave well.
The theorem is especially useful when an operator is defined indirectly, for example through limits or differential expressions. Rather than checking boundedness directly, one can examine the graph. This makes it a practical and elegant result in operator theory.
8 Special classes and constructions
Beyond the basic theory, Hilbert spaces admit several important refinements and constructions. These include countable-dimensional models, tensor products, duality properties, and spaces built from kernels. Each expands the range of phenomena that can be handled within the Hilbert framework.
8.1 Separable Hilbert spaces
A Hilbert space is separable if it contains a countable dense subset. Separable Hilbert spaces are especially manageable because they often admit countable orthonormal bases. Many spaces used in analysis, including \(L^2\) spaces on standard domains, are separable.
Separability is important for classification and for practical representation of elements. It allows many arguments to be reduced to sequences rather than uncountable families. This makes the space more accessible while preserving much of its infinite-dimensional richness.
8.2 Tensor products of Hilbert spaces
The tensor product of Hilbert spaces combines two spaces into a larger one that encodes interactions between them. It is used to represent composite systems and multilinear phenomena. The inner-product structure extends naturally to the completed tensor product.
Tensor products are essential in quantum theory, harmonic analysis, and operator theory. They allow separate systems to be studied jointly while retaining their individual structures. Many complex spaces arise as tensor products of simpler ones.
8.3 Dual spaces and reflexivity
The dual space consists of all continuous linear functionals on a Hilbert space. By the Riesz representation theorem, every functional corresponds to an inner product with a unique vector. This identification makes Hilbert spaces reflexive, meaning they naturally coincide with their double duals.
Reflexivity is a strong structural property. It supports compactness arguments and ensures that many limits can be interpreted within the space. In Hilbert spaces, duality is unusually transparent compared with general Banach spaces.
8.4 Reproducing kernel Hilbert spaces
A reproducing kernel Hilbert space is a Hilbert space of functions in which evaluation at each point is a continuous linear functional. Such spaces have a special kernel function that reproduces values from inner products. They appear in approximation theory, learning theory, and complex analysis.
The kernel encodes both geometry and function evaluation in a single object. This makes these spaces highly structured and computationally useful. Their theory connects Hilbert spaces with practical methods for interpolation and estimation.
9 Applications
Hilbert spaces provide a common language for many areas of mathematics and physics. Their combination of geometric clarity and analytic completeness makes them especially suited to problems involving expansions, operators, and energy methods. The following applications illustrate their broad reach.
9.1 Fourier analysis
Fourier analysis studies the decomposition of functions into oscillatory components. Hilbert spaces supply the natural setting for such decompositions, especially in \(L^2\) spaces. Orthogonal expansions and Parseval-type identities are central to the subject.
This framework clarifies convergence, approximation, and energy distribution among frequencies. Many classical results in Fourier analysis can be reformulated as statements about orthonormal bases and projections. The Hilbert space approach has become standard in modern analysis.
9.2 Partial differential equations
Many partial differential equations are studied by converting them into variational or operator equations in Hilbert spaces. This approach makes it possible to define weak solutions and to work with generalized derivatives. Energy estimates are often expressed using the inner product.
Hilbert space methods are especially effective for existence and uniqueness problems. They also help in understanding eigenvalue problems and boundary value problems. The geometry of the space often reflects the structure of the differential equation.
9.3 Quantum mechanics
In quantum mechanics, states are modeled using Hilbert spaces, and observables are represented by self-adjoint operators. The inner product encodes probabilities and transition amplitudes. Unitary operators describe time evolution and symmetry transformations.
This formalism captures superposition, measurement, and spectral decomposition in a rigorous mathematical language. Hilbert spaces provide the standard framework for nonrelativistic quantum theory. Their structure is well suited to the linear and probabilistic features of the subject.
9.4 Probability theory
Hilbert spaces arise in probability through spaces of square-integrable random variables. These spaces allow probabilistic quantities to be analyzed using inner products and orthogonality. Conditional expectation can often be interpreted as an orthogonal projection.
This perspective is useful for stochastic processes, estimation, and martingale theory. It connects randomness with geometry in a precise way. Many probabilistic inequalities and decompositions are naturally expressed in Hilbert space language.
9.5 Signal processing
Signal processing uses Hilbert spaces to represent signals as vectors and filters as operators. Orthogonal expansions help isolate frequencies, remove noise, and compress information. The language of inner products is especially useful for correlation and energy measurements.
Applications include communication theory, data analysis, and time-frequency methods. Fourier and wavelet techniques are often formulated in Hilbert space terms. The framework provides both theoretical clarity and computational tools.