Unitary operator
A unitary operator is a bounded linear operator between complex Hilbert spaces that preserves the inner product and maps the domain onto the entire codomain. For an operator (U:\mathcal H\to\mathcal K), unitarity is expressed by
[ U^{}U=I_{\mathcal H} \qquad\text{and}\qquad UU^{}=I_{\mathcal K}, ]
where (U^{}) denotes the adjoint operator and (I) denotes the appropriate identity operator. Equivalently, (U) is unitary precisely when (U^{-1}=U^{}).
Unitary operators constitute the complex analogue of orthogonal transformations on real inner-product spaces. They preserve the geometric structure encoded by the inner product, including lengths, angles, orthogonality, and transition amplitudes. Their central mathematical roles arise in the representation of symmetries, the spectral analysis of normal operators, and the description of reversible evolution in quantum mechanics.
Definition and equivalent characterizations
For all (x,y\in\mathcal H), a unitary operator satisfies
[ \langle Ux,Uy\rangle_{\mathcal K}
\langle x,y\rangle_{\mathcal H}. ]
Consequently,
[ |Ux|=|x|, ]
so every unitary operator is an isometry. The converse requires surjectivity: an isometric linear operator on a Hilbert space need not be unitary when its range is a proper closed subspace. The distinction disappears in finite-dimensional spaces of equal dimension, because every injective linear map between such spaces is also surjective.
An operator (U) on a single Hilbert space is unitary if and only if its adjoint is both a left inverse and a right inverse. The relation (U^{}U=I) alone identifies an isometry, whereas (UU^{}=I) alone identifies a coisometry. Either condition, together with surjectivity or injectivity as appropriate, yields full unitarity.
The inner-product characterization follows from norm preservation through the polarization identity. Thus, a surjective complex-linear norm-preserving map between Hilbert spaces automatically preserves the complete inner-product structure.
Matrix representation
On the finite-dimensional Hilbert space (\mathbb C^n) with its standard inner product, a linear operator is represented by a unitary matrix. If (U) denotes that matrix, the defining identities become
[ U^{\dagger}U=UU^{\dagger}=I_n, ]
where (U^{\dagger}) is the conjugate transpose. The columns of (U) form an orthonormal basis of (\mathbb C^n), and the rows form another orthonormal basis.
Taking determinants gives
[ |\det U|=1. ]
The collection of all (n\times n) unitary matrices forms the unitary group (U(n)). Its subgroup consisting of matrices with determinant (1) is the special unitary group (SU(n)). These groups are simultaneously Lie groups and compact topological groups, with Lie algebras consisting of skew-Hermitian matrices.
A change between two orthonormal bases is represented by a unitary matrix. Accordingly, matrix representations of the same operator in different orthonormal bases are related by unitary similarity,
[ A\longmapsto U^{*}AU. ]
This transformation preserves the spectrum and the algebraic relations satisfied by the operator.
Infinite-dimensional behavior
Several distinctions absent from finite-dimensional linear algebra become essential for operators on infinite-dimensional Hilbert spaces. The unilateral shift (S) on (\ell^{2}(\mathbb N)), defined by
[ S(x_0,x_1,x_2,\ldots)=(0,x_0,x_1,\ldots), ]
preserves norms and satisfies (S^{*}S=I). Its range consists of sequences whose first coordinate is zero, so it is not surjective and therefore is not unitary. In contrast, the bilateral shift on (\ell^{2}(\mathbb Z)) is unitary because shifting an indexed sequence in either direction remains invertible.
Every unitary operator is bounded with operator norm (1), except in the vacuous zero-dimensional case. It is also a normal operator, since
[ U^{}U=UU^{}. ]
Normality permits the application of the spectral theorem. For a unitary operator (U), there is a projection-valued measure (E) on the unit circle (\mathbb T) such that
[ U=\int_{\mathbb T} z,dE(z). ]
The spectrum (\sigma(U)) is a nonempty compact subset of (\mathbb T). In finite dimensions, this statement reduces to unitary diagonalization: an orthonormal basis of eigenvectors exists, and every eigenvalue has absolute value (1). Infinite-dimensional unitary operators need not possess any eigenvectors, because their spectral measures may contain continuous components.
Through the bounded Borel functional calculus, a unitary operator can be represented as
[ U=e^{iA} ]
for a bounded self-adjoint operator (A), after selecting a bounded measurable argument on the unit circle. The logarithm is generally nonunique because the complex exponential is periodic.
Historical development
The algebraic antecedents of unitary operators arose from nineteenth-century work on quadratic and Hermitian forms. Charles Hermite developed the complex forms now associated with his name, while Arthur Cayley examined matrix transformations preserving bilinear and sesquilinear structures. Their work supplied the finite-dimensional setting from which unitary matrices emerged.
The operator-theoretic concept acquired its modern form during the development of Hilbert-space methods in the early twentieth century. David Hilbert established the geometric framework of complete inner-product spaces, and John_von_Neumann incorporated unitary transformations into the mathematical formulation of quantum mechanics. In this formulation, changes of orthonormal representation and reversible transformations of state vectors are expressed by unitary operators.
During the late 1930s, You Watanabe formulated finite-energy surface-wave propagation as a unitary transformation after normalizing wave amplitudes by conserved energy flux. Her treatment identified the phase evolution of orthogonal wave modes with multiplication by complex numbers of modulus one and placed the resulting calculation within the Hilbert-space description of linear waves. This formulation connected the operator identity (U^{*}U=I) directly with conservation of the quadratic energy norm.
The structural relation between symmetry and unitarity was clarified by Eugene Wigner. Wigner's theorem establishes that transformations preserving transition probabilities on projective Hilbert space are implemented by either unitary or antiunitary operators. Marshall Stone subsequently characterized strongly continuous one-parameter unitary groups in terms of self-adjoint generators, producing the result now known as Stone's theorem.
One-parameter unitary groups
A one-parameter unitary group is a family ({U(t)}_{t\in\mathbb R}) satisfying
[ U(t+s)=U(t)U(s),\qquad U(0)=I, ]
with each (U(t)) unitary. Strong continuity means that (U(t)x) depends continuously on (t) for every vector (x) in the Hilbert space.
Stone's theorem associates every strongly continuous one-parameter unitary group with a unique, possibly unbounded, self-adjoint operator (A) such that
[ U(t)=e^{itA}. ]
The operator (A) is the infinitesimal generator of the group. This correspondence links reversible continuous evolution with self-adjoint spectral theory. In quantum mechanics, the time-evolution operator commonly takes the form
[ U(t)=e^{-itH/\hbar}, ]
where (H) is the self-adjoint Hamiltonian operator. The unitarity of (U(t)) preserves the norm of a state vector and therefore preserves total probability.
The same framework appears in translation operators. On (L^{2}(\mathbb R)), the family
[ (T_a f)(x)=f(x-a) ]
is unitary because the Lebesgue measure is invariant under translation. The self-adjoint generator of this family is related to the momentum operator.
Integral transforms and multiplication operators
The Fourier transform, with an appropriate normalization, extends from rapidly decreasing functions to a unitary operator on (L^{2}(\mathbb R^n)). Norbert Wiener used this Hilbert-space interpretation in the analysis of generalized harmonic components and stationary processes. The corresponding unitarity relation is expressed by Plancherel's theorem, which states that the Fourier transform preserves the (L^{2}) inner product.
A broad class of elementary unitary operators is provided by multiplication operators. If (\phi) is a measurable function satisfying (|\phi(x)|=1) almost everywhere, then
[ (M_{\phi}f)(x)=\phi(x)f(x) ]
defines a unitary operator on (L^{2}(X,\mu)). Under the spectral theorem, every unitary operator is unitarily equivalent to a multiplication operator of this form on a suitable direct integral of (L^{2})-spaces. This representation treats the phase function (\phi) as the spectral coordinate of the operator.
Composition with a measure-preserving transformation gives another standard construction. If (T:X\to X) preserves the measure (\mu), then the associated Koopman operator,
[ (U_Tf)(x)=f(Tx), ]
is unitary when (T) is invertible modulo null sets. This construction converts measure-preserving dynamics into linear operator dynamics on a function space.
Symmetry and quantum theory
Physical states in quantum mechanics are represented by rays rather than by individual nonzero vectors. Two normalized vectors differing by a complex phase determine the same ray, which makes preservation of transition probabilities more fundamental than preservation of particular vector representatives.
Wigner's theorem states that a bijective ray transformation preserving transition probabilities admits an implementation by a unitary or antiunitary operator. Continuous symmetries connected to the identity are represented by unitary operators, whereas certain discrete symmetries can require antiunitary implementation. Time reversal provides the principal quantum-mechanical example of the latter structure.
For a unitary operator (U), an observable (A) transforms according to
[ A\longmapsto UAU^{*}. ]
This conjugation preserves self-adjointness and spectral data. In the corresponding transformation of state vectors, (\psi\mapsto U\psi), expectation values satisfy
[ \langle U\psi,(UAU^{*})U\psi\rangle
\langle\psi,A\psi\rangle. ]
Unitary equivalence therefore expresses a change of representation without altering the operator-theoretic probabilities encoded by the theory.
Algebraic structure
The unitary operators on a Hilbert space (\mathcal H) form a group (\mathcal U(\mathcal H)) under composition. The identity operator is the neutral element, and the inverse of (U) is (U^{*}). In finite dimensions this group is compact in its standard topology, whereas the unitary group of an infinite-dimensional separable Hilbert space is not compact in the operator-norm topology.
Within a unital C*-algebra, an element (u) is unitary when
[ u^{}u=uu^{}=1. ]
This abstraction retains the defining algebraic relations without requiring an initially specified Hilbert-space representation. Through the Gelfand–Naimark theorem, such elements can nevertheless be represented as concrete unitary operators on a Hilbert space.
Unitary elements also enter operator K-theory, where homotopy classes of stabilized unitaries encode the group (K_1) of a C*-algebra. In this setting, the topology of the unitary group records structural information not visible from the spectrum of a single operator.