A linear Operator is a map satisfying

In the above:

Operators are a generalisation of matrices: They act on Vector Spaces linearly, without being attached to any specific basis. When working with finite Vector Spaces, the choice of a basis allows us to represent the Operator as a matrix. This won’t be always possible: Operators in infinite dimensional spaces do not always have a Matrix representation

Relations

The diagram below aims to make relations between Operators and their Matrix representation memorable. An arrow means an implication, and a label states the extra hypothesis needed. Operators act on a complex Hilbert Space , and the conventions for Symmetric, Self-Adjoint and Hermitian are those of In finite dimensional spaces, hermitian = symmetric = self-adjoint: symmetric means on a dense domain, self-adjoint means including domains, and Hermitian means symmetric and defined on all of .

stateDiagram
	Hermitian: Hermitian
	Positive: Positive
	Bounded: Bounded
	Normal: Normal
	RealEig: Real eigenvalues
	RealSpec: Real spectrum
	CircleSpec: Spectrum on the unit circle
	SelfAdjoint: Self Adjoint
	Spectral: Spectral decomposition by a projection valued measure
	Eigenbasis: Orthonormal basis of eigenvectors
	Symmetric: Symmetric
	Unitary: Unitary
	Involutory: Involutory
	Hermitian --> SelfAdjoint
	Hermitian --> Bounded : Hellinger Toeplitz
	SelfAdjoint --> Symmetric
	Symmetric --> Hermitian : If defined on all of H
	Symmetric --> RealEig
	SelfAdjoint --> RealSpec
	SelfAdjoint --> Spectral
	Positive --> Hermitian : If bounded and everywhere defined
	Hermitian --> Normal
	Unitary --> Normal
	Unitary --> Bounded
	Unitary --> CircleSpec
	Normal --> Spectral : If bounded
	Normal --> Eigenbasis : If compact
	Involutory --> Unitary : If self adjoint
	Unitary --> Involutory : If self adjoint

Remarks:

  • In finite dimensions every operator is bounded, compact and everywhere defined, so Symmetric, Self-Adjoint and Hermitian coincide, and every normal matrix () is unitarily diagonalisable (Spectral theorem).
  • A symmetric operator always has real eigenvalues (), but its spectrum need not be real unless it is self-adjoint.
  • Positive, for all , implies symmetric on a complex space by the polarisation identity. Positive operators have a unique positive square root (Positive operators have a unique positive square root).
  • An involutory operator () need not be unitary: squares to but is not unitary. An operator with any two of the properties self-adjoint, unitary and involutory has the third.

See more details on:

Properties