Operator theory studies transformations acting on function spaces and other infinite-dimensional vector spaces. Matrices act on finite-dimensional vectors. Operators generalise that idea to spaces whose “vectors” may themselves be functions, sequences, signals or quantum states.
The subject sits inside functional analysis but deserves its own machinery. It asks which operators are bounded, invertible, compact, self-adjoint or normal; how spectra generalise eigenvalues; how differential equations become operator equations; and how time evolution can be generated by operators.
Series route: Mathematics Learning Hub → How Mathematics Works → Operator Theory. Useful foundations include Functional Analysis, Linear Algebra, Harmonic Analysis and Partial Differential Equations.
1. Operators are transformations between spaces
An operator T maps vectors in one space to vectors in another.
In functional analysis, the vectors may be functions. Differentiation, integration, shifts and multiplication by a fixed function are all natural examples.
2. Linear operators preserve superposition
A linear operator satisfies T(ax+by)=aT(x)+bT(y).
This makes infinite-dimensional operator theory a continuation of matrix algebra rather than an unrelated subject.
3. Domains matter in infinite dimensions
An operator is not specified by a formula alone. Its domain and codomain are part of the object.
Differentiation on continuous functions, differentiable functions and Sobolev spaces creates different operator settings even when the symbolic formula looks identical.
4. Bounded linear operators are continuous
A linear operator T between normed spaces is bounded when ||Tx||≤C||x|| for some constant C and every x in its domain.
For linear operators defined everywhere between normed spaces, boundedness is equivalent to continuity.
5. The operator norm measures maximum amplification
The operator norm is ||T||=sup_{||x||=1}||Tx|| for bounded operators.
It generalises matrix norms and quantifies worst-case amplification relative to the chosen vector norm.
6. Composition creates an operator algebra
Compatible operators can be composed, added and multiplied by scalars.
Bounded operators on a Banach space form a Banach algebra under the operator norm.
7. Invertibility must include bounded inverse
An algebraically bijective bounded operator between Banach spaces has a bounded inverse by the bounded inverse theorem.
Completeness turns algebraic invertibility into stable analytical invertibility.
8. Eigenvalues are only part of the spectrum
For an operator T, λ is an eigenvalue when Tx=λx for some nonzero x.
In infinite dimensions, T−λI can fail to be suitably invertible even when no nonzero eigenvector exists. The spectrum is therefore broader than the set of eigenvalues.
9. The resolvent is where inversion works
The resolvent set contains λ for which T−λI has a bounded inverse defined on the whole space.
The spectrum is its complement.
10. Spectrum is compact for bounded operators
For a bounded operator on a complex Banach space, the spectrum is nonempty and compact.
This is a distinctly infinite-dimensional theorem: even operators without ordinary eigenvalues cannot have an empty spectrum.
11. Spectral radius measures asymptotic operator growth
The spectral radius r(T) is the largest modulus of spectral values.
It also satisfies the spectral-radius formula r(T)=lim ||T^n||^{1/n} in the standard Banach-algebra setting.
12. Adjoint operators reverse inner-product action
On a Hilbert space, the adjoint T* satisfies <Tx,y>=<x,T*y>.
The adjoint generalises the conjugate transpose of a matrix.
13. Self-adjoint operators generalise symmetric matrices
A densely defined operator is self-adjoint when T=T* with matching domains in the unbounded case.
Self-adjoint operators have real spectrum and are central to quantum observables and PDE theory.
14. Normal operators commute with their adjoints
T is normal when T*T=TT*.
Unitary and self-adjoint operators are important subclasses. Normality is the correct setting for the Hilbert-space spectral theorem.
15. Unitary operators preserve inner products
A unitary operator U satisfies U*U=UU*=I.
It preserves norms, angles and probabilities encoded by Hilbert-space inner products.
16. Projection operators separate subspaces
An orthogonal projection P satisfies P²=P and P*=P.
It extracts the component lying in a closed subspace and underlies least-squares approximation.
17. Compact operators imitate finite-dimensional behaviour
A compact operator maps bounded sets into relatively compact sets.
Nonzero spectral values of compact operators are eigenvalues with finite multiplicity, and accumulation can occur only at zero under standard infinite-dimensional settings.
18. Integral operators are important compact examples
An operator of the form Tf(x)=∫K(x,y)f(y)dy can be compact under suitable assumptions on the kernel K and function spaces.
Such operators arise in boundary-value problems and inverse problems.
19. Fredholm theory measures controlled failure of invertibility
A Fredholm operator has finite-dimensional kernel and cokernel and closed range.
Its index dim ker T−dim coker T is stable under compact perturbations.
20. The Fredholm alternative links solvability to adjoint kernels
For suitable compact-perturbation equations, either the homogeneous equation has only the trivial solution and the inhomogeneous problem is uniquely solvable, or nontrivial homogeneous solutions create compatibility conditions.
Solvability becomes an orthogonality question involving the adjoint.
21. Unbounded operators are unavoidable
Differential operators such as d/dx and the Laplacian are usually unbounded on natural Hilbert spaces.
They therefore require careful domain specification, closedness and dense definition.
22. Closed operators preserve limits of graph data
An operator is closed when convergence x_n→x together with Tx_n→y implies x is in the domain and Tx=y.
Closedness replaces boundedness as a crucial stability property for unbounded operators.
23. Closable operators admit a smallest closed extension
A densely defined operator may fail to be closed but still have a closure.
This allows initially defined differential expressions to be completed into analytically useful operators.
24. Spectral theorem replaces diagonalisation
Finite-dimensional normal matrices can be diagonalised by unitary matrices.
In infinite dimensions, normal operators are represented through projection-valued measures and spectral integrals.
Diagonalisation becomes a continuum of spectral components rather than necessarily a discrete list of eigenvectors.
25. Functional calculus applies functions to operators
Given appropriate spectral structure, one can define f(T) for scalar functions f.
Square roots, exponentials and spectral projections become operator constructions rather than formal substitutions.
26. Operator semigroups describe continuous time evolution
A strongly continuous semigroup T(t) satisfies T(0)=I and T(t+s)=T(t)T(s), with continuous dependence on t.
Its infinitesimal generator is often an unbounded operator.
27. Evolution PDEs become abstract differential equations
An equation u_t=Au can be treated as an ordinary differential equation in an infinite-dimensional function space.
Semigroup theory determines when u(t)=T(t)u_0 is well defined and stable.
28. Resolvent estimates control semigroups
The location and size of the resolvent (λI−A)^{-1} help determine whether A generates a stable semigroup.
Frequency-domain information can therefore control time-domain evolution.
29. Pseudospectra reveal nonnormal sensitivity
For nonnormal operators, tiny perturbations can move eigenvalues dramatically and transient growth can be large even when the spectrum looks stable.
Pseudospectra track where the resolvent becomes large and reveal this hidden sensitivity.
30. Operator theory enters quantum mechanics directly
Quantum states live in Hilbert spaces and observables are modelled by self-adjoint operators.
Spectral measures connect operators to possible measurement values, while unitary groups describe isolated time evolution.
Continue through Mathematical Physics for the wider physical setting.
31. Operator theory enters harmonic analysis
Fourier multipliers, singular integral operators and convolution operators are studied through their boundedness on function spaces.
Harmonic analysis therefore supplies operators whose behaviour is understood through symmetry and frequency.
32. Operator theory enters numerical analysis
Discretising an operator produces matrices that approximate its action.
Convergence requires more than matrix computation: one must show the discrete operators approximate the continuous operator in an appropriate sense.
33. A worked mechanism: multiplication operator
On L²([0,1]), define Tf(x)=x f(x).
- |x|≤1, so ||Tf||₂≤||f||₂ and ||T||≤1.
- Functions concentrated near x=1 show ||T||=1.
- No nonzero L² function is supported only at a single point λ, so ordinary eigenvectors are absent for most λ.
- Yet the spectrum is the interval [0,1].
This example shows why infinite-dimensional spectra extend beyond eigenvalue lists.
34. Common operator-theory failure modes
- Formula-only operators: omitting domain and codomain.
- Spectrum=eigenvalues confusion: importing finite-dimensional intuition.
- Bounded/unbounded collapse: applying bounded-operator theorems to differential operators without domain control.
- Adjoint-domain blindness: treating formal integration by parts as proof of self-adjointness.
- Discretisation certainty: assuming matrix eigenvalues automatically approximate operator spectra.
- Nonnormal blindness: judging transient stability from eigenvalues alone.
35. Operator theory as a mathematical machine
Function/State Space → Operator and Domain → Boundedness/Closedness → Spectrum/Adjoint → Structural Class → Semigroup or Functional Calculus → PDE/Quantum/Numerical Interpretation.
36. What mastery looks like
- treat domains as part of operator identity;
- use operator norms and boundedness correctly;
- distinguish spectrum from eigenvalues;
- work with adjoint, self-adjoint, normal and unitary classes;
- understand compact operators and Fredholm structure;
- handle unbounded operators through closedness and dense domains;
- interpret the spectral theorem as infinite-dimensional diagonalisation;
- connect generators and semigroups to evolution equations;
- recognise pseudospectral sensitivity in nonnormal systems.
37. Conclusion
Operator theory works by extending matrix thinking into infinite-dimensional spaces without pretending that every finite-dimensional theorem survives unchanged. Bounded operators preserve continuity. Unbounded operators require domains and closedness. Spectra replace simple eigenvalue lists. Self-adjointness supports physical observables. Compactness recovers finite-dimensional features. Semigroups turn operators into time evolution.
Functional analysis builds the spaces. Operator theory studies the transformations that make those spaces calculate, evolve and interact.
How Mathematics Works | Batch 11
- How Mathematics Works | Differential Topology
- How Mathematics Works | Riemannian Geometry
- Operator Theory — this article
- How Mathematics Works | Ergodic Theory
Return to the Mathematics Learning Hub or the How Mathematics Works root.