Observables and Operators

An observable is a physical quantity together with a rule for its possible outcomes and their probabilities. In the standard theory of sharp measurements, that rule is encoded by a self-adjoint operator and its spectral projectors. An operator is the broader mathematical object: a linear map that may represent an observable, a transformation, a dynamical generator, a projector, or some other structure.

This chapter develops the operator language without identifying every operator with a measurable quantity. Its central organizing idea is spectral: eigenvalues and spectral sets label possible outcomes, projectors identify the corresponding subspaces, and the state assigns probabilities to those alternatives.

Required background. States and Representations supplies the coordinate-independent meaning of a quantum state.

Helpful background. Bra–Ket Notation supplies matrix-element notation, while Linear Maps reviews the underlying algebra.

What This Chapter Owns

This chapter is the canonical home for

• sharp observables and their distinction from apparatuses and measurement procedures;
• linear operators, adjoints, unitary operators, and projection operators in the core formalism;
• eigenvalues, eigenvectors, eigenspaces, and degeneracy;
• discrete and continuous spectra;
• the finite-dimensional spectral decomposition and its spectral-measure extension;
• the distinction between Hermitian matrices, symmetric operators, and self-adjoint operators;
• complete sets of commuting observables as systems of state labels;
• functions of operators defined through spectral data;
• matrix, differential, multiplication, kernel, and diagonal representations of one abstract operator.

The Born Rule is the canonical home for probability assignment. State changes conditioned on sharp-measurement outcomes are developed in Projective Measurement. Generalized measurements, instruments, noise, and realistic detector models belong to Measurement and Open Quantum Systems.

Quantity, Operator, and Measurement Rule

Three levels should be kept distinct:

1. A physical quantity such as position, energy, or a spin component is specified by an experimental and theoretical context.
2. A sharp observable is represented mathematically by a self-adjoint operator $A$, equivalently by its projection-valued spectral measure.
3. A particular apparatus implements a measurement process with calibration, coupling, resolution, noise, and state disturbance.

The operator does not describe every detail of the apparatus. It specifies the idealized outcome structure and probability rule. Conversely, not every operator is an observable. A unitary operator usually represents a reversible transformation, an annihilation operator is generally not self-adjoint, and a density operator represents a state rather than a measured quantity.

Generalized measurements broaden the outcome rule from projection-valued measures to positive-operator-valued measures. That extension does not invalidate the operator picture for sharp observables; it identifies the larger class to which realistic and nonprojective measurements belong.

Spectral Organization

For a finite-dimensional observable $A$, the spectral theorem gives

$$A = \sum_{a\in\operatorname{spec}(A)} aP_a,$$

where $P_a$ projects onto the eigenspace with eigenvalue $a$. The projectors satisfy

$$P_aP_b = \delta_{ab}P_a, \qquad P_a^\dagger=P_a, \qquad \sum_a P_a=I.$$

For a normalized state $\lvert\psi\rangle$, the probability of outcome $a$ is

$$p(a) = \langle\psi\rvert P_a\lvert\psi\rangle.$$

The expectation value follows from the same spectral data:

$$\langle A\rangle_\psi = \langle\psi\rvert A\lvert\psi\rangle = \sum_a a\,p(a).$$

These equations separate three roles cleanly: $a$ is an outcome value, $P_a$ selects the associated alternative, and the state determines its probability. When $a$ is degenerate, $P_a$ projects onto an entire eigenspace. An observed eigenvalue then need not identify a unique state vector.

For a continuous or mixed spectrum, sums over ordinary eigenvectors are not generally adequate. The precise organization uses a projection-valued measure $E_A$:

$$A = \int_{\mathbb R} \lambda\,dE_A(\lambda),$$ $$\Pr_\psi(A\in\Delta) = \langle\psi\rvert E_A(\Delta) \lvert\psi\rangle.$$

The set $\Delta$ is a measurable range of outcomes. Ideal position and momentum kets are useful generalized eigenvectors, but they are not normalizable Hilbert-space vectors. The chapter’s Discrete and Continuous Spectra article gives the working distinction; Spectral Theorem: Practical Form supplies the mathematical bridge.

Definite Values, Degeneracy, and Labels

A state has a definite value $a$ of a sharp observable when it lies in the corresponding eigenspace:

$$P_a\lvert\psi\rangle = \lvert\psi\rangle,$$

equivalently, under the usual domain assumptions,

$$A\lvert\psi\rangle = a\lvert\psi\rangle.$$

Most states are not eigenstates of a given observable. A superposition spanning several eigenspaces defines a nontrivial probability distribution over its possible outcomes.

One observable may leave a degeneracy unresolved. A mutually commuting family can often be diagonalized simultaneously. It is called a complete set of commuting observables when its joint eigenvalue labels distinguish basis states, up to phase, in the space or sector under discussion. Completeness is therefore contextual: a set complete in a restricted symmetry sector need not be complete after the Hilbert space is enlarged.

Operators Beyond Observables

Operator algebra also organizes transformations and dynamics. A unitary operator $U$ preserves inner products,

$$U^\dagger U = UU^\dagger = I,$$

while a projector $P$ obeys $P^2=P$ and, for an orthogonal projector, $P^\dagger=P$. The Hamiltonian is both an energy observable and the generator of time translations. Its exponential defines time evolution for a time-independent Hamiltonian:

$$U(t) = \exp\!\left(-\frac{iHt}{\hbar}\right).$$

This is an example of the functional calculus. If

$$A=\sum_a aP_a,$$

then, whenever $f$ is defined on the spectrum,

$$f(A) = \sum_a f(a)P_a.$$

The operation acts on spectral values, not separately on matrix entries. Questions about domains become essential for unbounded operators and for unbounded functions of operators.

Abstract Operators and Their Representations

An abstract operator is independent of basis. Its matrix entries in an orthonormal basis $\{\lvert n\rangle\}$ are

$$A_{mn} = \langle m\rvert A\lvert n\rangle.$$

Under a unitary change of components,

$$A' = U^\dagger A U, \qquad \lvert\psi'\rangle = U^\dagger\lvert\psi\rangle.$$

Matrix entries and components change together; spectra, probabilities, and expectation values do not. The same momentum operator can be a differential operator in position space and a multiplication operator in momentum space. Its differential expression alone does not specify an unbounded operator: the Hilbert space, domain, and boundary conditions are also part of the definition.

Page Map

Question	Canonical page	Main distinction
What does an observable represent?	Observables	physical quantity versus apparatus
What is an operator?	Operators	linear map versus observable role
What makes a value definite?	Eigenvalues and Eigenstates	state versus eigenspace label
How do discrete and continuous outcomes differ?	Discrete and Continuous Spectra	normalizable eigenvectors versus generalized spectral labels
Why are domains important?	Hermitian vs Self-Adjoint Operators	matrix Hermiticity versus operator self-adjointness
What represents a yes-no alternative?	Projectors	subspace versus chosen vector
How is an observable reconstructed from outcomes?	Spectral Decomposition	eigenvalues together with spectral projectors
How can degeneracy be resolved?	Complete Sets of Commuting Observables	one label versus joint labels
What do exponentials and square roots of operators mean?	Functions of Operators	spectral calculus versus entrywise operations
Why can one operator have many concrete forms?	Operator Representations	abstract map versus basis-dependent representation

These ten articles form the planned chapter.

Suggested Routes

First systematic pass

Read Observables, Operators, Eigenvalues and Eigenstates, Projectors, and Spectral Decomposition. Continue to the Born rule only after the roles of outcomes and projectors are distinct.

Wave-mechanics route

Pair Discrete and Continuous Spectra with Operator Representations and Hermitian vs Self-Adjoint Operators. Then apply the language to bound and scattering states in Wave Mechanics and Model Systems.

Symmetry and dynamics route

The planned sequence next adds Complete Sets of Commuting Observables and Functions of Operators, then continues to Compatible Observables, Hamiltonians, and the Symmetry, Angular Momentum, and Spin volume. For a continuation available in this edition, use Projectors, then Probability Amplitudes, and finally the Born Rule to turn sharp alternatives into outcome probabilities.

Mathematical-care route

Pair the spectra and self-adjointness pages with Domains of Operators, Symmetric vs Self-Adjoint Operators, and the Spectral Theorem.

Operator Sanity Checks

• Verify dimensions and physical units. An eigenvalue of $A$ has the units of the represented quantity.
• For a finite-dimensional sharp observable, check $A^\dagger=A$ and confirm that numerical eigenvalues are real within tolerance.
• For spectral projectors, check idempotence, self-adjointness, mutual orthogonality, and completeness.
• Confirm that probabilities are nonnegative and sum to one.
• Under a basis change, transform states and operators consistently and verify that expectation values are unchanged.
• In a truncated basis, test convergence of spectra and matrix elements as the truncation grows.
• For differential operators, state the domain and boundary conditions before making claims about self-adjointness or spectra.

Common Mistakes

• Calling every operator an observable. Operators have many roles, and standard sharp observables require self-adjointness.
• Treating an observable as a pre-existing unknown classical value. Quantum states generally assign distributions, not hidden definite eigenvalues, to arbitrary observables.
• Equating an eigenvalue with an eigenvector. The value labels an eigenspace; degeneracy can leave many vectors with the same value.
• Assuming every state is an eigenstate. Generic states have support across several spectral alternatives.
• Writing only eigenvalue sums for continuous spectra. Spectral measures or carefully normalized generalized eigenvectors are needed.
• Using “Hermitian” without a domain caveat for unbounded operators. Symmetry and self-adjointness need not coincide.
• Applying a scalar function entry by entry to an operator matrix. Functional calculus acts on spectral values.
• Changing an operator matrix while leaving state components fixed. A representation change must be made consistently.
• Confusing outcome probabilities with post-measurement states. A POVM or spectral measure gives probabilities; an instrument or update rule supplies the conditioned transformation.

Cross-Links

• States and Representations
• Probability and the Born Rule
• Compatibility, Commutators, and Uncertainty
• Measurement and State Update
• Time Evolution
• Hilbert Spaces
• Measurement and Open Quantum Systems
• Observable glossary entry

References

• P. Busch, P. J. Lahti, J.-P. Pellonpää, and K. Ylinen, Quantum Measurement, Springer, 2016, doi:10.1007/978-3-319-43389-9.
• P. A. M. Dirac, The Principles of Quantum Mechanics, 4th ed., Oxford University Press, 1958.
• A. Peres, Quantum Theory: Concepts and Methods, Kluwer, 1995.
• M. Reed and B. Simon, Methods of Modern Mathematical Physics, Vol. I: Functional Analysis, rev. ed., Academic Press, 1980.
• R. Shankar, Principles of Quantum Mechanics, 2nd ed., Springer, 1994.
• J. von Neumann, Mathematical Foundations of Quantum Mechanics, Princeton University Press, 1955.