Algebraic Formulation Overview

The algebraic formulation of quantum mechanics starts from observables and their algebraic relations rather than from state vectors as the primary object. A physical system is described by an algebra $\mathcal A$ of observables, states are expectation-value functionals on that algebra, and dynamics is a structure-preserving map of the algebra into itself.

In ordinary finite-dimensional quantum mechanics, this sounds like a change of language:

$$\mathcal A=M_n(\mathbb C), \qquad \omega(A)=\operatorname{Tr}(\rho A).$$

But the change of language becomes powerful when Hilbert-space representations are not unique, as in infinite systems, thermodynamic limits, superselection sectors, and quantum field theory. The purpose of this page is orientation, not a rigorous operator-algebra treatise.

Why Start with Observables?

The usual Hilbert-space formulation begins with a Hilbert space $\mathcal H$, states such as $\lvert\psi\rangle$ or $\rho$, and operators acting on $\mathcal H$. The algebraic formulation asks for the structure that is directly probed by experiments: the set of quantities whose expectation values can be measured and compared.

This shift is natural for several reasons:

• the Heisenberg picture already emphasizes time-dependent observables;
• commutation relations and products encode compatibility, uncertainty, and dynamics;
• mixed states and thermal states are expectation-value assignments, not necessarily vectors;
• infinite systems can have inequivalent Hilbert-space representations;
• QFT is naturally organized by local algebras of observables.

The slogan is:

$$\text{state vector first} \quad\longrightarrow\quad \text{observable algebra first}.$$

This is a change of emphasis, not a new set of experimental predictions when applied to ordinary finite quantum systems.

Observables as an Algebra

The detailed canonical treatment is Observables as an Algebra. The summary here fixes the vocabulary needed for the rest of the overview.

An algebra of observables is a collection $\mathcal A$ equipped with addition, scalar multiplication, multiplication, and an adjoint operation $A\mapsto A^*$. In the standard Hilbert-space representation, $A^*$ is the operator adjoint.

The essential operations are:

$$A+B, \qquad \lambda A, \qquad AB, \qquad A^*.$$

The product $AB$ need not commute:

$$AB\neq BA.$$

The commutator

$$[A,B]=AB-BA$$

records noncommutativity. Self-adjoint elements,

$$A^*=A,$$

represent observables in the usual sense, while non-self-adjoint elements are still useful because products, ladder operators, and complex linear combinations belong to the same algebraic structure.

In finite-dimensional quantum mechanics, the model example is

$$\mathcal A=M_n(\mathbb C),$$

the algebra of $n\times n$ complex matrices. In classical mechanics, by contrast, observables are functions on phase space and multiply commutatively:

$$fg=gf.$$

Thus noncommutativity is one of the algebraic signatures of quantum theory.

States as Expectation-Value Functionals

The canonical treatment, including convexity, infinite-dimensional caveats, and the GNS reconstruction, is States as Positive Linear Functionals. The summary here supplies the definition used throughout this overview.

In the algebraic formulation, a state is a rule

$$\omega:\mathcal A\to\mathbb C$$

that assigns expectation values to observables. It is linear:

$$\omega(\alpha A+\beta B) = \alpha\omega(A)+\beta\omega(B),$$

normalized:

$$\omega(I)=1,$$

and positive:

$$\omega(A^*A)\geq0$$

for every $A\in\mathcal A$. Positivity is the algebraic form of nonnegative probabilities and variances.

In the Hilbert-space formulation, density matrices give examples:

$$\omega_\rho(A) = \operatorname{Tr}(\rho A), \qquad \rho\geq0, \qquad \operatorname{Tr}\rho=1.$$

Pure vector states are the special case

$$\omega_\psi(A) = \langle\psi\vert A\vert\psi\rangle.$$

This viewpoint makes mixed states feel less secondary. A state is not first a vector that sometimes becomes a density matrix; it is an expectation-value functional. Vectors and density matrices are representation-dependent ways to implement such functionals.

Dynamics as Automorphisms

The canonical abstract treatment is Dynamics as Automorphisms. The summary here introduces only the structure needed for this overview.

In the Heisenberg picture, observables evolve while states may be held fixed. Algebraically, time evolution is a family of maps

$$\alpha_t:\mathcal A\to\mathcal A$$

that preserve the algebraic structure:

$$\alpha_t(AB)=\alpha_t(A)\alpha_t(B), \qquad \alpha_t(A^*)=\alpha_t(A)^*, \qquad \alpha_t(I)=I.$$

Such a structure-preserving invertible map is called an automorphism. For time-independent Hamiltonian evolution on Hilbert space,

$$\alpha_t(A) = U^\dagger(t)AU(t), \qquad U(t)=e^{-iHt/\hbar}.$$

Differentiating gives the Heisenberg equation

$$\frac{d}{dt}\alpha_t(A) = \frac{i}{\hbar} \alpha_t([H,A]),$$

when the domains and differentiability assumptions are under control.

The infinitesimal generator is a derivation $\delta$, meaning

$$\delta(AB)=\delta(A)B+A\delta(B).$$

For Hamiltonian quantum mechanics,

$$\delta(A)=\frac{i}{\hbar}[H,A].$$

This is the algebraic version of “Hamiltonians generate time evolution.”

Representations and Hilbert Spaces

A representation realizes the abstract algebra as operators on a Hilbert space:

$$\pi:\mathcal A\to\mathcal B(\mathcal H).$$

In finite-dimensional textbook quantum mechanics, one usually begins with this representation and may not notice that it is a representation at all. Algebraic language separates two questions:

1. What are the algebraic relations among observables?
2. Which Hilbert-space representation realizes those relations for the state or sector of interest?

The GNS construction, named after Gelfand, Naimark, and Segal, says that under broad conditions a state $\omega$ produces a Hilbert-space representation

$$(\mathcal H_\omega,\pi_\omega,\Omega_\omega)$$

such that

$$\omega(A) = \langle\Omega_\omega\vert \pi_\omega(A) \vert\Omega_\omega\rangle.$$

This is only a preview, but it explains the logic: Hilbert space can be reconstructed from the algebra plus a state. The state chooses the representation.

Finite Systems and Infinite Systems

For finite numbers of canonical degrees of freedom, the usual Schrödinger representation is essentially unique under regularity and irreducibility assumptions. This is the content of the Stone–von Neumann Theorem.

For infinitely many degrees of freedom, the uniqueness statement fails. Different representations can describe different phases, temperatures, boundary conditions, or superselection sectors. This is one reason algebraic methods become important in many-body physics and QFT.

The algebraic formulation is therefore not merely a philosophical preference. It is a way to discuss physical structure before committing to a single Hilbert-space representation.

Why This Matters for QFT

Quantum field theory has infinitely many degrees of freedom in the continuum. Locality also matters: one wants to associate observables with spacetime regions. Algebraic QFT organizes this using local algebras,

$$\mathcal O \longmapsto \mathcal A(\mathcal O),$$

where $\mathcal O$ is a spacetime region. Causality is expressed by commutation of observables associated with spacelike separated regions.

Thermal equilibrium can be formulated through KMS states, which are algebraic characterizations of equilibrium relative to time evolution. Superselection sectors can be described through inequivalent representations. Gauge theories require care because not every field-like object is an observable.

These topics belong to dedicated mathematical and QFT treatments. The role of this page is to make the vocabulary legible before those subjects appear.

Relation to Other Formulations

The algebraic formulation is closest in spirit to the Heisenberg picture:

$$A(t)=\alpha_t(A),$$

with states assigning expectation values. It also connects naturally to density matrices, because

$$\rho \quad\longmapsto\quad \omega_\rho(A)=\operatorname{Tr}(\rho A)$$

turns a density operator into a state functional.

It differs from the phase-space formulation. Phase-space methods map operators to functions and replace products by star products. Algebraic methods keep the operator product abstract and study the algebra directly.

It also differs from geometric quantum mechanics. Geometric methods emphasize the projective geometry of pure states; algebraic methods emphasize the algebra of observables and expectation-value functionals.

Common Mistakes

• Thinking algebraic quantum mechanics is a different physical theory rather than a different formulation.
• Treating every algebraic statement as if it required full C*-algebra technical machinery.
• Forgetting positivity: a linear functional is not a physical state unless $\omega(A^*A)\geq0$.
• Assuming every state must be represented by a vector in a preselected Hilbert space.
• Ignoring domains when translating unbounded operators into algebraic language.
• Assuming finite-dimensional uniqueness of representations extends to QFT.
• Confusing local field symbols with gauge-invariant observable algebras.

Cross-Links

• Thermodynamic Limit
• Observables as an Algebra
• States as Positive Linear Functionals
• Dynamics as Automorphisms
• Operators
• Commutators
• Density Operators
• Trace Rule and Expectation Values
• Heisenberg Picture
• Symmetries and Dynamical Automorphisms
• Hamiltonians as Generators
• Liouville–von Neumann Equation
• Geometric Quantum Mechanics Overview
• Stone–von Neumann Theorem

References

• R. Haag, Local Quantum Physics: Fields, Particles, Algebras, 2nd ed., Springer, 1996.
• O. Bratteli and D. W. Robinson, Operator Algebras and Quantum Statistical Mechanics 1, 2nd ed., Springer, 1987.
• G. G. Emch, Algebraic Methods in Statistical Mechanics and Quantum Field Theory, Wiley-Interscience, 1972.
• F. Strocchi, An Introduction to the Mathematical Structure of Quantum Mechanics, 2nd ed., World Scientific, 2008.
• R. Kadison and J. Ringrose, Fundamentals of the Theory of Operator Algebras, Volume I, AMS, 1997.
• R. V. Kadison, “Transformations of states in operator theory and dynamics,” Topology 3, 177, 1965.

Exercises

1. Show that a density matrix defines a positive normalized functional.

Solution

Let

$$\omega_\rho(A)=\operatorname{Tr}(\rho A),$$

where $\rho\geq0$ and $\operatorname{Tr}\rho=1$. Normalization is

$$\omega_\rho(I)=\operatorname{Tr}\rho=1.$$

For positivity,

$$\omega_\rho(A^*A) = \operatorname{Tr}(\rho A^*A).$$

Write $\rho=\sum_n r_n\lvert n\rangle\langle n\rvert$ with $r_n\geq0$. Then

$$\operatorname{Tr}(\rho A^*A) = \sum_n r_n \langle n\vert A^*A\vert n\rangle = \sum_n r_n \|A\lvert n\rangle\|^2 \geq0.$$

2. Verify that unitary Heisenberg evolution is an automorphism.

Solution

Let

$$\alpha_t(A)=U^\dagger(t)AU(t),$$

with $U^\dagger U=I$. Then

$$\alpha_t(AB) = U^\dagger ABU = U^\dagger A(UU^\dagger)BU = \alpha_t(A)\alpha_t(B).$$

Also,

$$\alpha_t(A^*) = U^\dagger A^*U = (U^\dagger AU)^* = \alpha_t(A)^*.$$

Finally,

$$\alpha_t(I)=U^\dagger IU=I.$$

Thus $\alpha_t$ preserves the algebraic operations.

3. Show that the commutator generator is a derivation.

Solution

Define

$$\delta(A)=\frac{i}{\hbar}[H,A].$$

Then

$$[H,AB]=HAB-ABH.$$

Insert and subtract $AHB$:

$$[H,AB] = (HA-AH)B + A(HB-BH) = [H,A]B+A[H,B].$$

Multiplying by $i/\hbar$ gives

$$\delta(AB)=\delta(A)B+A\delta(B).$$

4. Compare a vector state and a general state functional.

Solution

A vector state has the form

$$\omega_\psi(A)=\langle\psi\vert A\vert\psi\rangle.$$

It is a special kind of positive normalized functional. A general state functional need not be given by a vector in a preselected Hilbert space. In ordinary finite-dimensional quantum mechanics, mixed states are represented by density matrices:

$$\omega_\rho(A)=\operatorname{Tr}(\rho A).$$

In algebraic settings, a state may instead determine its own Hilbert-space representation through the GNS construction.

5. Why do infinite systems make representations more important?

Solution

For finite canonical systems, regular irreducible representations of the canonical commutation relations are essentially unitarily equivalent. For infinitely many degrees of freedom, this uniqueness fails. Different representations can describe different phases, thermodynamic states, vacua, or superselection sectors. Therefore the algebraic relations alone do not pick out one universal Hilbert space; the state and physical sector matter.