Why Operator Algebras?

An observable algebra records which measurements and products belong to a theory before selecting a particular Hilbert-space realization. A state assigns expectation values to that algebra; a representation realizes it by operators. Keeping these three objects separate becomes useful when accessible observables are restricted, when infinitely many subsystems are involved, or when different states require inequivalent representations.

Helpful background. Operator Algebra Minimum introduces products, adjoints, and positivity. Positive State Functionals connects expectation rules with ordinary density operators.

An algebra preserves relations across realizations

Suppose bounded symbols $a,b,\ldots$ describe observables or combinations of observables. A representation $\pi$ must preserve their algebraic relations:

$$\pi(ab)=\pi(a)\pi(b),\qquad \pi(a^*)=\pi(a)^*.$$

A state $\omega$ supplies the numbers $\omega(a)$. One possible realization is $\omega(a)=\langle\Omega,\pi(a)\Omega\rangle$, but the vector and operator realization are additional data. The GNS construction proves that every state on a unital C*-algebra has a cyclic realization and specifies its uniqueness.

This order of construction does not remove Hilbert spaces. It explains which part of the description depends on a state and which relations are common to all states. In the CCR algebra, for example, the Weyl multiplication law is fixed before asking whether a representation is regular or irreducible.

A change of basis inside one representation and a change to an inequivalent representation are different operations. An intertwining unitary implements the former comparison. Such a unitary need not exist for the latter.

Norm completion controls every state at once

The C*-norm gives a useful operational bound. For any state and algebra elements $a,b$,

$$|\omega(a)-\omega(b)|\leq\|a-b\|.$$

Thus replacing $a$ by a norm approximation $b$ controls its expectation error uniformly over the state space. For effects, this controls a probability error. For self-adjoint elements, it controls the mean of a bounded observable. It does not automatically control an unbounded moment.

This estimate motivates completing an algebra of initially simple observables in norm. If $a_n$ is norm Cauchy, then $\omega(a_n)$ is Cauchy for every state, and each state extends continuously to the completion. The completion therefore retains consistent predictions for uniformly controlled approximations.

Other limits ask different questions. A sequence can converge strongly in one representation without converging in algebra norm. Such a limit can depend on which vectors and representation are being used. Von Neumann Algebras treats the corresponding closure after a representation has been selected.

Compatible local systems need no global density matrix

Consider a chain with a $d$-dimensional Hilbert space at each site. A finite set $\Lambda$ has matrix algebra

$$\mathcal A_\Lambda=\bigotimes_{j\in\Lambda}M_d(\mathbb C).$$

For $\Lambda\subset\Gamma$, identify $a\in\mathcal A_\Lambda$ with $a\otimes I_{\Gamma\setminus\Lambda}$. These embeddings preserve norms, products, and adjoints. The union of the finite-region algebras is therefore a consistent star algebra. Its norm completion is the quasilocal C*-algebra.

Suppose density operators $\rho_\Lambda$ satisfy compatibility under partial trace. Define $\omega(a)=\operatorname{Tr}(\rho_\Lambda a)$ when $a$ is supported in $\Lambda$. Compatibility makes the answer independent of the larger region used to evaluate it. Positivity, normalization, and the norm bound then extend this functional uniquely to the quasilocal algebra.

This is already an infinite-system state specification. It neither assumes a trace-class density operator on an infinite tensor-product Hilbert space nor selects such a Hilbert space in advance. The infinite tensor-product discussion explains the representation choices hidden by a formal infinite product of vector spaces.

The construction alone does not prove existence of interacting equilibrium limits or phases. Those require dynamical and thermodynamic estimates. It supplies a common algebra on which candidate limits can be compared.

Accessible measurements and physical sectors are distinct inputs

Choosing a smaller observable algebra can identify preparations that a larger algebra distinguishes. This is a statement about the measurement specification. To call the restriction a fundamental superselection rule requires a physical argument for excluding the missing observables.

Similarly, a representation’s commutant records operators invisible to its algebraic action, while its center identifies classical decomposition data within the represented algebra. Neither object should be inferred solely from degeneracies of one Hamiltonian. The algebraic superselection page keeps those distinctions explicit.

The algebraic approach is consequently useful even for finite systems. Its purpose is to make the chosen observables, continuity assumptions, and state comparisons explicit, not to declare ordinary matrix quantum mechanics obsolete.

Exercise: compatible local probabilities

For each site let $\rho_j$ be a density matrix and set $\rho_\Lambda=\bigotimes_{j\in\Lambda}\rho_j$. Show that these define a state of the quasilocal algebra. Does this require convergence of an infinite product of partition functions?

Solution

Tracing out any added factor gives one, so the finite density matrices are compatible. Their expectations define a positive normalized functional on the local union, bounded by the operator norm, and hence a unique state on its completion. No infinite product of partition functions occurs. The data are normalized local expectations and their compatibility; a global trace representation would be a further assertion.

References

• Bratteli, Ola, and Derek W. Robinson, Operator Algebras and Quantum Statistical Mechanics 1, 2nd ed., Springer, 1987, doi:10.1007/978-3-662-02520-8 — observable algebras, states, and representations.
• Naaijkens, Pieter, Quantum Spin Systems on Infinite Lattices, Lecture Notes in Physics 933, Springer, 2017, author’s notes, arXiv:1311.2717 — compatible local algebras and infinite-system states.