Functional Analysis
C*-algebras and von Neumann algebras
Цели урока
- Understand the C*-identity ||a*a|| = ||a||^2 and its consequence that the norm is algebraically determined
- State the Gelfand theorem (commutative case) and the Gelfand-Naimark theorem (general case)
- Classify von Neumann factors by type: I_n, I_inf, II_1, II_inf, III
- Understand the GNS construction connecting states to representations
Предварительные знания
- Spectral theory of self-adjoint operators
- Hilbert spaces and bounded operators
- Banach algebras
How can one algebraic identity ||a*a|| = ||a||^2 completely determine the norm of an algebra - and why does this force every abstract C*-algebra to be concretely realizable as operators on a Hilbert space?
- Standard Model of particle physics: observable algebra is a C*-algebra; Weyl algebras (bosons) and CAR algebras (fermions) are specific C*-algebras of quantum fields
- Quantum computing: n-qubit algebra = M_{2^n}(C); tensor product structure of C*-algebras describes multi-qubit entanglement
- Topological insulators (Nobel 2016): K-theory of C*-algebra of quasi-periodic Hamiltonian classifies topological phases
- MIP* = RE (Ji et al., 2020): Connes embedding conjecture was disproved using quantum interactive proofs - connecting operator algebra to quantum complexity theory
Von Neumann and the Birth of Operator Algebras
Von Neumann formalized quantum mechanics in 1929 using Hilbert spaces. With Murray, he studied weakly closed *-subalgebras of B(H) in 1936-1943, classifying factors into types I, II, III. Gelfand and Naimark gave the abstract axiomatization (C*-identity) and the representation theorem in 1943. Segal added the GNS construction in 1947. The 1970s brought Connes's classification of type III factors (Fields Medal, 1982). The 2020 result of Ji et al. connecting the Connes conjecture to quantum complexity was perhaps the most surprising development in operator algebras in decades.
C*-Algebras: Axioms and Gelfand-Naimark Theorem
In 1943, Gelfand and Naimark showed: every abstract C*-algebra is concretely realized as an algebra of operators on a Hilbert space. A single algebraic identity ||a*a|| = ||a||^2 replaces infinitely many compatibility conditions between norm and involution. This is the algebraic foundation of quantum mechanics.
Gelfand duality is a categorical equivalence: the category of commutative unital C*-algebras (with *-homomorphisms) is dual to the category of compact Hausdorff spaces (with continuous maps). Commutative C*-algebra <-> compact space. Noncommutative geometry extends this: replace 'commutative C*-algebra' with any C*-algebra, and study the implied 'noncommutative space'.
What does the Gelfand-Naimark theorem state?
Gelfand-Naimark: every C*-algebra A embeds isometrically in B(H) for some Hilbert space H. Proof: construct H via GNS construction from all states on A.
Von Neumann Algebras and Classification of Factors
Von Neumann algebras are C*-algebras additionally closed in the weak operator topology. Von Neumann classified factors (trivial center) in 1936-1943: types I, II_1, II_inf, III. Type II_1 factors with finite normalized traces have 'continuous dimension' - between I_n and I_inf. The Connes conjecture (1976, proved 2020 via quantum complexity) asked whether all separable II_1 factors embed into matrix ultraproducts.
| Type | Projections | Trace | Example |
|---|---|---|---|
| I_n | Ranks 0,1,...,n | tr(p) = rank/n | M_n(C) |
| I_inf | All projections in B(H) | No finite trace | B(l^2) |
| II_1 | Projections: tau in [0,1] | Finite tau(I) = 1 | L(F_2): free group algebra |
| II_inf | Projections: tau in [0,inf) | Semi-finite trace | B(l^2) tensor L(F_2) |
| III | All projections equivalent | No trace | Araki-Woods factors |
What distinguishes a factor from a general von Neumann algebra?
Factor: Z(M) = M cap M' = C*I - trivial center. Murray-von Neumann classification by projection structure: I_n, I_inf, II_1, II_inf, III.
States, GNS Construction, and Tensor Products
A state on a C*-algebra A is a positive normalized linear functional phi: A -> C. The GNS (Gelfand-Naimark-Segal) construction builds from any state phi a Hilbert space H_phi with a representation pi_phi and a cyclic vector xi_phi such that phi(a) = <pi_phi(a)*xi_phi, xi_phi>. This is the mathematical foundation of quantum statistical mechanics.
Quantum entanglement and tensor products
C*-algebraic description of entanglement
A two-qubit system: algebra A = M_2(C) tensor M_2(C) = M_4(C). A product state phi = phi_1 tensor phi_2 is separable - no entanglement. Entangled states are states on M_4(C) that are not product states. The C*-algebra approach makes this precise: entangled = state not factorable across the tensor product. Nuclear C*-algebras: tensor product is unique, no ambiguity in the product state structure.
What does the GNS construction produce from a state phi on A?
GNS: from phi construct H_phi = A / {a: phi(a*a)=0} with inner product <a+N, b+N> = phi(b*a), representation pi_phi(a)(b+N) = ab+N, cyclic vector xi_phi = 1+N. Then phi(a) = <pi_phi(a)*xi_phi, xi_phi>.
Tensor Products of C*-Algebras
Quantum entanglement physically: the state of two particles does not factor into a tensor product of individual states. Mathematically: the tensor product of C*-algebras A⊗B admits several inequivalent norms (minimal and maximal). Nuclear algebras are those for which the norm is unique. Most algebras of quantum mechanics are nuclear.
The Connes embedding conjecture (1976, open until 2020): is every separable type II_1 von Neumann algebra embeddable into an ultraproduct of matrix algebras? In 2020, Ji, Natarajan, Vidick, Wright, and Yuen (MIP* = RE) gave a negative answer, connecting the Connes problem to quantum interactive proofs.
What does nuclearity of a C*-algebra A mean?
Nuclearity: A⊗_{min}B = A⊗_{max}B for all C*-algebras B. This is the infinite-dimensional analogue of finite-dimensionality.
Connections to other topics
C*-algebras and von Neumann algebras are the language of quantum physics and noncommutative mathematics.
- Noncommutative Geometry — Von Neumann algebras are the algebraic backbone of Connes' noncommutative manifolds
- Spectral theory of self-adjoint operators — GNS construction and the spectral theorem are the technical foundation of C*-algebra theory
- Operator K-Theory — K-theory is the topological invariant classifying projections and unitaries in a C*-algebra
Итоги
- C*-identity ||a*a|| = ||a||^2: norm is algebraically determined; unique C*-norm on any *-algebra
- Gelfand theorem: commutative C*-algebra with unit = C(X); Gelfand duality between algebras and spaces
- Gelfand-Naimark: any C*-algebra embeds isometrically in B(H); abstract = concrete
- Von Neumann algebra: M = M'' (bicommutant); factors (trivial center) classified as I, II_1, II_inf, III
- GNS: state phi -> (H_phi, pi_phi, xi_phi); phi(a) = <pi_phi(a)*xi_phi, xi_phi>
Вопросы для размышления
- Why does the C*-identity force the norm to be uniquely determined by the algebraic structure - and what would go wrong if there were two different C*-norms?
- The bicommutant theorem says M = M''. What is the geometric meaning of this for the operator topology?
- Type II_1 factors have 'continuous dimension' tau(p) in [0,1] for projections. How does this interpolate between type I_n (dimensions 0,1,...,n) and I_inf?