Functional Analysis
Connes' noncommutative geometry
Цели урока
- Understand the construction of a spectral triple (A, H, D) and its parts
- Know the Connes metric d(x, y) and its link to geodesic distance
- Master the Dixmier integral and spectral dimension
- Understand how the Standard Model of physics is encoded in a spectral triple
Предварительные знания
- C*-algebras and the Gelfand-Naimark theorem
- Self-adjoint operators and the spectral theorem
- Operator K-theory
Why can the 17 elementary particles of the Standard Model be derived from one geometric object: a spectral triple over a 12-dimensional algebra?
- Standard Model of physics (Connes-Marcolli, 2007): the Lagrangian of every particle and interaction comes out of the spectral action Tr(f(D / Lambda^2))
- Topological insulators: K-theory of spectral triples classifies topological phases of matter (the 2016 Nobel-class of problems)
- Quantum torus: noncommutative geometry of quasiperiodic systems, Cantor fractals, foliations, areas with noncommutative 'coordinates'
- Spectral methods in ML: Laplacian Eigenmaps and Graph Neural Networks are discrete analogs of spectral triples on graphs
Connes' path from von Neumann algebras to physics
Alain Connes started by classifying type III factors in the 1970s, earning the Fields Medal in 1982 for it. Cyclic cohomology (1983) is his noncommutative analog of de Rham cohomology. Spectral triples appeared in the late 1980s as an attempt to describe 'quantum spaces' in operator language. The 1994 book 'Noncommutative Geometry' systematized everything. In 1997 Connes and Lott, then Connes and Marcolli (2007), showed that the Standard Model equals a spectral triple. The match is not coincidence: Connes' programme predicted a Higgs mass of 170 GeV in 2008. After the 2012 discovery at 125 GeV the triple required refinement.
Connes' spectral triples
Alain Connes received the Fields Medal in 1994. His programme: replace a geometric space by the algebra of functions on it, then generalize the construction to noncommutative algebras. The payoff: the Standard Model of particle physics arises as a spectral triple over a C*-algebra.
Classical case: a Riemannian manifold
Spectral triple for M
For a smooth compact Riemannian manifold M: A = C^inf(M), H = L^2(M, S) (spinors), D = the Dirac operator on M. The Connes metric agrees with the geodesic distance. The Dixmier integral agrees with the integral over M. This confirms that classical geometry is a special case of noncommutative geometry.
Noncommutative geometry reverses the usual flow: instead of studying spaces, study the algebras of functions on them. For noncommutative algebras there is no 'space', and yet the geometry persists. Quantum groups, groupoids, foliations all have a noncommutative geometric meaning.
The condition [D, a] in B(H) inside a spectral triple is the noncommutative analog of what?
On a classical manifold [D, f] = grad(f) (more precisely the Clifford symbol of the gradient). Boundedness of [D, a] is the analog of finiteness of the gradient norm, i.e. Lipschitz continuity.
Noncommutative integration: traces and dimension
The classical integral of f(x) dx is the trace of the multiplication operator in a suitable sense. Connes showed that the integral over an n-dimensional manifold equals the Dixmier trace Tr_omega(f * |D|^{-n}). This makes it possible to define an integral on 'spaces' with no points: quantized manifolds, fractals, foliations.
| Space | Algebra A | Operator D | Spectral dimension |
|---|---|---|---|
| Circle S^1 | C^inf(S^1) | -i d/dx | 1 |
| Sphere S^n | C^inf(S^n) | Dirac operator on S^n | n |
| Two-point space | C^2 | Lambda * sigma_1 | 0 |
| Torus T^n | C^inf(T^n) | -i * nabla | n |
| Quantum torus A_theta | Rotation algebra | Deformed D | 2 |
What is the Dixmier trace Tr_omega(T)?
Dixmier trace: Tr_omega(T) = lim_omega (1 / log N) * sum_{n <= N} mu_n(T). It regularizes a logarithmically divergent sum. On an n-dimensional manifold Tr_omega(f * |D|^{-n}) = c_n * integral of f.
The Standard Model as a spectral triple
2007. Connes and Marcolli publish: the Standard Model Lagrangian (17 particles, 4 interactions) follows from a spectral triple with algebra A = C tensor H tensor M_3(C), where H denotes the quaternions. This is not curve fitting. Algebraic conditions uniquely fix the particles and the coupling constants.
Standard Model as a spectral triple: A = C + H + M_3(C) (direct sum), H = L^2(M) tensor F (spinors plus internal degrees of freedom), D = D_M tensor 1 + gamma_5 tensor D_F. The Dirac operator D_F encodes fermion masses and the CKM mixing matrix. The strong, electromagnetic, and weak interactions appear as internal symmetries of fluctuations of the Dirac operator.
What does the spectral action S = Tr(f(D_A / Lambda^2)) fix in its Lambda expansion?
The expansion S[D_A] = Tr(f(D_A / Lambda^2)) in Lambda gives Lambda^4 * (volume), Lambda^2 * (Yang-Mills action), Lambda^0 * (Standard Model Lagrangian + Einstein-Hilbert) plus decreasing corrections.
Cyclic cohomology and the index theorem
The Atiyah-Singer index theorem (1963) computes the index of elliptic operators via topological invariants. Connes generalized it to the noncommutative case using cyclic cohomology, a noncommutative analog of de Rham cohomology. That opened the door to a noncommutative geometry of foliations and quantum spaces.
Connes' index theorem for spectral triples: ind(D^+) = <tau, ch([e])>, where tau is a cyclic cocycle and ch is the noncommutative Chern character. This is the noncommutative generalization of Gauss-Bonnet: for a 0-dimensional triple it yields the formula for the Euler number.
What does the Atiyah-Singer index theorem compute?
Atiyah-Singer theorem: ind(D) = dim ker D - dim ker D* = integral of A_hat(M) wedge ch(E). The analytic index (kernel dimensions) equals the topological index (integral of characteristic classes).
Links to other topics
Noncommutative geometry brings algebra, geometry, and physics together.
- C*-algebras and von Neumann algebras — Spectral triples are built over C*-algebras rather than commutative function algebras
- Operator K-theory — Connes' cyclic cohomology pairs with K-theory through the Chern-Connes formulas
- Free probability — Free probability is the probabilistic extension of Connes' non-abelian geometry
Итоги
- Spectral triple (A, H, D): algebra plus Hilbert space plus self-adjoint Dirac operator
- Connes metric: d(x, y) = sup{|f(x) - f(y)| : ||[D, f]|| <= 1}. Generalizes geodesic distance
- Dixmier integral: Tr_omega(f * |D|^{-n}) = c_n * integral of f. Noncommutative integration
- Standard Model: A = C + H + M_3(C). Spectral action = Lagrangian plus gravity
- Index theorem: ind(D) = <[D], K_0(A)>. Analytic = topological via cyclic cohomology
Вопросы для размышления
- What is the principled difference between ordinary geodesic distance and the Connes metric, and what does the Connes metric 'feel' in the noncommutative case?
- How does one object, a spectral triple, give both the geometry of spacetime and the internal structure of elementary particles?
- What does a non-integer spectral dimension mean for the Cantor fractal?