English · Español · Français · Deutsch · 中文 · 日本語 · Русский

Finite-dimensional von Neumann algebras: lecture notes¶

This course teaches the theory of finite-dimensional von Neumann algebras — the algebras \(\bigoplus_cM_{k_c}(\mathbb C)\) of block-diagonal matrices — together with their uses in quantum physics, quantum information, random matrix theory and machine learning: symmetry and superselection sectors, the double commutant theorem and the Wedderburn structure theorem, traces, states and entropy, Haar measure and the Weingarten calculus, quantum channels (Kraus, Stinespring, Choi), the Lindblad (GKSL) equation, spin chains, quantum error correction, equivariant neural networks and Jordan algebras.

It is written for two kinds of readers at once:

Mathematicians get a self-contained text: every theorem, formula, constant and inequality that is stated is proved in the course — in the chapter itself or in one of the appendices A–F, cited by number — from the prerequisites listed below. Numbers that are only computed (for instance large-\(n\) level statistics) are labelled numerical observation and are never used as premises or test targets. References are historical attribution or further reading, never a substitute for a proof.
Physicists and engineers get the physical meaning of each notion in boxes marked Physical meaning. These boxes are physical interpretation, not theorems of the course: nothing proved elsewhere depends on them. Working code they can adapt is in the tutorial.

Chapters 1–13 have the same structure: definitions and results with complete proofs, worked examples, and exercises with solutions. After every result that the library torch_vn_algebra computes or checks, a line Computation: or In the library: links to the place in the tutorial where this is done.

Contents¶

	chapter	main results
0	What this course is about	the algebra as a data type: block structure = batched tensors; cost; what exists and what is new
1	Matrix algebras and functional calculus	spectral theorem, functional calculus, polar decomposition, trace-norm duality, a GUE moment
2	Finite-dimensional von Neumann algebras	double commutant theorem, structure theorem, centre, symmetry groups and commutants, conditional expectations
3	Traces, states and entropy	classification of traces, densities, entropy of sectors, Klein's inequality, monotonicity under the conditional expectation onto the centre, Michelson contrast
4	Symmetries and representations	Schur's lemma, charge sectors, \(U(1)\) and \(SU(2)\), spin systems, real, complex and quaternionic type
5	Random matrices in an algebra	Haar measure on \(\prod_cU(k_c)\), Weingarten calculus, level statistics and why sectors must not be mixed
6	Quantum channels	Kraus and Stinespring, duals with respect to a trace, maps between sectors, decoherence
7	Tensor products and entanglement	partial trace, Page's formula, fusion of charges
8	Dynamics	Schrödinger and Lindblad equations, jumps between sectors, post-selection
9	Spin chains	sectors of magnetisation and momentum, Lanczos, Bethe ansatz checks, disorder
10	Quantum computing	decoherence-free subspaces, operator-algebra error correction, constrained mixers
11	Optimisation and trace inequalities	constrained parametrisations, sampling versus optimisation
12	Equivariant machine learning	equivariant layers as commutants
13	Jordan algebras: why quantum mechanics uses *-algebras	Jordan product and quadratic representation, spin factors and the Lorentz cone, the Jordan–von Neumann–Wigner classification, the transpose and the partial transpose, a trace inequality in Jordan form
A	Facts from analysis and probability	power series and null sets, Gaussian vectors, spherical measures, eigenvalue inequalities, Isserlis–Wick
B	Schur–Weyl duality and the Weingarten calculus	Schur–Weyl duality, Weingarten functions, Wigner–Eckart theorem, Haar measure on \(O(k)\)
C	Eigenvalue densities and Page's formula	Weyl integration formula, GOE/GUE, Wishart and induced densities, Page's formula
D	Generators of quantum dynamical semigroups	the GKSL theorem, generators on direct sums, Frigerio's theorem, Bény–Kempf–Kribs
E	Classification of Euclidean Jordan algebras	Peirce decomposition, Hurwitz and Frobenius theorems, coordinatisation, the Albert algebra, the Jordan–von Neumann–Wigner theorem
F	Symmetric cones, order isomorphisms and Jordan maps	cones of squares, the Koecher–Vinberg theorem, Kadison's theorem, Jordan automorphisms of \(\bigoplus_cM_{k_c}\)

Prerequisites¶

The course assumes only the following, and proves everything else.

Linear algebra over \(\mathbb R\) and \(\mathbb C\) (finite-dimensional), and the fundamental theorem of algebra.
Standard real analysis and multivariable calculus: limits, continuity, compactness in \(\mathbb R^n\), differentiation, the mean value theorem, Taylor's formula, the inverse and implicit function theorems.
Standard measure and integration theory: Lebesgue measure, product measures and Fubini–Tonelli, dominated and monotone convergence, change of variables in \(\mathbb R^n\), pushforward measures, densities, independence.
Elementary probability: expectation, variance, independence, the Gaussian distribution and its moment generating function, the Markov and Chebyshev inequalities, the law of large numbers.

No physics and no functional analysis are assumed.

Finite-dimensional von Neumann algebras: lecture notes¶

Contents¶

Prerequisites¶

Further reading¶