PhyslibAlpha.AlgebraicFramework.CStarAlgebra.JordanDecomposition
The Jordan decomposition of a self-adjoint element
Every self-adjoint element `a` of a C⋆-algebra splits canonically as `a = a⁺ - a⁻`, a difference of two *orthogonal* positive elements (`a⁺ * a⁻ = 0`) — the noncommutative analogue of splitting a real-valued function into its positive and negative parts, or a signed measure into its positive and negative variation (Camille Jordan's decomposition theorem, 1881/1892). This is not the same "Jordan" as `StarAlgebra/Jordan.lean`'s Jordan *product* `a ∘ b := a * b + b * a` — that one is named for Pascual Jordan, no relation, and the shared name is an unfortunate but standard clash in the operator-algebra literature.
Mathlib already builds `a⁺`, `a⁻` from the continuous functional calculus (`cfcₙ` applied to the functions `t ↦ max t 0` and `t ↦ max (-t) 0`) and proves the decomposition, orthogonality, and uniqueness facts at the level of a bare C⋆-algebra element. This file packages exactly those facts one level up, at the level of `Observable A := selfAdjoint A` and `PositiveObservable A`, so that "take the positive/negative part" is available as an operation *on observables*, landing in `PositiveObservable A` rather than requiring the caller to separately track self-adjointness and nonnegativity of a bare element of `A` after the fact.
Genuinely needing a full C⋆-algebra here (rather than a bare order-unit space) is not a corner that was cut: the continuous functional calculus behind `a⁺`, `a⁻` needs completeness and the C⋆-identity to exist at all, so `[CStarAlgebra A] [PartialOrder A] [StarOrderedRing A]` is the correct, load-bearing hypothesis, not one to weaken.
Main definitions
- `Observable.posPart`, `Observable.negPart` : the positive and negative parts of an observable, as `PositiveObservable A`. - `Observable.posPart_sub_negPart` : `a⁺ - a⁻ = a`. - `Observable.posPart_mul_negPart`, `Observable.negPart_mul_posPart` : the two parts are orthogonal. - `Observable.posPart_negPart_unique` : this is the *only* decomposition of `a` into a difference of orthogonal positive observables.
A. Positive observables
B. Positive and negative parts
7 declarations
An observable iff for some observable
Let be a -algebra. For any observable , if and only if there exists an observable such that .
Positive part of an observable
For a self-adjoint element of a -algebra (referred to as an observable), its positive part is the positive observable defined via the continuous functional calculus as where . The result is returned as a `PositiveObservable`, which is a structure containing an element along with proofs that is self-adjoint and .
Negative part of an observable
For an observable (a self-adjoint element) in a -algebra , its negative part is the positive element defined via the continuous functional calculus by the function . It is one of the two components in the Jordan decomposition .
Let be a -algebra and be an observable (a self-adjoint element). Let and denote the positive part and negative part of , respectively. Then is the difference of these two parts:
for an observable
Let be a -algebra and be an observable (a self-adjoint element) in . Let and denote the positive part (`posPart a`) and the negative part (`negPart a`) of as defined by the continuous functional calculus. Then their product is zero, i.e., .
for an observable
For any observable in a -algebra , the product of its negative part and its positive part is zero: This signifies that the negative and positive parts of the Jordan decomposition are orthogonal.
Uniqueness of the Jordan decomposition
Let be a -algebra and be an observable (a self-adjoint element of ). Suppose there exist two positive observables and (self-adjoint elements such that ) that satisfy the decomposition and the orthogonality condition . Then is the positive part of (denoted ) and is the negative part of (denoted ).
