Physlib

PhyslibAlpha.ProbabilisticTheory.State.Convex

Convex state spaces

i. Overview

States mix: a probabilistic combination of two states is again a state, and the state space embeds convexly into the algebraic dual. A pure state is one that is never a proper mixture of two others, an extreme point of that convex set. A mixed state is one that is a mixture.

ii. Key results

- `UnitalPositiveLinearMap.mix` : randomize between two states with a given probability. - `UnitalPositiveLinearMap.convex_algebraicStateSpace` : the state space is convex in the algebraic dual. - `UnitalPositiveLinearMap.isPure_iff_forall_mix_eq` : a state is pure iff every proper mixture producing it is trivial.

iii. Table of contents

  • A. Mixing states
  • B. The state space
  • C. Pure and mixed states

A. Mixing states

B. The state space

C. Pure and mixed states

10 declarations

definition

Probabilistic mixture of states tω+(1−t)φt\omega + (1-t)\varphi

Given two states ω,φ∈S[R,E]\omega, \varphi \in \mathcal{S}[\mathbb{R}, E] and a probability t∈[0,1]t \in [0, 1], this function defines a new state representing the probabilistic mixture of ω\omega and φ\varphi. The resulting state is the convex combination tω+(1−t)φt\omega + (1-t)\varphi, which maps an element A∈EA \in E to the value tω(A)+(1−t)φ(A)t \omega(A) + (1 - t) \varphi(A).

theorem

mix(ω,φ,t)(A)=tω(A)+(1−t)φ(A)\text{mix}(\omega, \varphi, t)(A) = t \omega(A) + (1 - t) \varphi(A)

For any two states ω,φ∈S[R,E]\omega, \varphi \in \mathcal{S}[\mathbb{R}, E], any probability t∈[0,1]t \in [0, 1], and any element A∈EA \in E, the evaluation of the probabilistic mixture state mix(ω,φ,t)\text{mix}(\omega, \varphi, t) at AA is given by the convex combination of the individual evaluations: mix(ω,φ,t)(A)=tω(A)+(1−t)φ(A). \text{mix}(\omega, \varphi, t)(A) = t \omega(A) + (1 - t) \varphi(A).

definition

Algebraic state space of EE

The algebraic state space of a space EE is the set of linear maps in the algebraic dual space E→ℓRE \to_{\ell} \mathbb{R} that are states. Specifically, it is the range of the function that maps each state ω∈S[R,E]\omega \in \mathcal{S}[\mathbb{R}, E] (a unital positive linear map) to its underlying linear map.

theorem

The algebraic state space is convex

The algebraic state space of a space EE, which consists of the linear maps in the algebraic dual E→ℓRE \to_{\ell} \mathbb{R} that are states (unital positive linear maps), is a convex set over the real numbers R\mathbb{R}.

definition

ω\omega is a pure state

A state ω\omega in the state space S[R,E]\mathcal{S}[\mathbb{R}, E] is **pure** if its underlying linear map is an extreme point of the convex set of all states. Geometrically, this means ω\omega cannot be expressed as a non-trivial convex combination of two distinct states in the state space.

definition

Mixed state ω\omega

A state ω\omega in the state space S(E)\mathcal{S}(E) is mixed if it is not a pure state. Formally, this means ω\omega is not an extreme point of the convex set S(E)\mathcal{S}(E).

theorem

ω∈openSegment(φ,ψ)\omega \in \text{openSegment}(\varphi, \psi) iff ω\omega is a Proper Mixture of φ\varphi and ψ\psi

Let S[R,E]\mathcal{S}[\mathbb{R}, E] be the state space consisting of unital positive linear maps. For any states ω,φ,ψ∈S[R,E]\omega, \varphi, \psi \in \mathcal{S}[\mathbb{R}, E], the linear map ω\omega lies in the open segment between φ\varphi and ψ\psi if and only if there exists a probability tt in the unit interval [0,1][0, 1] with t≠0t \neq 0 and t≠1t \neq 1 such that ω\omega is the proper mixture tφ+(1−t)ψt\varphi + (1-t)\psi.

theorem

ω\omega is pure iff every proper mixture producing it is trivial

Let S(E)\mathcal{S}(E) be the space of states (unital positive linear maps). A state ω∈S(E)\omega \in \mathcal{S}(E) is pure if and only if for every pair of states ϕ,ψ∈S(E)\phi, \psi \in \mathcal{S}(E) and every tt in the unit interval [0,1][0, 1] with t≠0t \neq 0 and t≠1t \neq 1, the mixture tϕ+(1−t)ψ=ωt\phi + (1-t)\psi = \omega implies that ϕ=ω\phi = \omega and ψ=ω\psi = \omega.

theorem

ω\omega is pure   ⟺  F(ω)\iff F(\omega) is an extreme point of F(S)F(\mathcal{S})

Let EE be an ordered vector space and S\mathcal{S} denote the state space of unital positive linear maps on EE. Let XX be a real vector space and F:S→XF: \mathcal{S} \to X be an injective map that preserves convex mixtures, such that for all states ϕ,ψ∈S\phi, \psi \in \mathcal{S} and all t∈[0,1]t \in [0, 1], F(tϕ+(1−t)ψ)=tF(ϕ)+(1−t)F(ψ)F(t\phi + (1-t)\psi) = t F(\phi) + (1-t)F(\psi). Then a state ω∈S\omega \in \mathcal{S} is pure if and only if its image F(ω)F(\omega) is an extreme point of the image of the state space F(S)F(\mathcal{S}) in XX.

theorem

ω\omega is mixed iff ω=tϕ+(1−t)ψ\omega = t\phi + (1-t)\psi for t∈(0,1)t \in (0, 1) and ϕ≠ω\phi \neq \omega

A state ω\omega is mixed if and only if it can be represented as a proper convex combination of states that are not equal to ω\omega. Formally, ω\omega is mixed if and only if there exist states ϕ,ψ\phi, \psi and a scalar t∈(0,1)t \in (0, 1) such that ω=tϕ+(1−t)ψ\omega = t\phi + (1-t)\psi and (ϕ≠ω∨ψ≠ω)(\phi \neq \omega \lor \psi \neq \omega).