Physlib

PhyslibAlpha.AlgebraicFramework.CStarAlgebra.Uncertainty

Uncertainty relations

Positivity of a state gives a Cauchy–Schwarz inequality for expectation values. Applied to centered observables, this yields the Robertson–Schrödinger and Robertson uncertainty relations.

Unlike `StarAlgebra.Statistics`, this file genuinely needs a C⋆-algebra: `gns_cauchy_schwarz`, the seed of every inequality below, is Cauchy–Schwarz for the sesquilinear form `(x, y) ↦ ω(x⋆y)`, and the cleanest route to it is through the GNS Hilbert space `ω.GNS` of `CStarAlgebra.GNS` — the completion needs `CStarAlgebra A`, not just a bare star-ordered ring. We get it from the already-built `UnitalPositiveLinearMap.gnsRep`/`gnsCyclicVector` API (rather than reaching past `GNS.lean` into mathlib's raw `PreGNS` machinery) via the identity `⟪π_ω(x) Ω_ω, π_ω(y) Ω_ω⟫ = ω(x⋆y)`: expand the inner product using that `π_ω` is a ⋆-representation and that `Ω_ω` reproduces `ω` (`inner_gnsCyclicVector_gnsRep_gnsCyclicVector`), then invoke the general Cauchy–Schwarz inequality `inner_mul_inner_self_le` on the Hilbert space `ω.GNS`.

The commutator observable used to state the relations is the Lie bracket `⁅a, b⁆` already built in `StarAlgebra/Lie.lean`; the only fact about it needed here beyond what that file already proves is that centering leaves it unchanged (`bracket_centered`).

Main results

- `gns_cauchy_schwarz` : Cauchy–Schwarz for the state-induced sesquilinear form on `A`. - `robertson_schrodinger` : the Robertson–Schrödinger uncertainty inequality, jointly bounding covariance and the commutator's expectation by the product of the spreads. - `covariance_cauchy_schwarz`, `robertson` : the familiar `|correlation| ≤ σ_a σ_b` and Heisenberg-type `|⟨⁅a,b⁆⟩| ≤ σ_a σ_b` relations obtained from it by dropping one term.

Cauchy–Schwarz

Uncertainty relations

17 declarations

theorem

The commutator is invariant under centering: ⟦a~,b~⟧=⟦a,b⟧\llbracket \tilde{a}, \tilde{b} \rrbracket = \llbracket a, b \rrbracket

Let AA be a unital C∗C^*-algebra and ω\omega be a state on AA. For any observables a,b∈Aa, b \in A, let the centered observables be defined by a~=a−ω(a)1\tilde{a} = a - \omega(a)\mathbf{1} and b~=b−ω(b)1\tilde{b} = b - \omega(b)\mathbf{1}, where 1\mathbf{1} is the identity element of AA. Then the commutator of the centered observables is equal to the commutator of the original observables: ⟦a~,b~⟧=⟦a,b⟧\llbracket \tilde{a}, \tilde{b} \rrbracket = \llbracket a, b \rrbracket where ⟦x,y⟧=xy−yx\llbracket x, y \rrbracket = xy - yx denotes the Lie bracket (commutator).

theorem

ω(ΔaΔb)=Covω(a,b)+12ω([a,b])\omega(\Delta a \Delta b) = \text{Cov}_\omega(a, b) + \frac{1}{2} \omega([a, b])

Let ω\omega be a state (a unital positive linear map) on a C∗C^*-algebra AA. For any two observables a,b∈Aa, b \in A, let Δa=a−ω(a)1\Delta a = a - \omega(a)1 and Δb=b−ω(b)1\Delta b = b - \omega(b)1 denote their centered fluctuations. The expectation value of the product of these fluctuations decomposes into a real symmetric part (the covariance) and an imaginary antisymmetric part (half the expectation of the commutator): ω(ΔaΔb)=Covω(a,b)+12ω([a,b])\omega(\Delta a \Delta b) = \text{Cov}_\omega(a, b) + \frac{1}{2} \omega([a, b]) where Covω(a,b)=12ω(ΔaΔb+ΔbΔa)\text{Cov}_\omega(a, b) = \frac{1}{2} \omega(\Delta a \Delta b + \Delta b \Delta a) is the covariance and [a,b]=ab−ba[a, b] = ab - ba is the commutator of aa and bb.

theorem

⟨πω(x)Ωω,πω(y)Ωω⟩=ω(x∗y)\langle \pi_\omega(x) \Omega_\omega, \pi_\omega(y) \Omega_\omega \rangle = \omega(x^* y)

Let AA be a unital C∗C^*-algebra and ω\omega be a state (a unital positive linear map) on AA. Let πω:A→B(Hω)\pi_\omega: A \to \mathcal{B}(H_\omega) be the GNS representation of AA on the Hilbert space HωH_\omega, and let Ωω∈Hω\Omega_\omega \in H_\omega be the GNS cyclic vector. For any elements x,y∈Ax, y \in A, the inner product in HωH_\omega of the vectors πω(x)Ωω\pi_\omega(x)\Omega_\omega and πω(y)Ωω\pi_\omega(y)\Omega_\omega is given by the value of the state on the product x∗yx^* y: ⟨πω(x)Ωω,πω(y)Ωω⟩=ω(x∗y)\langle \pi_\omega(x) \Omega_\omega, \pi_\omega(y) \Omega_\omega \rangle = \omega(x^* y) where x∗x^* denotes the adjoint of xx.

theorem

Cauchy–Schwarz Inequality for State-induced Sesquilinear Forms

Let AA be a C∗C^*-algebra and ω\omega be a state on AA (a unital positive linear map). For any elements x,y∈Ax, y \in A, the following inequality holds: ∣ω(x∗y)∣⋅∣ω(y∗x)∣≤Re(ω(x∗x))⋅Re(ω(y∗y))|\omega(x^* y)| \cdot |\omega(y^* x)| \leq \text{Re}(\omega(x^* x)) \cdot \text{Re}(\omega(y^* y)) where x∗x^* denotes the adjoint of xx, ∣⋅∣|\cdot| denotes the complex modulus, and Re(⋅)\text{Re}(\cdot) denotes the real part.

theorem

∥ω(aˉbˉ)∥⋅∥ω(bˉaˉ)∥≤Varω(a)Varω(b)\|\omega(\bar{a}\bar{b})\| \cdot \|\omega(\bar{b}\bar{a})\| \leq \text{Var}_\omega(a) \text{Var}_\omega(b)

Let ω\omega be a state (a unital positive linear map) on a C∗C^*-algebra AA. For any observables a,b∈Aa, b \in A, let aˉ=a−ω(a)1\bar{a} = a - \omega(a)1 and bˉ=b−ω(b)1\bar{b} = b - \omega(b)1 denote the centered observables (fluctuations). Then the product of the norms of the expectations of their cross-fluctuations is bounded by the product of their variances: ∥ω(aˉbˉ)∥⋅∥ω(bˉaˉ)∥≤Varω(a)⋅Varω(b)\|\omega(\bar{a} \bar{b})\| \cdot \|\omega(\bar{b} \bar{a})\| \leq \text{Var}_\omega(a) \cdot \text{Var}_\omega(b) where Varω(x)=ω(xˉ2)\text{Var}_\omega(x) = \omega(\bar{x}^2) is the variance of the observable xx in the state ω\omega.

theorem

Cauchy–Schwarz Inequality for Centered Observables: ∣ω(aˉbˉ)∣2≤Varω(a)Varω(b)|\omega(\bar{a}\bar{b})|^2 \le \text{Var}_\omega(a) \text{Var}_\omega(b)

Let AA be a unital C∗C^*-algebra and ω:A→C\omega: A \to \mathbb{C} be a state (a unital positive linear map). For any two observables a,b∈Aa, b \in A (self-adjoint elements), let aˉ=a−ω(a)1\bar{a} = a - \omega(a)1 and bˉ=b−ω(b)1\bar{b} = b - \omega(b)1 be the centered observables. The squared magnitude of the expectation value of the product of these centered observables is bounded by the product of their variances: ∣ω(aˉbˉ)∣2≤Varω(a)⋅Varω(b)|\omega(\bar{a}\bar{b})|^2 \le \text{Var}_\omega(a) \cdot \text{Var}_\omega(b) where Varω(x)=ω(xˉ2)\text{Var}_\omega(x) = \omega(\bar{x}^2) denotes the variance of the observable xx with respect to the state ω\omega.

theorem

The Robertson–Schrödinger Uncertainty Inequality

Let AA be a C∗C^*-algebra and ω\omega be a state (a unital positive linear map) on AA. For any two observables (self-adjoint elements) a,b∈Aa, b \in A, the Robertson–Schrödinger uncertainty inequality holds: Covω(a,b)2+(12iω([a,b]))2≤Varω(a)⋅Varω(b)\text{Cov}_\omega(a, b)^2 + \left( \frac{1}{2i} \omega([a, b]) \right)^2 \le \text{Var}_\omega(a) \cdot \text{Var}_\omega(b) where Varω(a)\text{Var}_\omega(a) and Varω(b)\text{Var}_\omega(b) are the variances of aa and bb with respect to the state ω\omega, Covω(a,b)\text{Cov}_\omega(a, b) is their covariance, and [a,b]=ab−ba[a, b] = ab - ba is the commutator of aa and bb.

theorem

Covariance Cauchy-Schwarz Inequality: covω(a,b)2≤varω(a)⋅varω(b)\text{cov}_\omega(a, b)^2 \leq \text{var}_\omega(a) \cdot \text{var}_\omega(b)

Let ω\omega be a state on a C∗C^*-algebra AA. For any two observables aa and bb in AA, the square of their covariance with respect to ω\omega is less than or equal to the product of their variances: covω(a,b)2≤varω(a)⋅varω(b)\text{cov}_\omega(a, b)^2 \leq \text{var}_\omega(a) \cdot \text{var}_\omega(b) where covω(a,b)\text{cov}_\omega(a, b) denotes the covariance of aa and bb, and varω(x)\text{var}_\omega(x) denotes the variance of an observable xx under the state ω\omega.

theorem

Robertson Uncertainty Relation: ω(12i[a,b])2≤Varω(a)Varω(b)\omega(\frac{1}{2i}[a, b])^2 \leq \text{Var}_\omega(a) \text{Var}_\omega(b)

Let ω\omega be a state on a unital C∗C^*-algebra AA. For any two observables a,b∈Aa, b \in A, the square of the expectation value of their commutator (multiplied by 1/2i1/2i to ensure it is real) is less than or equal to the product of their variances: ω(12i[a,b])2≤Varω(a)⋅Varω(b) \omega\left( \frac{1}{2i} [a, b] \right)^2 \leq \text{Var}_\omega(a) \cdot \text{Var}_\omega(b) where [a,b]=ab−ba[a, b] = ab - ba is the commutator and Varω(x)=ω(x2)−ω(x)2\text{Var}_\omega(x) = \omega(x^2) - \omega(x)^2 is the variance of the observable xx in the state ω\omega. This result is a version of Heisenberg's uncertainty relation obtained by neglecting the covariance term in the Robertson–Schrödinger inequality.

definition

Gram defect Varω(a)Varω(b)−∣ω(ΔaΔb)∣2\text{Var}_\omega(a)\text{Var}_\omega(b) - |\omega(\Delta a \Delta b)|^2

Let ω\omega be a state (a unital positive linear map) on a C∗C^*-algebra AA, and let a,ba, b be observables (self-adjoint elements) in AA. The **centered Gram defect** is the real-valued difference between the product of the variances of aa and bb and the squared absolute value of the expectation of the product of their centered fluctuations: Varω(a)Varω(b)−∣ω(ΔaΔb)∣2\text{Var}_\omega(a) \text{Var}_\omega(b) - |\omega(\Delta a \Delta b)|^2 where Δa=a−ω(a)1\Delta a = a - \omega(a)\mathbf{1} and Δb=b−ω(b)1\Delta b = b - \omega(b)\mathbf{1} are the centered observables, and Varω(x)=ω(Δx2)\text{Var}_\omega(x) = \omega(\Delta x^2) is the variance of xx in the state ω\omega. This quantity represents the non-negative gap in the Robertson–Schrödinger uncertainty relation.

theorem

The centered Gram defect is non-negative: 0≤Varω(a)Varω(b)−∣ω(ΔaΔb)∣20 \le \text{Var}_\omega(a) \text{Var}_\omega(b) - |\omega(\Delta a \Delta b)|^2

Let ω\omega be a state (a unital positive linear map) on a C∗C^*-algebra AA. For any two observables (self-adjoint elements) a,b∈Aa, b \in A, the centered Gram defect is non-negative: 0≤Varω(a)Varω(b)−∣ω(ΔaΔb)∣20 \leq \text{Var}_\omega(a) \text{Var}_\omega(b) - |\omega(\Delta a \Delta b)|^2 where Δa=a−ω(a)1\Delta a = a - \omega(a)\mathbf{1} and Δb=b−ω(b)1\Delta b = b - \omega(b)\mathbf{1} are the centered observables, and Varω(x)=ω(Δx2)\text{Var}_\omega(x) = \omega(\Delta x^2) is the variance of xx in the state ω\omega.

theorem

∣ω(ΔaΔb)∣2=Covω(a,b)2+ω(⟦a,b⟧)2|\omega(\Delta a \Delta b)|^2 = \text{Cov}_\omega(a, b)^2 + \omega(\llbracket a, b \rrbracket)^2

Let AA be a star-algebra and ω\omega be a state (a unital positive linear map) on AA. For any two observables (self-adjoint elements) a,b∈Aa, b \in A, let Δa=a−ω(a)1\Delta a = a - \omega(a)\mathbf{1} and Δb=b−ω(b)1\Delta b = b - \omega(b)\mathbf{1} denote the centered observables representing the fluctuations of aa and bb about their mean values. The squared absolute value of the expectation of the product ΔaΔb\Delta a \Delta b is equal to the sum of the square of the covariance of aa and bb and the square of the expectation of their Lie bracket: ∣ω(ΔaΔb)∣2=Covω(a,b)2+ω(⟦a,b⟧)2|\omega(\Delta a \Delta b)|^2 = \text{Cov}_\omega(a, b)^2 + \omega(\llbracket a, b \rrbracket)^2 where Covω(a,b)=Re(ω(ΔaΔb))\text{Cov}_\omega(a, b) = \text{Re}(\omega(\Delta a \Delta b)) and ⟦a,b⟧=−i2(ab−ba)\llbracket a, b \rrbracket = -\frac{i}{2}(ab - ba).

theorem

Robertson's Gap Equals Gram Defect plus Squared Covariance

Let AA be a star-algebra and ω\omega be a state (a unital positive linear map) on AA. For any two observables a,b∈Aa, b \in A (self-adjoint elements), the difference between the product of their variances and the square of the expectation of their Lie bracket ⟦a,b⟧=−i2(ab−ba)\llbracket a, b \rrbracket = -\frac{i}{2}(ab - ba) is equal to the sum of the centered Gram defect and the square of their covariance: Varω(a)Varω(b)−ω⟨⟦a,b⟧⟩2=centeredGramDefectω(a,b)+Covω(a,b)2\text{Var}_\omega(a) \text{Var}_\omega(b) - \omega\langle \llbracket a, b \rrbracket \rangle^2 = \text{centeredGramDefect}_\omega(a, b) + \text{Cov}_\omega(a, b)^2 In this expression, Varω(x)\text{Var}_\omega(x) denotes the variance of observable xx, Covω(a,b)\text{Cov}_\omega(a, b) is the covariance of aa and bb, and the centered Gram defect is defined as Varω(a)Varω(b)−∣ω(ΔaΔb)∣2\text{Var}_\omega(a) \text{Var}_\omega(b) - |\omega(\Delta a \Delta b)|^2, where Δx=x−ω⟨x⟩1\Delta x = x - \omega\langle x \rangle \mathbf{1} represents the fluctuation of xx about its mean.

theorem

Robertson–Schrödinger Equality iff Zero Centered Gram Defect

Let ω\omega be a state (a unital positive linear map) on a C∗C^*-algebra AA, and let aa and bb be observables (self-adjoint elements) in AA. The Robertson–Schrödinger uncertainty relation is saturated, satisfying the equality Covω(a,b)2+ω(⟦a,b⟧)2=Varω(a)Varω(b),\text{Cov}_\omega(a, b)^2 + \omega(\llbracket a, b \rrbracket)^2 = \text{Var}_\omega(a) \text{Var}_\omega(b), if and only if the centered Gram defect of aa and bb in the state ω\omega is zero. Here, Covω(a,b)\text{Cov}_\omega(a, b) denotes the covariance, Varω(x)\text{Var}_\omega(x) the variance, and ⟦a,b⟧=−i2(ab−ba)\llbracket a, b \rrbracket = -\frac{i}{2}(ab - ba) the Lie bracket of the observables.

theorem

ω⟨⟦a,b⟧⟩2=Varω(a)Varω(b)\omega\langle\llbracket a, b \rrbracket\rangle^2 = \text{Var}_\omega(a) \text{Var}_\omega(b) iff Centered Gram Defect and Covariance are Zero

Let ω\omega be a state (a unital positive linear map) on a C∗C^*-algebra AA. For any two observables a,b∈Aa, b \in A (self-adjoint elements), let ⟦a,b⟧=−i2(ab−ba)\llbracket a, b \rrbracket = -\frac{i}{2}(ab - ba) be their Lie bracket, Varω(x)\text{Var}_\omega(x) be the variance of an observable, and Covω(a,b)\text{Cov}_\omega(a, b) be their covariance. The Robertson uncertainty relation saturates, meaning ω⟨⟦a,b⟧⟩2=Varω(a)Varω(b),\omega\langle\llbracket a, b \rrbracket\rangle^2 = \text{Var}_\omega(a) \text{Var}_\omega(b), if and only if the centered Gram defect is zero (Varω(a)Varω(b)−∣ω(ΔaΔb)∣2=0\text{Var}_\omega(a) \text{Var}_\omega(b) - |\omega(\Delta a \Delta b)|^2 = 0) and the covariance between the observables is zero (Covω(a,b)=0\text{Cov}_\omega(a, b) = 0). This equivalence holds generally, including cases where the variance is zero, without the need for division.

theorem

4Varω(a)Varω(b)≥14 \text{Var}_\omega(a) \text{Var}_\omega(b) \ge 1 for normalized commutator expectation ω(⟦a,b⟧)2=14\omega(\llbracket a, b \rrbracket)^2 = \frac{1}{4}

Let ω\omega be a state (a unital positive linear map) on a C∗C^*-algebra AA, and let aa and bb be observables (self-adjoint elements). If the expectation value of their Lie bracket ⟦a,b⟧=−i2(ab−ba)\llbracket a, b \rrbracket = -\frac{i}{2}(ab - ba) satisfies the normalization condition ω(⟦a,b⟧)2=14,\omega(\llbracket a, b \rrbracket)^2 = \frac{1}{4}, then the product of their variances satisfies the inequality 1≤4Varω(a)Varω(b),1 \le 4 \text{Var}_\omega(a) \text{Var}_\omega(b), where Varω(x)\text{Var}_\omega(x) denotes the variance of an observable xx in the state ω\omega.

theorem

ω⟨⟦a,b⟧⟩2=14  ⟹  Varω(a)>0∧Varω(b)>0\omega\langle\llbracket a, b \rrbracket\rangle^2 = \frac{1}{4} \implies \text{Var}_\omega(a) > 0 \land \text{Var}_\omega(b) > 0

Let ω\omega be a state (a unital positive linear map) on a C∗C^*-algebra AA. For any two observables a,b∈Aa, b \in A (self-adjoint elements), let ⟦a,b⟧=−i2(ab−ba)\llbracket a, b \rrbracket = -\frac{i}{2}(ab - ba) be their Lie bracket and Varω(x)\text{Var}_\omega(x) be the variance of an observable xx in the state ω\omega. If the expectation value of the Lie bracket satisfies ω⟨⟦a,b⟧⟩2=14,\omega\langle \llbracket a, b \rrbracket \rangle^2 = \frac{1}{4}, then the variances of aa and bb are strictly positive: Varω(a)>0andVarω(b)>0.\text{Var}_\omega(a) > 0 \quad \text{and} \quad \text{Var}_\omega(b) > 0.