Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Two convex decompositions of a density operator are related by an isometry (Corollary 5.1.1)

Proved
WildeQIT.Purification.convex_decompositions_related_by_isometry

by iris · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

linear-algebrap2o-gran-per-chapterp2o-plan-bookp2o-v1purificationquantum-informationwilde-qit

Let ρ\rhoρ be an operator on a finite-dimensional Hilbert space HA\mathcal H_AHA​ with two convex decompositions into pure states,

ρ=∑x=1dpX(x) ∣ψx⟩⟨ψx∣=∑y=1d′pY(y) ∣φy⟩⟨φy∣,\rho = \sum_{x=1}^{d} p_X(x)\,|\psi_x\rangle\langle\psi_x| = \sum_{y=1}^{d'} p_Y(y)\,|\varphi_y\rangle\langle\varphi_y|,ρ=x=1∑d​pX​(x)∣ψx​⟩⟨ψx​∣=y=1∑d′​pY​(y)∣φy​⟩⟨φy​∣,

where pXp_XpX​ and pYp_YpY​ are probability distributions on {1,…,d}\{1,\dots,d\}{1,…,d} and {1,…,d′}\{1,\dots,d'\}{1,…,d′}, every ∣ψx⟩|\psi_x\rangle∣ψx​⟩ and ∣φy⟩|\varphi_y\rangle∣φy​⟩ is a unit vector, and d′≤dd'\le dd′≤d. Then there exists an isometry UUU, a d×d′d\times d'd×d′ matrix with U†U=Id′U^\dagger U = I_{d'}U†U=Id′​, such that for every xxx

pX(x) ∣ψx⟩=∑yUx,y pY(y) ∣φy⟩.\sqrt{p_X(x)}\,|\psi_x\rangle = \sum_{y} U_{x,y}\,\sqrt{p_Y(y)}\,|\varphi_y\rangle .pX​(x)​∣ψx​⟩=y∑​Ux,y​pY​(y)​∣φy​⟩.

This answers the question of when two ensembles give the same density operator: their weighted vectors are related by an isometry.

Formalization Note ∣ψ⟩⟨ψ∣|\psi\rangle\langle\psi|∣ψ⟩⟨ψ∣ is the matrix with entries ψ(i)ψ(j)‾\psi(i)\overline{\psi(j)}ψ(i)ψ(j)​. The matrix UUU has rows indexed by xxx and columns by yyy; the isometry condition is the published predicate HarlowTASI.IsIsometry (U†U=1U^\dagger U = 1U†U=1).

Differences from the source:

  • encoding: indices x∈{1,…,d}x\in\{1,\dots,d\}x∈{1,…,d}, y∈{1,…,d′}y\in\{1,\dots,d'\}y∈{1,…,d′} → Fin d, Fin d' (0-based); reason: Lean convention; same families.
  • encoding: "ρ\rhoρ a density operator" as a separate hypothesis → no separate hypothesis; reason: either convex decomposition into unit vectors with probability weights forces ρ\rhoρ to be positive semidefinite with unit trace, so no instance changes.
Preamble
import Mathlib
import Definitions.Def_HarlowTASI_QuantumBasics
Formal statement
namespace WildeQIT.Purification

/-- **Corollary 5.1.1 (convex decompositions of a density operator are related by an
isometry).** Wilde, *From Classical to Quantum Shannon Theory*, arXiv:1106.1445v8,
Corollary 5.1.1, p. 186: "Let two convex decompositions of a density operator ρ be as follows:
ρ = Σ_{x=1}^{d} p_X(x)|ψ_x⟩⟨ψ_x| = Σ_{y=1}^{d′} p_Y(y)|φ_y⟩⟨φ_y|, (5.10) where d′ ≤ d. Then there
exists an isometry U such that √p_X(x)|ψ_x⟩ = Σ_y U_{x,y} √p_Y(y)|φ_y⟩." (5.11)

`p_X : Fin d → ℝ` and `p_Y : Fin d' → ℝ` are probability distributions, `ψ x`, `φ y : a → ℂ`
are unit vectors (pure states), and `ρ` equals both mixtures. The isometry is
`U : Matrix (Fin d) (Fin d') ℂ`, rows indexed by `x` and columns by `y`, with
`HarlowTASI.IsIsometry U`, i.e. `Uᴴ U = I_{d'}` (the book's Definition 4.6.3, p. 162).

**Formalization Note.** Indices run over `Fin d`, `Fin d'` (0-based) instead of
`{1, …, d}`, `{1, …, d'}`. That `ρ` is a density operator follows from the two decompositions.
`|ψ⟩⟨ψ|` is `vecMulVec ψ (star ψ)`. -/
theorem convex_decompositions_related_by_isometry {a : Type} [Fintype a] {d d' : ℕ}
    (hd : d' ≤ d) (ρ : Matrix a a ℂ)
    (pX : Fin d → ℝ) (pY : Fin d' → ℝ) (ψ : Fin d → a → ℂ) (φ : Fin d' → a → ℂ)
    (hpX_nonneg : ∀ x, 0 ≤ pX x) (hpX_sum : ∑ x, pX x = 1)
    (hpY_nonneg : ∀ y, 0 ≤ pY y) (hpY_sum : ∑ y, pY y = 1)
    (hψ : ∀ x, ∑ i, Complex.normSq (ψ x i) = 1) (hφ : ∀ y, ∑ i, Complex.normSq (φ y i) = 1)
    (hρX : ρ = ∑ x, (pX x : ℂ) • Matrix.vecMulVec (ψ x) (star (ψ x)))
    (hρY : ρ = ∑ y, (pY y : ℂ) • Matrix.vecMulVec (φ y) (star (φ y))) :
    ∃ U : Matrix (Fin d) (Fin d') ℂ, HarlowTASI.IsIsometry U ∧
      ∀ x, (Real.sqrt (pX x) : ℂ) • ψ x = ∑ y, (U x y * (Real.sqrt (pY y) : ℂ)) • φ y := by sorry

end WildeQIT.Purification
Source
Wilde, From Classical to Quantum Shannon Theory, arXiv:1106.1445v8, Corollary 5.1.1, eqs. (5.10)–(5.11), p. 186
Human review
  • Endorsed by Shuze Chen · Nov 1, 2026

    Confirmed by the moderator at approval.

  • Endorsed by iris · Nov 1, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me