Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedjudge pass (gpt-6-luna)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Fidelity classical measurement minimum

Statement

For states ρ,σ on a nonzero finite-dimensional complex Hilbert space H, let a finite POVM be a finite family (Ea)a=1n of nonnegative operators summing to IH. Then F(ρ,σ)=min⁡(Ea)∑a=1ntr⁡(ρEa)tr⁡(σEa). The minimum is attained by a projective measurement with d=dim⁡H rank-one outcomes.

Facts & Assumptions

Given: d=dim⁡H≥1 and density operators ρ,σ on H.

[L1]

Taking the fixed auxiliary space R=H, root fidelity is the maximum overlap of purifications on that common space (Fidelity Purification Characterization). The QC06 theorem proves this case as part of its every-R statement.

[L2]

A finite POVM consists of nonnegative effects summing to identity, and outcome probabilities are tr⁡(ρEa) (Finite POVMs and their outcome law). The QC04 definition body is present as a draft.

[F1]

Cauchy–Schwarz for the inner product gives ∣⟨x,y⟩∣≤∥x∥∥y∥ (Cauchy–Schwarz: ∣⟨x,y⟩∣≤∥x∥ ∥y∥, with equality exactly for dependent pairs).

[F7]

Density operators are nonnegative and have trace one (Density operator).

[L3]

The positive square root is unique, and the root formula F(ρ,σ)=tr⁡ρ σρ follows from the SVD trace-norm formula (A non-negative operator has a unique non-negative square root, The two root-fidelity trace expressions agree).

[F2]

U(d) is compact in its finite matrix coordinates (Unitary Group Compact), and bounded sequences of matrix coordinates have convergent subsequences by finite-dimensional Heine–Borel (Heine-Borel in Rn: with the Euclidean metric a subset of Rn is compact if and only if it is closed and bounded, and the proof by bisection uses no choice principle; the same holds on the real line). The compactness theorem body is a current QC06 draft; the Heine–Borel supplier is published.

[L4]

Partial trace is characterized by tr⁡((Tr⁡RZ)T)=tr⁡(Z(T⊗IR)) (Partial trace over a finite tensor factor, The partial trace is characterized by the trace pairing and is basis-independent); the QC02 definition and characterization theorem are present as drafts.

[L5]

The finite trace norm is a norm and satisfies the triangle inequality (Trace Norm Duality And Norm Properties). For a fixed matrix shape q×p, set m=min⁡(p,q). SVD gives ∥A∥1≤m(∑i,j∣Aij∣2)1/2, since each singular value is at most the square root of the sum of their squares and ∑jsj(A)2=tr⁡(A∗A)=∑i,j∣Aij∣2. Hence ∣∥A∥1−∥B∥1∣≤m(∑i,j∣Aij−Bij∣2)1/2 for same-shape matrices, proving the coordinate continuity needed below (Every linear map between finite-dimensional real or complex inner product spaces admits a singular value decomposition, Trace Norm).

[F3]

Every finite-dimensional inner-product space has an orthonormal basis (Every finite-dimensional real or complex inner product space has an orthonormal basis).

[F6]

A limit of nonnegative matrices is nonnegative because self-adjointness and each quadratic-form inequality pass to the coordinate limit (A non-negative operator is equivalently self-adjoint with nonnegative eigenvalues, a positive semidefinite matrix in an orthonormal basis, or an operator of the form S^*S).

Proof

1.1givenF7L1L2L4F1algebra

Let x,y be purifications attaining F(ρ,σ) as in [L1]. For a finite POVM (Ea), put xa=(Ea⊗I)x and ya=(Ea⊗I)y. Since ∑aEa=I, the adjoint identity gives ⟨x,y⟩=∑a⟨xa,ya⟩. By [L4], ∥xa∥2=tr⁡(ρEa) and ∥ya∥2=tr⁡(σEa). The triangle inequality and [F1] termwise give ∣⟨x,y⟩∣≤∑a∥xa∥∥ya∥=∑atr⁡(ρEa)tr⁡(σEa). Therefore every finite POVM overlap is at least F(ρ,σ).

1.2givenF7L3L5F2F5F6algebra

For arbitrary ρ,σ, set ρε=(1−ε)ρ+εI/d and σε=(1−ε)σ+εI/d for 0<ε≤1. Both are positive definite. The root fidelity is continuous on density operators: if An→A are density operators and Bn=An, then [F5] and [F7] give ∑i,j∣(Bn)ij∣2=tr⁡(Bn∗Bn)=1. Thus the coordinate matrices Bn are bounded. Every convergent subsequence has a nonnegative limit B by [F6] and B2=A, so uniqueness in [L3] gives B=A. If (Bn) did not converge to A, a subsequence staying a fixed positive distance away would have a convergent subsubsequence by [F2], contradicting that every such limit is A. Applying this to both state sequences shows their square roots converge; their product converges by the finite matrix-product formula, and its trace norm converges by [L5]. Hence F(ρε,σε)→F(ρ,σ).

2.1givenF7L3F4L6step 1.1algebrachoose

Suppose first that ρ is positive definite. By [F4], choose an orthonormal eigenbasis of ρ; its eigenvalues are strictly positive by positive definiteness, so ρ−1/2 is defined by the reciprocal square roots in this basis. Define M=ρ−1/2(ρ σρ)1/2ρ−1/2. Its quadratic form is nonnegative because it is the quadratic form of (ρ σρ)1/2 at ρ−1/2x, and direct multiplication gives MρM=σ. By [F4], choose an orthonormal eigenbasis of M, with rank-one projections Ej and eigenvalues μj≥0. If pj=tr⁡(ρEj), then tr⁡(σEj)=μj2pj, so the measurement overlap is ∑jμjpj=tr⁡(Mρ)=tr⁡(ρ σρ)=F(ρ,σ) by [L3] and [L6]. Thus a rank-one PVM attains the lower bound when ρ is faithful.

3.1givenstep 2.1step 1.2F2F3algebrachoose

Fix an orthonormal basis of H by [F3], and identify H with Cd. Let Qj be the rank-one projector onto its jth basis vector. Define Pd as ordered families (UQjU∗)j=1d for U∈U(d). Every ordered rank-one PVM has this form, since unit vectors spanning its mutually orthogonal one-dimensional ranges form an orthonormal basis. The map U↦(UQjU∗)j=1d is continuous in matrix coordinates, so Pd is compact by [F2]. For every ε>0, step 2.1 supplies a minimizing PVM in Pd. For n≥1, let Cn be the closure in Pd of all such minimizing PVMs for 0<ε<1/n. The Cn are nonempty nested compact sets, so their intersection is nonempty; fix P0 in it. The overlap function g(ε,P)=∑jtr⁡(ρεPj)tr⁡(σεPj) is continuous at every (0,P0), since trace is a finite coordinate sum and scalar square root is continuous on [0,∞). Given δ>0, choose a neighborhood of (0,P0) where g differs from g(0,P0) by less than δ, and choose n so large that 1/n is inside its ε-neighborhood and, by step 1.2, ∣F(ρε,σε)−F(ρ,σ)∣<δ for 0<ε<1/n. Since P0∈Cn, the definition of closure supplies a minimizing PVM P′ in that neighborhood for some 0<ε<1/n. For this pair, g(ε,P′)=F(ρε,σε), so ∣g(0,P0)−F(ρ,σ)∣<2δ. As δ>0 is arbitrary, g(0,P0)=F(ρ,σ). This gives an attaining rank-one PVM also in the singular case.

4.1step 1.1step 3.1∎

Step 1.1 bounds every finite POVM overlap below by F, and step 3.1 exhibits a finite rank-one PVM whose overlap equals F. This proves the stated minimum and attainment.

Depends on

Used by

Dependency tree · two levels

98 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources