Manuscript En07-A@1 · En07-A.tex · example · 44229ce9
Example: Complete preservation without either local inverse
Classical records and exact destination examples › A channel-level check of the two-state control · lines 1067–1176 · proof included or linked
En07-A@1#ex:control, En07-A@1#eq:controlN, En07-A@1#eq:controlR, En07-A@1#eq:controlB, En07-A@1#eq:controlleak
Let , , and choose with . Define
and the isometry
Then
The complete output is exactly reversible. For , neither local output exactly recovers the code.
Proof. Orthogonal -labels give (eq:controlN). Put . Their trace distance is .
For any qubit decoder, its Bell overlap is
The two diagonal output tests pull back to complementary effects. Optimal binary discrimination therefore gives
The Bell-projector test yields error at least .
The two states are interchanged by . Symmetrizing an optimal discrimination measurement makes its two error probabilities equal to . Measuring and preparing the inferred logical basis state then realizes
Its Pauli probabilities are
Lemma lem:unitary-mixture gives error , proving (eq:controlR).
Tracing gives
The identity decoder has error . The outputs of and have trace distance . If a decoder has error , contraction and the triangle inequality give . This proves (eq:controlB).
The channel is Pauli covariant. Averaging a replacement state over Pauli conjugations cannot increase its distance from this channel and gives . That replacement is therefore optimal. Relative to the uniform Pauli representation of , the coefficient differences in the order are
Lemma lem:unitary-mixture gives
which is (eq:controlleak). Complete restoration is Theorem thm:global.
Displayed formulas in this passage
En07-A@1#eq:controlN · line 1082
En07-A@1#eq:controlR · line 1082
En07-A@1#eq:controlB · line 1082
En07-A@1#eq:controlleak · line 1082
unlabeled · line 1072
unlabeled · line 1078
unlabeled · line 1107
unlabeled · line 1117
unlabeled · line 1128
unlabeled · line 1132
unlabeled · line 1141
unlabeled · line 1164
unlabeled · line 1170