Logical information leakage versus erasure flags

PDF p. 5: Discussion, first paragraph

The article argues that when a frequency shift lets the environment distinguish which state decayed, the information–disturbance tradeoff can increase the damage caused by decay. It says distinguishability is harmful when the resulting record depends on the logical state, and contrasts this with erasure conversion: an error flag is designed to reveal an error syndrome without revealing logical information. Here a decay is a specified physical error process, a syndrome is a label for an error outcome, and an erasure flag is a distinguishable outcome indicating a designated erased branch. These are the article's qualitative claims; they do not specify a numerical measure of damage or a universal inequality.

For a precise finite-dimensional comparison, encode a logical system by an isometry $V$ into a code subspace $C$ of a physical system, and let a quantum channel $\mathcal{N}$ describe the noise. A quantum channel is a completely positive trace-preserving linear map. Fix a finite Kraus representation $\mathcal{N}(X)={\sum }_a{E}_{a}X{E}_a^{†}$ and its complementary channel, which gives the environment output in the associated dilation. The full environmental record is logical-state dependent if two encoded logical density operators give different environment outputs. Exact correction means that a recovery channel after the noise returns every logical operator unchanged. Under these hypotheses, logical-state dependence of the full record rules out exact correction of the channel on the code.

A flag's distribution being independent of the logical input is a weaker statement than independence of the full environmental record. In the standard erasure channel, the flag probability is input independent, while the erased logical state need not be recoverable from the channel output. Thus the comparison concerns what the flag itself reveals; it does not assert that every environment associated with an erasure process contains no logical information.

Code restriction and decoded channel types

Complementary error records in a Kraus basis

Syndrome flags and logical-state-independent environmental records

Given: The finite-dimensional encoding, channel, Kraus representation and complementary record described above.

Goal: Establish the stated exact-correction consequence and distinguish flag statistics from the full record.

1.1 Suppose exact correction exists. The scalar-overlap conditions follow by [F3], and [F4] then gives the same full environmental record for every density operator supported on the code. Encoded logical density operators are supported on that code. Consequently, if two such inputs give different full records, exact correction cannot exist. [given, F1, F2, F3, F4]

2.1 By [F5], logical-state-independent flag probabilities alone do not imply independence of the full record. In the specified erasure channel the flag probability is the same $ϵ$ for every input; at $ϵ=1$ the channel output is the fixed erasure state for every input, so a decoder cannot recover two distinct logical states. This illustrates why the flag's lack of logical information does not by itself establish recovery or characterize the entire environment. [F1, F5, step 1.1] ∎

The article's qualitative information–disturbance and erasure-conversion comparison is attributed above. The finite-dimensional conclusion follows from the named mathematical suppliers and is not a quantitative claim that every increase in distinguishability causes a specified increase in recovery error.