Source-linked AI summary
Quantum data processing and error correction
Benjamin Schumacher, M. A. Nielsen
TL;DR
Noisy quantum channels raise the question of how much quantum information their dynamics retain. The paper defines coherent information for a channel and shows that subsequent processing cannot increase it.
Problem
The paper addresses how much quantum information is retained when Q undergoes noisy channel dynamics.
Source: §6 · p. 14Method
The paper defines coherent information as an intrinsic quantity for the input state and channel.
Source: §4 · p. 9Results
The coherent information in the first stage cannot be increased by the additional dynamics of a second stage.
Source: §5 · p. 13Takeaways & Limitations
The paper considers whether a subsequent quantum evolution can make a channel’s overall entanglement fidelity unity, its definition of perfect error correction.
Source: §6 · p. 14The error-correction discussion uses a particular input state ρQ and defines perfect restoration with respect to the subspace supporting that state.
Source: §6 · pp. 14, 17
Abstract
from arXiv · showhide
This paper investigates properties of noisy quantum information channels. We define a new quantity called {\em coherent information} which measures the amount of quantum information conveyed in the noisy channel. This quantity can never be increased by quantum information processing, and it yields a simple necessary and sufficient condition for the existence of perfect quantum error correction.
1 Introduction
The paper frames noisy-channel evolution as a trace-preserving, completely positive map and uses reference and environment systems to describe its effects. It relates this map to equivalent unitary and operator-sum representations.
1 Introduction
- A noisy quantum channel maps initial density operators to final states while preserving trace and positivity.When extended to a reference system, the map must also preserve positivity; this is complete positivity.Source: §1 · pp. 2–3
- A reference system purifies Q’s initial state, allowing the channel’s action on Q to be considered within a joint system RQ.The reference is introduced as a mathematical device and remains isolated from the channel dynamics.Source: §1 · pp. 2–3
- Every allowable channel has equivalent descriptions as a unitary evolution on a larger system or as a normalized operator-sum representation.These representations are not unique.Source: §1 · pp. 3, 5
- The environment can be a mathematical device rather than the channel’s actual physical surroundings; only the dynamics of Q matter.Many choices of environment and unitary operator can represent the same channel.Source: §1 · p. 4
2 Entanglement fidelity
Entanglement fidelity is an intrinsic measure of how well a channel preserves a state’s entanglement with a reference. It bounds ensemble-average fidelity and connects perfect or near-perfect fidelity to behavior on the input state’s support.
2 Entanglement fidelity
- Entanglement fidelity is intrinsic to ρQ and the channel, measuring how faithfully the entangled state is preserved.Its value does not depend on the particular purification of ρQ.Source: §2 · p. 6
- Entanglement fidelity never exceeds an ensemble’s average fidelity, so it is a lower bound for that average.This relation follows by representing ensemble states as relative states prepared by measurements on the reference.Source: §2 · p. 7
- Entanglement fidelity equals one if and only if every pure state in the support of ρQ has fidelity one.The paper also gives a quantitative lower bound when all such pure-state fidelities are close to one.Source: §2 · p. 7
- The passage states Fe ≥ 1 − (3η/2) and says Fe = 1 if and only if F = 1 for every pure state in the support of ρQ.Source: §2 · p. 7
3 Entropy production
Entropy production is an intrinsic measure of information exchanged between Q and the rest of the world during channel evolution. The quantum Fano inequality relates it to entanglement fidelity.
3 Entropy production
- Entropy production can be interpreted as the post-evolution entropy of RQ, or equivalently of E when the environment starts pure.It remains intrinsic to ρQ and the channel.Source: §3 · p. 8
- Entropy production measures information exchanged between Q and the rest of the world during the channel evolution.It is not generally equal to the entropy change of Q or of the actual physical environment.Source: §3 · p. 8
- If Fe = 1, the quantum Fano inequality implies Se = 0.The inequality relates entanglement fidelity and entropy production for a system with d complex dimensions.Source: §3 · pp. 8–9
4 Coherent quantum information
Coherent information measures retained quantum coherence through the entanglement between reference R and system Q after channel evolution. It is bounded above by the initial entropy of Q, with equality under a stated independence condition.
4 Coherent quantum information
- Coherent information measures the degree of entanglement retained between R and Q, or the quantum coherence retained by the process.It may be positive, negative, or zero.Source: §4 · p. 9
- Coherent information cannot exceed the initial entropy of Q, which measures the initial entanglement of R and Q.Equality holds if and only if the final state of R and E is a product state.Source: §4 · p. 10
- The initial entropy of Q also measures the resources needed to faithfully store the initial RQ entanglement.This identifies the bound’s initial-state quantity with a storage resource.Source: §4 · p. 10
5 Quantum data processing inequality
The paper proves that coherent information cannot increase when a quantum process is followed by further processing. The proof applies strong subadditivity to the environments involved in successive stages.
5 Quantum data processing inequality
- Classical mutual information between a channel’s input and output likewise cannot be increased by processing the output.The paper presents this as the classical data processing inequality.Source: §5 · p. 10
- Applying strong subadditivity to the compound system RE1E2 yields the inequality that establishes coherent information’s nonincrease across stages.The resulting relation is expressed using the entropy productions of the first stage and the overall process.Source: §5 · pp. 12–13
- Overall entropy production is not generally the sum of the stages’ entropy productions and can be less than either individual stage’s production.The paper denotes the first-stage and overall entropy productions by Se1 and Se12.Source: §5 · p. 13
- Coherent information in the first stage cannot be increased by additional dynamics in the second stage.This is the quantum data processing inequality.Source: §5 · p. 13
6 Error correction
The paper shows that perfect error correction is possible exactly when the channel’s coherent information equals the input entropy. It constructs a recovery process and extends its guarantee to pure states in the input state’s support.
6 Error correction
- Perfect error correction means the overall process has entanglement fidelity Fe12=1, restoring Q’s entanglement with reference system R.The paper notes this is equivalent to every pure state in the subspace supporting ρQ having fidelity one.Source: §6 · p. 14
- Perfect error correction is possible if and only if the channel’s coherent information equals the input entropy S(ρQ).The paper establishes equality as both a necessary and a sufficient condition.Source: §6 · pp. 14–15
- The constructed recovery process restores the original entangled state, giving the overall process entanglement fidelity Fe12=1.The paper describes this as a perfect error-correction scheme.Source: §6 · pp. 15, 17
- The recovery scheme works for every pure state in the support of ρQ and any other entangled state with the same support.Although constructed using a particular input state, the guarantee extends across that support.Source: §6 · p. 17
- Classically, an input can be reconstructed from a channel’s output with zero error if and only if H(X)=I(X:Y).The paper compares this criterion with its quantum error-correction result.Source: §6 · p. 17
7 Remarks
The paper presents coherent information as a measure of distinctively quantum information conveyed through a channel, and says it cannot increase under further dynamics. It also gives a necessary and sufficient condition for perfect quantum error correction: S(ρQ) = Ie.
7 Remarks
- The paper defines entanglement fidelity, entropy production, and coherent information as intrinsic quantities; a quantum Fano inequality relates the first two.Entanglement fidelity is also closely related to input-output fidelities for pure states of Q.Source: §7 · p. 18
- Coherent information measures the distinctively quantum information passing through a channel, and in general S(ρQ) ≥ Ie.Source: §7 · p. 18
- Coherent information cannot increase under further dynamics; for successive independent processes, Ie1 ≥ Ie12.Source: §7 · p. 18
- Perfect quantum error correction is possible if and only if S(ρQ) = Ie, when the environment has obtained no information about Q's state.At equality, ρRE′ = ρR ⊗ ρE′, so no observable on E alone provides information about the outcome of a measurement on R or which ensemble state is present.Source: §7 · pp. 17–18
- The authors expect coherent information to play a role analogous to classical mutual information, but note its time asymmetry and that it cannot be positive for classical channels.Source: §7 · p. 18
- The paper develops intrinsic quantities using an extrinsic construction with a reference system R to purify the input and an environment E for unitary dynamics.Source: §7 · p. 19