Quantum conditional mutual information and channel capacity

Abstract: Information measures acquire operational meaning through coding theorems. The quantum conditional mutual information (QCMI) is nonnegative due to strong subadditivity, yet a direct connection with channel coding has remained elusive. In this work, we propose a quantum communication task-conditional quantum communication-that fills this gap. We show that the optimal rate for establishing quantum correlation between two parties, assisted by a third system, is given by half the QCMI. This result naturally extends the classical key generation capacity of Csiszár and Ahlswede to the quantum domain. We place our model within the family tree of quantum protocols and compute the conditional capacity for several example channels. Our results provide new insights for code design in reliable quantum information processing.

• The paper introduces an operational quantum channel capacity defined as half the quantum conditional mutual information (QCMI).
• It rigorously links QCMI to conditional communication by employing side information from a third system to enhance reliable transmission.
• Numerical examples and proof techniques extend classical wiretap models to inform novel code designs for quantum error correction.

Quantum Conditional Mutual Information and Channel Capacity

Introduction and Motivation

The paper "Quantum conditional mutual information and channel capacity" (2606.25264) addresses the operational significance of quantum conditional mutual information (QCMI) in the context of quantum communication protocols. While QCMI, defined by $I(A:B|C) = S(AC) + S(BC) - S(C) - S(ABC)$, is a fundamental information-theoretic quantity known for its non-negativity via strong subadditivity, its direct connection to channel capacities has historically been elusive. The authors close this gap by proposing a conditional quantum communication task and rigorously demonstrating that its optimal rate is given by half the QCMI. This construction extends the classical work on key generation and conditional capacities (e.g., Csiszár-Ahlswede's capacity for wiretap channels) into the quantum regime and shows that QCMI serves as a natural capacity measure when a third system, $C$, provides side information.

Quantum Conditional Mutual Information: Properties and Examples

The QCMI quantifies the residual correlations between subsystems $A$ and $B$ when conditioned on a third system $C$. Its non-negativity is a direct consequence of strong subadditivity, and equality to zero characterizes quantum Markov chains. QCMI also connects closely to the Petz recovery map, with operational consequences in recovery and data-processing inequalities. The paper considers six paradigmatic tripartite quantum states ($\psi_1$ to $\psi_6$) to illustrate various configurations of mutual and conditional information.

Figure 1: Venn diagram for the mutual information of the six states $\psi_1$ to $\psi_6$ (from left to right, top to bottom).

States such as $\psi_1$ (post-measurement state in entanglement swapping) exhibit maximal $C$0, representing optimal conditional correlations, whereas states like $C$1 (pre-measurement) have vanishing $C$2, demonstrating that measurement on $C$3 is necessary to unlock conditional correlations. These concrete examples make explicit that QCMI detects and quantifies the presence of conditional quantum correlations and highlights its operational meaning beyond abstract entropy inequalities.

Conditional Quantum Channel Capacity: Main Theorem and Protocol

The central result of the paper is the introduction of conditional quantum channel capacity as an operational quantity. For a broadcast quantum channel $C$4, the capacity for establishing quantum correlations between $C$5 and $C$6, with the assistance of system $C$7, is shown to be:

$C$8

where optimization is over the input ensemble to $C$9, with $A$0 taken as the purifying system and $A$1 an external reference system.

The proof leverages quantum generalizations of the packing (reliable transmission) and covering (privacy against $A$2) lemmas, with $A$3 playing a dual role: it may act as an assistant (side channel) to $A$4 or as an eavesdropper, echoing the structure of classical wiretap and key generation models. Critically, an ebit (entanglement bit) can always simulate a debit (common randomness), but a debit is never secure in the quantum regime—a consequence of quantum monogamy of entanglement.

This capacity reduces to well-known quantities in certain limits:

• If $A$5 is the environment, or when $A$6 form a quantum Markov chain, it reduces to the private channel capacity $A$7.
• If $A$8 is used as an input (environment-assisted/entanglement-assisted protocol), the capacity reduces to entanglement-assisted channel capacity.

The model thus interpolates between existing quantum capacities, generalizing them and providing a strict operational interpretation for QCMI as a rate measure.

Relationship to Quantum Protocol Family Tree

By casting conditional quantum communication in the language of resource inequalities, the paper situates its results within the broader family of quantum information protocols (merging, distillation, and channel coding). The new model occupies a decisive place alongside standard protocols such as state merging, state redistribution, and entanglement distillation. The resource requirements and achievable rates unify and clarify the operational landscape, showing, for instance, that in the presence of side information ($A$9), quantum and classical capacities relate by a factor of two, lining up with prior results in entanglement-assisted settings.

Applications: Example Channels and Code Design

The paper provides concrete scenarios where the conditional capacity $B$0 surpasses the standard quantum mutual information-based capacity. Notably, in correlated noise settings (e.g., the scenario of quantum repeaters or environment splitting), the auxiliary output $B$1 can carry syndrome or diagnostic information. Such an architecture enables higher communication rates between $B$2 and $B$3 by using $B$4 to correct or account for correlated errors, a benefit not captured by conventional (marginalized) channel capacities.

For example, by constructing a broadcast channel from two correlated noise processes (bit-flip and phase-flip), the conditional mutual information $B$5 can exceed the bare $B$6, with numerically significant gains. This has substantial implications for quantum error correction, quantum key distribution, and quantum networking, particularly in architectures with environmental monitoring or pilot tones.

Theoretical and Practical Implications

The main implication is the explicit operational role of QCMI as a capacity for conditional quantum communication. This reframes and extends the function of QCMI from a property of quantum states (recovery, Markov chains, and entanglement measures such as squashed entanglement) to a capacity that governs achievable rates in channel coding settings with auxiliary or conditional systems.

Practically, this suggests new code designs and error correction strategies that treat passive syndrome outputs or monitorable environments as side channels to be exploited, rather than nuisances to be suppressed. This is a departure from the traditional assumption of adversarial or fully inaccessible noise and paves the way for codes with improved rates and thresholds in realistic, correlated noise environments.

The framework unifies disparate areas of quantum information—theory of privacy, broadcast channels, quantum networks, and code design—under the language and machinery of conditional mutual information. It also suggests new benchmarks for protocols like entanglement swapping, quantum repeaters, and multi-user quantum key distribution.

Conclusion

This work establishes QCMI as a single-letter, operational quantum channel capacity in the presence of side information, filling a notable gap in the quantum information theory literature. The conditional communication model provides a rigorous foundation for assessing the ultimate limits of quantum coding with auxiliary outputs. It implies that existing architectures in distributed quantum computing and quantum networking can benefit from structured, QCMI-based code design, especially where diagnostics and environmental monitoring are available. Future developments will likely explore explicit code constructions achieving these rates, robustness in realistic physical networks, and further connections to multipartite and network quantum information theory.