Fundamental Quality Bound on Optical Quantum Communication

1. Problem Statement

The paper addresses the fundamental limits of transmitting quantum information over noisy optical channels (e.g., optical fibers or free-space links). Traditionally, quantum communication theory focuses on capacity (quantity): the maximum asymptotic rate of qubits that can be transmitted reliably. However, calculating the quantum capacity is notoriously difficult because:

• It is defined via a regularized formula involving an optimization over an infinite number of channel uses.
• The underlying entropic measures (like coherent information) are generally non-additive.
• For many physically relevant channels (e.g., thermal attenuators), the capacity is unknown or computationally intractable.

The authors propose a paradigm shift from quantity to quality. Instead of asking "how many qubits can be sent?", they ask: "How fast does the transmission error decay as the number of channel uses increases?" This leads to the study of the error exponent (specifically the zero-rate error exponent), which quantifies the reliability of the transmission when the rate approaches zero.

2. Methodology and Theoretical Framework

A. Teleportation-Simulable Channels

The authors focus on the class of teleportation-simulable channels. A channel $\mathcal{N}$ is teleportation-simulable if its action can be simulated by a Local Operations and Classical Communication (LOCC) protocol using its own Choi state (or quasi-Choi state) as a resource. This class includes all physically relevant Gaussian channels in optical communication.

• Reduction: For these channels, the problem of two-way assisted quantum communication is reduced to the problem of entanglement distillation from the Choi state of the channel.

B. Reverse Relative Entropy of Entanglement

The core theoretical tool is the reverse relative entropy of entanglement, defined as:
$$D(\text{SEP} \parallel \rho) := \inf_{\sigma \in \text{SEP}} D(\sigma \parallel \rho)$$
where $\text{SEP}$ is the set of separable states and $D(\sigma \parallel \rho)$ is the quantum relative entropy.

• Key Property: Unlike the standard relative entropy of entanglement (which requires regularization due to non-additivity), the reverse relative entropy is additive on product states. This allows for a "single-letter" characterization without regularization.
• Operational Interpretation: The paper extends previous finite-dimensional results to infinite-dimensional systems, proving that this measure exactly characterizes the zero-rate error exponent of entanglement distillation under non-entangling operations (NE).

C. Gaussian Systems and Convex Optimization

A major technical hurdle is that optimizing over separable states in infinite-dimensional systems is generally NP-hard. The authors overcome this for Gaussian states (the standard model for optical communication) by proving:

1. Equivalence: For Gaussian input states, the regular reverse relative entropy of entanglement coincides with the Gaussian reverse relative entropy (optimization restricted to Gaussian separable states).
2. Finite-Dimensional Reduction: This optimization reduces to a convex program over finite-dimensional covariance matrices with simple positive semi-definite (PSD) constraints. This makes the bound efficiently computable using standard interior-point solvers.

3. Key Contributions

1. Fundamental Upper Bound: The authors prove that for any teleportation-simulable channel $\mathcal{N}$, the zero-rate error exponent of two-way assisted quantum communication, denoted $Q_{\leftrightarrow, \text{err}}(\mathcal{N})$, is upper-bounded by the reverse relative entropy of entanglement of the channel's asymptotic Choi state:
   $$Q_{\leftrightarrow, \text{err}}(\mathcal{N}) \leq \liminf_{r \to \infty} D(\text{SEP} \parallel \rho_{\mathcal{N}}(r))$$
   This parallels the Pirandola et al. (2017) bound on capacity but uses the reverse relative entropy and applies to the error exponent rather than the rate.

2. Operational Interpretation in Infinite Dimensions: They rigorously extend the operational interpretation of the reverse relative entropy of entanglement to infinite-dimensional systems, linking it to the Sanov exponent of entanglement testing (distinguishing an entangled state from the set of separable states).

3. Efficient Computability: They provide a closed-form convex optimization program to compute this bound for Gaussian states, making it the first operational entanglement measure that is both meaningful and efficiently computable for continuous-variable systems.

4. Achievability Bounds: Using random-coding arguments based on the Choi state, they derive matching lower bounds (achievability) for specific channels, demonstrating that their upper bounds are tight or near-tight.

4. Results for Specific Channels

The authors derive explicit analytical expressions for the upper bound (UB) and lower bound (LB) for three prototypical one-mode Gaussian channels:

A. Thermal Attenuator Channel ($\mathcal{N}_{\lambda, n}$)

• Model: Beam splitter with transmissivity $\lambda$ and thermal noise $n$.
• Upper Bound:
  $$\text{UB} = \frac{n_{\text{sep}}(\text{arcoth}(n) - \text{arcoth}(n_{\text{sep}}))}{\ln 2} + \log_2 \sqrt{\frac{n^2 - 1}{n_{\text{sep}}^2 - 1}}$$
  where $n_{\text{sep}}(\lambda) = \frac{1+\lambda}{1-\lambda}$. The bound diverges for pure loss ($n=1$), consistent with the fact that pure loss channels can transmit error-free entanglement via dual-rail encoding.
• Lower Bound: Derived via random coding, showing the UB is not far from optimal, especially for high thermal noise.

B. Thermal Amplifier Channel ($\mathcal{A}_{\eta, n}$)

• Model: Amplification with gain $\eta$ and thermal noise $n$.
• Result: The error exponent follows a qualitatively identical expression to the attenuator, with $n_{\text{sep}}(\eta) = \frac{\eta+1}{\eta-1}$. It diverges for the quantum-limited amplifier ($n=1$).

C. Additive-Noise Gaussian Channel ($\mathcal{N}_{\mu}$)

• Model: Random Gaussian displacement with noise parameter $\mu$.
• Result:
  $$\text{UB} = \frac{2-\mu}{\mu \ln 2} + \log_2 \left(\frac{\mu}{2}\right)$$
  Valid for $0 \leq \mu \leq 2$. The lower bound matches the upper bound in the low-noise regime.

5. Significance and Impact

• New Benchmark: This work provides a new, computable benchmark for the performance of future quantum optical networks. Unlike capacity, which is often unknown, this quality bound can be calculated for any Gaussian channel.
• Theoretical Insight: It resolves the tension between operational meaning and computational tractability in continuous-variable entanglement theory. It proves that for Gaussian states, the "quality" of entanglement (error exponent) is governed by a single-letter, additive measure.
• Practical Application: The derived bounds allow engineers and physicists to quickly assess the reliability limits of optical links under various noise conditions without resorting to complex, non-regularized capacity calculations.
• Foundational Link: It establishes a rigorous duality between the standard relative entropy (governing capacity/quantity) and the reverse relative entropy (governing error exponents/quality), deepening the understanding of quantum resource theories.

In summary, the paper successfully shifts the focus from "how much" to "how well," providing a mathematically rigorous, computationally efficient, and operationally meaningful framework for analyzing the reliability of quantum communication in noisy optical environments.