Abstract
We investigate cat codes that can correct multiple excitation losses and identify two types of logical errors: bit-flip errors due to excessive excitation loss and dephasing errors due to quantum backaction from the environment. We show that selected choices of logical subspace and coherent amplitude significantly reduce dephasing errors. The trade-off between the two major errors enables optimized performance of cat codes in terms of minimized decoherence. With high coupling efficiency, we show that one-way quantum repeaters with cat codes feature a boosted secure communication rate per mode when compared to conventional encoding schemes, showcasing the promising potential of quantum information processing with continuous variable quantum codes.
Introduction: The Challenge of Excitation Loss
In the quest for reliable quantum computing and communication, "excitation loss" (losing a photon) is a primary obstacle. Traditional Discrete Variable (DV) encodings address this by spreading a logical qubit across many physical qubits or bosonic modes, which is hardware-intensive and often inefficient.
Continuous Variable (CV) encodings, such as cat codes, offer a more efficient alternative by utilizing the high-dimensional Hilbert space of a single bosonic mode (e.g., a microwave cavity or optical fiber). However, cat codes face a fundamental "Goldilocks" problem:
- Bit-flip errors: If the coherent amplitude ($\alpha$) is too small, the states overlap, making them indistinguishable after loss.
- Dephasing errors: If the amplitude is too large, the environment "observes" the system more easily. When a photon is lost, the environment gains information about the state, causing a "back-action" that destroys quantum coherence.
Methods: Orthonormal Basis and Recovery
The authors model the system as a Lossy Bosonic Channel (LBC) and propose a specific Quantum Error Correction (QEC) recovery routine. Logical states are defined using superpositions of $2d$ coherent states arranged in a circle in phase space, allowing the code to correct up to $d-1$ photon losses. The recovery process involves a $Z_d$ measurement using a $d$-level ancilla to measure the photon number modulo $d$, followed by conditional loss compensation and amplitude restoration to return the state to its original size.
The Core Trade-off: Balancing Errors
A major insight of this research is that photon loss is not just a missing particle; it is a measurement by the environment. Because different logical states have slightly different average photon numbers, the environment "learns" which state the system is in, leading to dephasing. The authors demonstrate that the choice of logical subspace can be used to cancel out leading-order dephasing terms.
By deriving analytical expressions using the Lambert W function, the authors identified the optimal amplitude ($\alpha_o$) that minimizes the total error (measured by diamond distance). This optimization allows dephasing to be suppressed from $O(\Delta\pi^2/d^2)$ to $O([\Delta\pi^2/d^2]^2)$.
Key Results: Superior Performance in Quantum Repeaters
The theoretical optimization of cat codes translates into significant performance gains for long-distance quantum communication. In simulations of one-way quantum repeaters over transcontinental distances (>1000 km), cat codes achieved significantly higher Secure Key Rates Per Mode (SKRPM) than conventional DV schemes like Quantum Parity Codes (QPyC) or Quantum Polynomial Codes (QPC).
This performance boost is most prominent when coupling efficiency ($\eta$) is high (around 99.5%). At these levels, cat codes emerge as a clear winner for high-speed, long-distance quantum key distribution.
Limitations and Future Work
The primary limitation is the requirement for very high coupling efficiency; if the interface between the fiber and the repeater is lossy, the advantages of cat codes diminish. Furthermore, implementing the required $Z_d$ measurements and deterministic amplitude restoration requires high-fidelity quantum control. Future work will compare cat codes against other single-mode CV encodings, such as GKP (Gottesman-Kitaev-Preskill) codes, to determine robustness under various noise models.
Summary updated