TL;DR
This paper introduces new ways to measure how well quantum information can be stored and transmitted over time, even when the storage medium is noisy. They define "subdivision capacities" that account for the ability to intervene and correct errors during storage, showing that under certain conditions, the ideal storage capacity can be achieved.
What Problem Does This Paper Address?
Quantum information needs to be stored reliably, but quantum memories are susceptible to noise that degrades the information over time. Traditional quantum capacity measures optimize encoding and decoding at the beginning and end of a transmission, but they don't account for the possibility of actively intervening and correcting errors during the storage period. This is particularly relevant for quantum memories where information is stored in time, not transmitted through space. The paper addresses this by developing new capacity measures that incorporate the ability to subdivide the time evolution and apply intermediate quantum operations, effectively allowing for continuous error correction.
Methods
The authors introduce two main concepts: Quantum Subdivision Capacities and Continuous-Time Quantum Coding.
- Quantum Subdivision Capacities ($Q_C(t\mathcal{L})$): These capacities quantify the optimal rate of information storage when the noisy time evolution, generated by a Liouvillian $\mathcal{L}$, can be broken into arbitrary many parts. Intermediate "coding channels" from a predefined set $C$ can be applied between these parts. The authors analyze this for different choices of $C$, including arbitrary quantum channels ($C = \text{ch}$), channels generated by quantum dynamical semigroups ($C = \text{sg}^*$), and unitary channels ($C = \text{un}$).
- Continuous-Time Quantum Coding: This extends the subdivision idea to a continuous setting. Instead of discrete subdivisions, they consider adding supplementary "coding Liouvillians" ($\mathcal{L}_C$) that act alongside the noise Liouvillian ($\mathcal{L}$) on multiple copies of the system. This allows for continuous error correction. They analyze this for time-dependent coding Liouvillians and then focus on a specific case with time-independent, quasi-local coding Liouvillians.
Key techniques include:
- Decoupling approach: Used to show that certain rates are achievable by demonstrating that the "environment" (representing noise) can be made to forget the information.
- Coherent information: A crucial quantity used to bound capacities and prove achievability.
- Schumacher compression: Employed to generate "almost pure" ancilla states from fixed points of the noise, which are then used in coding schemes.
- Entropy growth arguments: Used to derive upper bounds on capacities, particularly for unitary coding.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Intervention is key: The ability to intervene and correct errors during the storage of quantum information can significantly improve the achievable rates, sometimes reaching the ideal capacity.
- Subdivision capacity can match ideal capacity: For broad classes of intermediate coding operations (arbitrary channels or channels from quantum dynamical semigroups), the subdivision capacity can achieve the maximum possible value, $\log(d)$, for a $d$-dimensional system.
- Unitary interventions have limits: When only unitary operations are allowed for intermediate coding, the subdivision capacity can be arbitrarily small, especially for certain types of noise (like depolarizing noise). This highlights the importance of the type of intervention allowed.
- Dissipation can help: Adding purely dissipative terms to the noise evolution can sometimes increase the quantum capacity, a counter-intuitive but important finding.
- Continuous coding schemes are powerful: Continuous error correction, by adding coding Liouvillians, can overcome limitations imposed by noise, even for channels that would otherwise lead to complete information loss.
Key Results
- $Q_{\text{ch}}(t\mathcal{L}) = \log(d)$: The subdivision capacity with arbitrary intermediate quantum channels achieves the maximum possible capacity.
- $Q_{\text{sg}^*}(t\mathcal{L}) = \log(d)$: The subdivision capacity with intermediate channels generated by quantum dynamical semigroups also achieves the maximum capacity.
- Upper bound for unitary interventions: For depolarizing Liouvillians, $Q_{\text{un}}(t\mathcal{L}) \leq \log(d) - (1 - e^{-rt}) S(\rho_0)$, showing it can be less than $\log(d)$.
- Lower bounds for unitary interventions: For Liouvillians with pure fixed points, $Q_{\text{un}}(t\mathcal{L}) > 0$, demonstrating that even with unitary constraints, some information can be reliably stored.
- Continuous coding can improve capacity: For certain noisy channels, continuous coding schemes can achieve rates higher than what's possible without intervention, even overcoming scenarios where the standard capacity is zero.
Limitations and Future Work
- General unitary case: While lower bounds for unitary interventions are provided, a general closed-form expression for $Q_{\text{un}}(t\mathcal{L})$ remains an open problem.
- Choice of coding sets: The paper explores specific sets $C$ for coding channels and Liouvillians. Investigating other choices and their impact on capacities is a direction for future work.
- Classical vs. Quantum: The paper hints at similar phenomena for classical capacities and asks if they differ significantly.
- Zero-error capacities: The behavior of zero-error capacities in these scenarios is also an open question.
- Practical implementation: While the theoretical framework is developed, practical implementation of these continuous coding schemes and the generation of ancilla states would be a significant engineering challenge.
Summary updated