Sheron Blair, Francesco Arzani, Giulia Ferrini, and Alessandro Ferraro demonstrate error probabilities below the threshold for fault-tolerant memory using concatenated GKP-stabilizer codes. The researchers numerically optimized the generation of GKP states from vacua using circuits comprised solely of universal continuous-variable gates, addressing a long-standing question in the field. This work builds directly on the 1999 Lloyd and Braunstein paper in Phys. Rev. Lett. 82, 1784, establishing a lineage to foundational work in continuous-variable quantum computation. Achieving this demonstrates that universal CV gates can enable encoded fault tolerance, a critical step toward scalable quantum memory.

Continuous-Variable Quantum Computation via Lloyd-Braunstein Universality

GKP code states of sufficient quality for fault-tolerant quantum memory can now be prepared using a finite set of gates, a development that addresses a fundamental gap in continuous-variable (CV) quantum computation. The work focuses on demonstrating encoded universal computation and establishing the LB model as a fully fledged framework for fault tolerance, rather than achieving full fault-tolerant quantum computation at this stage.

The LB universality hinges on the ability to host a finite set of parametrized gates capable of approximating any dynamics with a polynomial Hamiltonian in the field variables. Researchers established this by using a SNAIL-terminated nonlinear resonator, a system that allows for the creation of gates essential for single-mode and, with the addition of a two-mode gate, multimode LB universality.

An important element of this approach is the adoption of the Gottesman-Kitaev-Preskill (GKP) state as the fiducial initial state and measurement set, providing a standardized foundation for computation within the LB model. The team’s strategy involved generating GKP-encoded qubits using the LB universal gate set, defined by a specific set of equations, to demonstrate the compatibility of LB universality with fault tolerance. Their primary objective was to prove that imperfections introduced during state preparation within the LB framework do not inherently prevent the realization of a fault-tolerant quantum memory.

This analysis intentionally excludes the examination of how gates might contribute to error propagation, focusing instead on the initial state preparation as the limiting factor for memory stability. The results show that the LB model is not only computationally universal but also capable of supporting the necessary conditions for a fault-tolerant quantum memory.

Gottesman-Kitaev-Preskill Encoding for Fault Tolerance

Generating Gottesman-Kitaev-Preskill (GKP) states using only universal continuous-variable (CV) gates has long been a theoretical goal, and recent work demonstrates the numerical optimization of these states from vacuum inputs, a key step toward realizing practical fault-tolerant quantum memory. The ability to create GKP states solely with universal CV gates addresses a key limitation in the field, as prior approaches often relied on resources or techniques incompatible with a purely continuous-variable system.

This achievement establishes a pathway for encoded universality and fault tolerance within the Lloyd-Braunstein model, a framework initially proposed in 1999. The demonstrated state preparation uses circuits designed to minimize errors, focusing on Gaussian inputs and measurements alongside classical feedforward to maintain unitarity and avoid introducing extraneous resources.

Researchers computed the probability of uncorrectable errors by measuring the quadratures of the generated GKP states, applying the Glancy-Knill condition as an upper bound for shift errors correctable through Steane-type GKP error correction. Achieving fault tolerance requires the error probability to fall below a threshold value, dependent on the specific discrete-variable error correcting code operating at the logical level and the noise model employed.

This result is significant because it confirms that the Lloyd-Braunstein approach, despite its reliance on approximations in unbounded operator constructions, can indeed generate states suitable for robust quantum computation. The team isolated the effect of encoding imperfections by assuming ideal implementation of all LB universal gates used in the error correction procedure, allowing them to focus specifically on the quality of the generated GKP states.

The work deliberately avoids methods that fall outside the scope of the LB approach, such as those involving auxiliary qubits, discrete-variable measurements, or ground-state preparation. Protocols relying on conditional CV measurements were also excluded due to their non-deterministic nature, which poses challenges for encoded fault tolerance and the associated overhead. This focus on a purely CV gate set emphasises the versatility of the Lloyd-Braunstein model and its potential for scalability, as it does not require the integration of disparate quantum technologies.

The researchers’ calculations build upon earlier work demonstrating the potential of infinite-energy GKP states as qubits in a surface code for fault-tolerant quantum memory. By demonstrating the generation of GKP states with sufficient quality, even with finite energy and inherent imperfections, they establish that the LB universal gate set can support the necessary conditions for quantum error correction.

LB Universal Gate Set: Displacements, Squeezing, and Kerr Gates

The construction of approximate GKP states from vacuum using only displacements in position, squeezing, rotations, and Kerr gates demonstrates a capability previously lacking analytical proof, according to recent work. These circuits use gate parameters organized as vectors, with each element defining the settings for a specific block, beginning with a vacuum input and targeting the square GKP state as defined in prior research.

A single block of the circuit consists of a sequence of gates, displacements in both position and momentum, a Kerr gate, and squeezing, acting on the input state. The generated states, specifically their position probability distributions, were compared to the target GKP state for a value of 10, revealing the circuit’s ability to approximate the desired quantum state.

Further analysis, detailed in the Supplemental Material, extends this comparison to momentum space and the Wigner function, providing a comprehensive assessment of the generated state’s fidelity. This adherence to a purely CV approach is significant because it addresses a longstanding question: can a finite set of parametrized gates, without resorting to techniques outside the Lloyd-Braunstein (LB) framework, achieve encoded fault tolerance? The researchers’ approach deliberately excludes methods involving conditional CV measurements, which introduce non-deterministic behavior problematic for encoded fault tolerance and associated overhead.

Instead, they implemented an inverse-SUM gate acting on inputs, followed by momentum-basis measurement and classical feedforward to a displacement gate, ensuring all circuit elements remain within the defined LB universal model. “All of these methods lie outside the LB approach, which is our focus,” the researchers state, emphasizing the deliberate constraints of their investigation.

The work builds on earlier findings that identified finite squeezing of auxiliary GKP states as the dominant noise source when concatenating GKP error correction with the surface code for qubit-level error correction. Nohet et al. demonstrated this in a 2023 paper, providing important context for the current research. By focusing on generating GKP states with sufficient quality using only the LB gate set, the team aimed to address this critical limitation.

The numerical optimization process involved adjusting gate parameters to minimize the difference between the generated and target states, a complex task requiring significant computational resources. The researchers’ calculations confirm that the defined set of CV gates can, in fact, generate approximate GKP states from vacuum, a result not previously guaranteed analytically. This capability is essential for realizing LB universality and, consequently, fault tolerance within the CV model.

The ability to generate these states is not merely theoretical; the demonstrated error probabilities fall below the threshold required for concatenated GKP-stabilizer codes, indicating a potential breakthrough in achieving practical quantum memory. The team’s approach offers a deterministic, unitary and fully CV pathway to encoded fault tolerance. The researchers’ focus on a specific encoding, and the demonstration of its feasibility, represents an advancement in the field, offering a clear path toward realizing fault-tolerant quantum information processing.

Error Thresholds for Fault-Tolerant Quantum Memory

Establishing a reliable quantum memory demands surpassing specific error thresholds, and recent work demonstrates that GKP-encoded qubits generated using a specific set of continuous-variable gates can meet these requirements. Calculations reveal that error probabilities fall below 0.0174 at a squeezing level of 10.5 decibels, an important benchmark for achieving fault tolerance. This finding indicates that imperfections inherent in the state preparation process do not automatically preclude the possibility of a stable quantum memory.

To confirm the viability of this approach, the team sought to demonstrate that the generated GKP states, despite being finite and imperfect, possess sufficient quality to satisfy the conditions for fault tolerance. The work represents the first error analysis of approximate GKP states prepared using a physically realizable procedure within the Lloyd-Braunstein framework.

Using a twirling approximation, the researchers identified a 9.9-decibel threshold for fault-tolerant quantum memory in terms of required squeezing, which was then converted into an equivalent error probability threshold. This threshold aligns closely with a 10.5-decibel threshold previously found for a measurement-based fault-tolerant quantum memory. The team’s calculations confirm that the generated states, when used as qubits in a surface code, exhibit error probabilities below this threshold, suggesting a viable pathway toward practical quantum memory.

The error probability for target states closely mirrored that of the twirling-approximated states, validating the accuracy of the approximation method used. The study’s conclusions extend beyond simply achieving a low error rate. They demonstrate the ability to generate encoded GKP qubits suitable for universal encoded computation and, critically, that these qubits can support a fault-tolerant quantum memory.

The team employed Monte Carlo methods to estimate the logical error probability of the overall concatenated error correction procedure, a technique that allowed for a detailed assessment of the system’s performance. Although the current work focuses on fault tolerance within the preparation of GKP states, and does not address error propagation during gate operations, it establishes a significant foundation for future research. Further analysis is needed to fully establish fault-tolerant quantum computation, as even perfect gates can introduce errors when acting on states with finite squeezing and non-unit fidelity.

For example, applying a perfect gate to a generated GKP state with a fidelity of 0.97 yields an error probability of 0.0313, compared to 0.0197 for the original state. Despite these considerations, the generated states remain well below the established fault-tolerance threshold. The authors write that their analysis focuses on the error correction of the prepared GKP states, and does not encompass the complexities of error propagation during quantum computation. However, they emphasize that the demonstrated ability to generate high-quality GKP states within the Lloyd-Braunstein framework represents a step toward realizing practical fault-tolerant quantum technologies.

👉 More information
🗞 Toward fault-tolerant quantum memory with universal continuous-variable gates
✍️ Sheron Blair, Francesco Arzani, Giulia Ferrini and Alessandro Ferraro
🧠 DOI: http://link.aps.org/doi/10.1103/45v3-h82q

More like this