Quantum Code Design Boosts Error Correction

Researchers are continually seeking methods to build robust and reliable quantum computers, and a significant hurdle remains the susceptibility of quantum bits (qubits) to errors. Omid Khosravani from the Duke Quantum Center at Duke University, Guillermo Escobar-Arrieta from the Escuela de Física at Universidad de Costa Rica, and colleagues, working with Kenneth R. Brown and Mauricio Gutierrez from Duke University and Universidad de Costa Rica respectively, have developed a novel approach to quantum error correction using heterogeneous codes composed of different qubit types. Their findings demonstrate that strategically placing qubits with varying noise characteristics – specifically, noisier qubits within the code’s core and cleaner qubits on its edges – can substantially improve error-correction thresholds and performance, exceeding those of conventional placements by a significant margin. This research, conducted in collaboration between the Duke Quantum Center, Duke University, Escuela de Física, Universidad de Costa Rica, and Escuela de Química, Universidad de Costa Rica, not only reveals a counterintuitive bias-inversion property within these codes but also offers a unifying theoretical framework with potential implications for the design of future, more resilient quantum computing architectures.

Tensor network decoding reveals qubit arrangement impacts on error correction

Maximum-likelihood tensor network decoding underpinned a sophisticated method for identifying and correcting errors in a quantum system, akin to using a complex algorithm to reconstruct a blurry image. This technique enabled simulations of the performance of quantum error-correcting codes with varying qubit arrangements, allowing assessment of how different qubit placements impacted the code’s ability to withstand noise and maintain quantum information.

Employing maximum-likelihood tensor network decoding, the process represents qubits and their relationships as a network of interconnected tensors to determine the most likely error configuration. Simulations of quantum error correction utilised codes with varying distances of 5, 7, and 9 qubits to assess performance under differing noise conditions.

A bond dimension of 16 was used, verified as sufficient for accurate threshold estimation, and thresholds were estimated using a standard critical exponent method. Two regimes were studied: differing error rates with identical bias, and identical error rates with differing bias, each with roughly equal numbers of qubit types. Error rates dropped significantly during the simulations. The team investigated how qubit placement affected code performance.

Strategic qubit placement enhances quantum error correction performance

Error rates fell to 0.6 per cent, a substantial improvement achieved by strategically arranging qubits within a quantum error-correcting code. This surpasses the typical threshold of approximately 0.2 attained with conventional, uniform qubit arrangements, previously limiting the scale and reliability of quantum computations. Placing noisier qubits in the core and cleaner qubits on the boundary of the code—or high-bias qubits on the boundary—yields significant gains in error correction.

This counterintuitive approach unlocks performance exceeding that of randomly arranged qubits, with the advantage growing exponentially with code distance. Naren Manjunath from the Perimeter Institute and colleagues observed that with a tenfold difference in error rates between qubit types, the ‘Bulk-Noisy’ placement yielded thresholds at least 0.43, while a reversed ‘Boundary-Noisy’ arrangement only reached 0.23. Furthermore, when qubits shared the same error rate but varied in bias—predictability of errors—placing high-bias qubits on the boundary increased the threshold from 0.292 to 0.360 at a bias ratio of 100. Thresholds exceeding 0.4 were achieved when employing a quantum error correction strategy utilising differing qubit qualities, compared to approximately 0.2 attained with standard arrangements. The logical error channel exhibited a bias-inversion property, becoming dominated by errors orthogonal to the initial physical noise, despite strongly Z-biased physical qubits.

Strategic qubit placement enhances error correction but relies on computationally intensive methods

Building quantum computers durable to errors demands key innovation beyond improving individual qubit quality. This work convincingly demonstrates that strategically mixing qubits—placing those prone to noise internally within error-correcting codes and more stable qubits on the periphery—can sharply boost performance. However, the simulations underpinning these gains rely on maximum-likelihood tensor network decoding, a computationally intensive technique.

These calculations, utilising this powerful but demanding method, cannot yet fully replicate the unpredictable behaviour of real quantum hardware. Despite this limitation, demonstrating a pathway to significantly improved error correction by strategically arranging qubits remains vital, guiding future experimental designs and prioritising research into more efficient decoding techniques for practical quantum computers.

Speed doubled with this new arrangement. The simulations also revealed a bias-inversion property, where the logical error channel becomes dominated by errors orthogonal to the initial physical noise. Strategically arranging qubits—the basic units of quantum information—can dramatically improve error correction within quantum computers. This placement strategy isn’t simply about average qubit quality; it exploits differences in error characteristics, including bias—the predictability of errors. This arrangement also induced a ‘bias-inversion’ effect, altering the nature of errors within the system, and further enhanced error correction when high-bias qubits were positioned on the boundary.