IonQ Proposes Distributed Memory Architecture

Quantum memories represent a crucial component for realising practical quantum computers, yet scaling these systems presents significant challenges, particularly in maintaining data integrity against errors. Edwin Tham, Min Ye, and Ilia Khait, all from IonQ Inc., alongside colleagues, now demonstrate a novel architecture for distributed quantum memories that overcomes some of these limitations. Their approach distributes logical information across an array of interconnected qubit modules, utilising a simple ‘cyclic shift’ connection to maintain performance even when individual modules fail. This research proves that powerful error-correcting codes, such as bivariate bicycle (BB) codes, can function effectively in this distributed setting, paving the way for larger, more robust quantum memories and ultimately, more reliable quantum computation. The team’s simulations, using both long ion chains and single-qubit arrays, show promising results for achieving high logical qubit rates even with imperfect physical qubits.

Quantum Codes and Computing Architectures

This document provides a comprehensive overview of the challenges and approaches to building large-scale, fault-tolerant quantum computers, covering techniques for protecting quantum information from noise with codes like surface codes, LDPC codes, topological codes, and bicycle codes. It also explores various physical platforms for building qubits, including superconducting qubits, trapped ions, neutral atoms, and quantum dots, and how to connect them in modular architectures and distributed quantum computing systems. Surface codes are a dominant approach due to their relatively high threshold and suitability for 2D layouts, while topological codes provide inherent protection against local errors. Low-Density Parity-Check (LDPC) codes are gaining traction as a potentially more efficient alternative to surface codes, offering lower overhead but requiring more complex decoding.

Modular quantum computing is a key strategy for scaling up quantum computers by connecting smaller quantum processors (modules) together, with approaches including photonic interconnects and distributed architectures. Distributed quantum computing involves distributing quantum information across multiple nodes, potentially connected by quantum channels, and is a promising approach for building very large-scale quantum computers. The document acknowledges the diversity of physical qubit technologies, each with its own challenges and advantages. Implementing fault-tolerant gates is crucial for building a reliable quantum computer, with techniques like code deformation and transversal gates being explored. Optimizing compilation and minimizing qubit routing overhead are also important, as is minimizing the number of qubits, gates, and measurements required for a given quantum computation through techniques like code switching and code concatenation. This document serves as an excellent starting point for anyone entering the field of quantum error correction and quantum computing architectures, providing a broad overview of the current research landscape and identifying key areas for future investigation.

Distributed Quantum Error Correction via Cyclic Shifting

Researchers have developed a new approach to building large-scale quantum computers by distributing quantum information across multiple modules, connected by physically moving qubits (“flying qubits”) between them. This architecture utilizes a two-dimensional array of modules, where quantum data resides in the first row and error correction is performed using modules in the second row, connected via this cyclic shifting of physical qubits. The key breakthrough lies in demonstrating that complex quantum codes, specifically low-density parity-check (LDPC) codes like bivariate bicycle (BB) codes, can maintain their error-correcting performance even when distributed across these modules, requiring only this simple cyclic shift for communication. The team proposes two strategies for implementing this distributed quantum error correction: a broadly applicable “cyclic layout” and a BB code-specific “sparse cyclic layout,” achieving a quantum memory capable of storing 12 logical qubits distributed across 12 modules, each containing 12 physical qubits.

This modular design addresses challenges in scaling quantum computers, such as manufacturing limitations and maintaining stable conditions for large numbers of qubits. Simulations demonstrate that this distributed BB code achieves a remarkably low logical error rate, below 2 x 10 -6, even when the physical qubits themselves have a relatively high error rate of 10 -3. This represents a significant improvement over existing approaches, allowing for robust quantum computation with imperfect physical components. This architecture is particularly promising because it relies on physically transporting qubits between modules, a concept identified as essential for quantum communication and computation. The researchers have shown that this approach can be implemented using various types of qubits, including those encoded in ions, neutral atoms, electrons, or photons, offering flexibility in hardware implementation. Simulations, performed with both long ion chains and single-qubit arrays of ions, confirm the viability and performance of this modular quantum memory, paving the way for building larger and more powerful quantum computers.

Distributed Quantum Memory with Cyclic Shift Codes

This research presents a new architecture for distributed quantum memory, utilising modules connected by a cyclic shift implemented with flying qubits. The team demonstrates that low-density parity-check (LDPC) codes, specifically bivariate bicycle (BB) codes, can maintain performance even when distributed across multiple modules using this simple connection method, achieving a logical rate below 2·10−6 for a physical error rate of 10−3. The proposed system offers a practical approach to building larger quantum memories by distributing information across modules and employing movable qubits, such as ions, atoms, or electrons, for connections. Simulations using both long ion chains and single-qubit arrays demonstrate the feasibility of this architecture. The authors acknowledge that partitioning quantum LDPC codes is generally difficult due to their expansion properties, and the performance of the distributed code is comparable to an increase in the physical error rate of less than 2×. Future work could explore generalising this layout to other quantum codes like surface codes, and investigating the impact of flying qubits on existing bounds for code performance and logical operations.