Single-shot preparation of hypergraph product codes via dimension jump
Yifan Hong
TL;DR
The paper addresses the challenge of reliably preparing the codespace for constant-rate hypergraph product (HGP) quantum LDPC codes, which is a bottleneck in fault-tolerant implementations. It introduces a two-stage, single-shot protocol that uses a thickened code, formed via a homological product with a classical code, to perform a dimension-jump from a 2D to a 3D complex and then collapses back to the original HGP codespace, enabling constant-depth initialization with $O(√n)$ spatial overhead. The method relies on metachecks and a spacetime-decoding perspective to tolerate measurement and data errors, with a secondary improvement via a star-code variant that reduces spatial overhead further. Numerical simulations provide evidence of fault-tolerance and threshold-like behavior under local stochastic noise for selected LDPC ensembles, and the work outlines pathways to generalize the approach and integrate it with fault-tolerant architectures. Overall, the protocol offers a practical route to fast, error-tolerant preparation of LDPC HGP code states, laying groundwork for scalable, high-speed quantum initialization in architectures capable of long-range connectivity.
Abstract
Quantum error correction is a fundamental primitive of fault-tolerant quantum computing. But in order for error correction to proceed, one must first prepare the codespace of the underlying error-correcting code. A popular method for encoding quantum low-density parity-check codes is transversal initialization, where one begins in a product state and measures a set of stabilizer generators. In the presence of measurement errors however, this procedure is generically not fault-tolerant, and so one typically needs to repeat the measurements many times, resulting in a deep initialization circuit. We present a protocol that prepares the codespace of constant-rate hypergraph product codes in constant depth with $O(\sqrt{n})$ spatial overhead, and we show that the protocol is robust even in the presence of measurement errors. Our construction is inspired by dimension-jumping in topological codes and leverages two properties that arise from the homological product of codes. We provide some improvements to lower the spatial overhead and discuss applications to fault-tolerant architectures.
