Enhancing the Clique Local Decoder to Correct Length-2 Space Errors in the Surface Code

TL;DR

This paper introduces Clique_L2, an improved local decoder for surface codes that corrects length-2 errors, significantly reducing bandwidth needs and enhancing error correction at higher code distances and error rates.

Contribution

It extends the Clique decoder by enabling correction of length-2 error chains, improving bandwidth efficiency and error correction in quantum surface codes.

Findings

Up to 8.95x bandwidth reduction for data-qubit errors.

Up to 18.3x bandwidth reduction with clustered errors.

Effective across various noise models and higher code distances.

Abstract

The growing demand for fault-tolerant quantum computing drives the need for efficient, scalable Quantum Error Correction (QEC) strategies. Conventional decoders designed for worst-case error scenarios incur significant overhead, prompting the development of local decoders, that leverage the sparse and often trivial nature of many quantum errors, to support the conventional decoders. The previously proposed Clique decoder addresses this by handling isolated, length-1 space and time errors within the cryogenic environment with minimal hardware costs, thereby mitigating I/O bandwidth constraints between cryogenic quantum systems and room-temperature processors. Building on this foundation, we propose Clique_L2 that extends the Clique-based approach by relaxing some original constraints and incorporating additional low-cost logic to also correct length-2 error chains in space, which become non-trivial occurrences at higher physical error rates and code distances. This enhanced capability not only further reduces out-of-the-fridge data transmission but also adapts more effectively to clustered errors observed under a variety of noise models. Specifically, under data-qubit-only errors and uniformly random noise, Clique_L2 achieves up to 8.95x decoding bandwidth reduction over the original Clique (or Clique_L1) decoder, especially beneficial at higher code distances. When clustered errors and longer error chains are more likely to occur, Clique_L2 achieves up to 18.3x decoding bandwidth reduction over Clique_L1, achieving substantial benefits across a wide range of physical qubit error rates.