Error correction in neutral atom quantum computers benefits from accounting for correlated atom losses during computation, when one atom fails, its gate partner often does too. Maximal-Matching-Mixture (MMM) is a decoder that exploits these correlations and lowers logical error rates by as much as twenty-five times compared with previous methods and two times better than existing loss-aware approaches. Neutral atom quantum computers are susceptible to errors caused by atoms being lost during calculations; key to this susceptibility is the fact that when one atom fails, its connected partner often does too.

An improved error correction method accounts for this linked behaviour, sharply boosting performance compared with existing techniques. To address these issues, an improved error correction method accounts for this linked behaviour and boosts performance compared with existing techniques.

The Maximal-Matching-Mixture (MMM) decoder builds a complex map of potential error scenarios, akin to connecting clues at a crime scene, and then analyses it using an algorithm similar to how rumours spread: information passes along until the most probable cause emerges. This approach uses the fact that correlated losses, where two connected qubits fail simultaneously, are far more likely than independent failures. MMM lowers logical error rates by up to twenty-five times over previous methods; but can this technique scale effectively as quantum computers grow in complexity.

Correlated Atom Loss Enables Twentyfivefold Reduction

Logical error rates dropped as much as ~25x with the Maximal-Matching-Mixture (MMM) decoder versus those not accounting for atom losses, a gain approximately twofold better than prior state-of-the-art approaches. MMM constructs a weighted map of potential errors based on correlated loss scenarios and utilizes standard belief propagation to interpret it, enabling stronger error correction compared to existing methods. This breakthrough addresses an important limitation in neutral atom quantum computing: qubit failures often occur in pairs when one entangled ‘gate partner’ fails alongside another during operations, a phenomenon previously unexploited.

A twenty-five times reduction in logical error rate is achieved by the Maximal-Matching-Mixture decoder over decoders that do not account for atom losses while applied to the Gross code; this also represents roughly two times improvement versus previous loss-aware techniques from Baranes et al. Neutral atom quantum computers offer potential scalability, yet qubit losses present an obstacle because these aren’t random, when one atom is lost, its entangled partner frequently follows suit. Exploiting this correlation allows construction of a better decoder focused on syndrome extraction circuits within the Gross code framework and creates weighted mixtures of decoding hypergraphs based on likely loss scenarios.

UCLA researchers, collaborating with University of California scientists, have demonstrated novel error correction for neutral atom quantum computers by exploiting correlations arising when entangled qubits fail simultaneously during computation, something existing methods previously overlooked. Constructing a weighted map representing probable failure scenarios alongside standard belief propagation sharply reduces logical errors compared to approaches treating qubit losses randomly or simply acknowledging they occur together. Achieving up to twenty-five times lower error rates than decoders ignoring correlated events offers substantial gains within the Gross code; however, whether these benefits extend beyond this specific structure remains an open question.

This research demonstrates that accounting for correlated qubit loss improves error correction in neutral atom quantum computing systems. The authors found that when one of two entangled atoms fails, its partner is often lost as well, and developed a decoder, the Maximal-Matching-Mixture (MMM), that utilizes this information with standard belief propagation techniques on [[144,12,12]] Gross codes.

Applying MMM lowered logical error rates by up to twenty-five times compared to decoders unaware of losses, and twice the improvement over previous loss-aware methods; these results suggest exploiting correlations between failures can enhance performance. Researchers indicate further work will explore whether these benefits extend beyond the specific code structure used in their experiments.

👉 More information
🗞 Maximally-Matched Mixture Decoding for Correlated Atom Loss in Quantum LDPC Codes
✍️ Jacob W. Murri (UCLA)
🧠 ArXiv: https://arxiv.org/abs/2610.38283

More like this