Loss-biased fault-tolerant quantum error correction

TL;DR

This paper introduces loss biasing in quantum error correction for neutral-atom processors, transforming errors into erasures to enable fault-tolerance with faster QEC cycles and reduced hardware overhead.

Contribution

It proposes a practical loss biasing technique using autoionization to suppress correlated errors and improve fault-tolerant quantum computing performance.

Findings

Loss biasing converts errors into erasures, restoring fault-tolerant scaling.

Shorter QEC cycles amplify platform-specific errors without loss biasing.

Loss-aware decoding with loss biasing achieves optimal erasure scaling.

Abstract

We investigate the limits of quantum error correction (QEC) in neutral-atom processors approaching high-fidelity gates and fast cycle times. We show that shorter QEC cycles amplify platform-specific errors, notably Rydberg excitation hopping, and hinder decay of residual Rydberg population, leading to non-Markovian correlated errors that degrade logical performance. To address this, we introduce loss biasing, where spurious Rydberg excitations are rapidly converted into atom loss via mid-circuit ionization, transforming errors into erasure-like noise and suppressing their propagation. Loss biasing restores the fault-tolerant logical error scaling for intra-cycle Pauli errors; furthermore, we argue that when supported with loss-aware decoding, it can achieve the optimal scaling of erasures while enabling shorter QEC cycles with reduced hardware overhead. We outline an implementation using fast autoionization in alkaline-earth(-like) atoms, establishing loss biasing as a practical route toward fault-tolerant quantum computing with sub-millisecond QEC cycles.