Fast Simulation and Decoding of Superconducting Erasure Qubits at Scale

Erasure qubits offer a compelling approach to fault-tolerant quantum computation by converting dominant noise sources into detectable leakage events. These events can be corrected using dedicated erasure checks, ideally performed after every gate. While such idealized scenarios are efficiently simulated using Pauli frame methods like Stim, realistic implementations must contend with faulty checks and reduced check frequency, requiring scalable simulation tools and adaptive decoding strategies for leakage.

We introduce a high-performance leakage simulator built on Stim that supports depolarizing leakage models with no performance penalty compared to Pauli noise. Leveraging this tool, we investigate how the frequency of erasure checks influences fault-tolerance thresholds and sub-threshold scaling, using noise parameters informed by more granular physical simulations of Dimon devices developed by OQC. Integrated with the Local Clustering Decoder, which dynamically adapts to erasure outcomes, our framework enables rapid evaluation of erasure-based error correction strategies. This co-design approach not only guides the development of superconducting erasure qubit platforms but also provides a general-purpose tool for modelling leakage and erasure across diverse quantum architectures.