Radiation-Induced Fault Detection in Superconducting Quantum Devices

The quest for universal superconducting quantum computing is hindered by noise and errors. It has been proven that Quantum Error Correction (QEC) codes will lay at the foundation of fault tolerant quantum computing. However, cosmic-ray induced correlated errors, which are the most detrimental events that can impact superconducting quantum computers, are yet to be efficiently tackled. In order to reach fault tolerance, we must also develop radiation aware methods to complement QEC. In this paper, we propose the first algorithm to effectively exploit syndrome information for the efficient detection of radiation events in superconducting quantum devices at runtime. We perform a thorough analysis of simulated Rotated Surface codes injecting over 11 million physics-modeled radiation-induced faults. We consider the properties of the X and Z check bases, the impact of code distance, and the decoder's time to solution constraints. Our technique detects $100\%$ of injected faults, regardless of the impact's position. Moreover, we accurately identify both the radiation impact centre and the area affected, with an overhead lower than $0.3\%$ the decoding time. Additionally, we use the fault identification information to propose a radiation fault correction technique that improves of up to $20\%$ the output correctness compared to existing decoders.