Simulating leakage and LRUs in QEC codes beyond the stabilizer formalism

Leakage out of the computational subspace can be destructive to quantum error correction. A significant roadblock to scalability of error correcting codes is the difficulty of characterizing, simulating and addressing leakage and ancillary effects. Proposals to address leakage include hardware-adapted schemes and circuit-based leakage-reduction units (LRUs). In this work, we employ state-of-the-art statevector and tensor network simulation methods to perform realistic modeling of repetition codes, and color codes under a wide range of leakage errors, avoiding the approximations used in simplified stabilizer formalism-compatible leakage noise models. We identify the improvements afforded by the more accurate simulation of error correcting codes. Our analysis reveals how thresholds and logical error rates vary across leakage models and different circuit-level noise assumptions.