Abstract
Stabilizer-based simulation of quantum error-correcting codes typically relies on the Pauli-twirling approximation (PTA) to render non-Clifford noise classically tractable, but PTA can distort the behavior of physically relevant channels such as thermal relaxation. Physically accurate noise simulation is needed to train decoders and understand the noise suppression capabilities of quantum error correction codes. In this work, we develop an exact and stabilizer-compatible model of qubit thermal relaxation noise and show that the combined amplitude damping and dephasing channel admits a fully positive probability decomposition into Clifford operations and reset whenever $T_2 \leqslant T_1$. For $T_2 > T_1$, the resulting decomposition is negative, but allows a smaller sampling overhead versus independent channels. We further introduce an approximated error channel with reset that removes the negativity of the decomposition while achieving higher channel fidelity to the true thermal relaxation than PTA, and extend our construction to finite temperature relaxation. We apply the exact combined model to investigate large surface codes and bivariate bicycle codes on superconducting platforms with realistic thermal relaxation error. The differing logical performances across code states further indicate that noise-model-informed decoders will be essential for accurately capturing thermal-noise structure in future fault-tolerant architectures.
Introduction
To build fault-tolerant quantum computers, researchers rely heavily on software simulators to evaluate Quantum Error Correction (QEC) codes. While stabilizer simulators (like Stim) can simulate thousands of qubits efficiently, they are traditionally limited to Clifford operations and Pauli noise. Real quantum hardware, however, suffers from thermal relaxation—energy loss ($T_1$) and dephasing ($T_2$)—which are non-Clifford processes.
Historically, researchers have used the Pauli-Twirling Approximation (PTA) to map thermal noise onto Pauli errors. However, as shown in Figure 1, PTA averages out the intrinsic directional bias of thermal relaxation. While a qubit naturally relaxes toward the $|0\rangle$ state, PTA treats it as a random bit or phase flip. This distortion can lead to misestimates of logical error rates by factors of 2x to 10x, making it an unreliable tool for designing high-performance decoders.
Methodology: Unified Channel Decomposition
The core innovation of this work is the combination of amplitude damping and pure dephasing into a single, unified thermal relaxation channel before applying Quasi-Probabilistic Decomposition (QPD). QPD typically breaks non-Clifford noise into a linear combination of Clifford operations, but often introduces negative probabilities that lead to an exponential sampling overhead.
The authors prove that when $T_2 \le T_1$, the negative terms in the combined decomposition perfectly cancel out. This results in a fully positive probability distribution over Clifford and reset operations, enabling exact Monte Carlo simulation with zero sampling overhead. Figure 5 illustrates this "sweet spot" in the parameter space where the sampling overhead factor $\Gamma$ equals 1.
For regimes where $T_2 > T_1$, a "reset-based approximation" is introduced. By dropping negative terms and rescaling, the authors maintain higher fidelity to physical noise than PTA while keeping simulation tractable.
Key Results and Scaling
The mathematical condition for zero-overhead simulation ($T_2 \le T_1$) aligns well with modern superconducting hardware (e.g., IBM Heron), where environmental dephasing typically ensures $T_2$ does not exceed $T_1$. Leveraging this efficiency, the authors performed large-scale simulations of distance-11 surface codes and high-rate Bivariate Bicycle (BB) qLDPC codes.
Key findings include:
- PTA Inconsistency: PTA is not only inaccurate but inconsistent, underestimating errors for some logical states while overestimating others. This confirms that decoders must be "noise-aware" to handle the asymmetric structure of thermal noise.
- Logical State Sensitivity: Because thermal noise is biased toward the ground state, the logical $|0\rangle_L$ state often exhibits different error rates than $|+\rangle_L$, a nuance captured only by the exact model.
- Code Viability: Bivariate Bicycle codes were shown to be highly viable under realistic thermal noise, often outperforming surface codes in encoding efficiency while maintaining low logical error rates.
Conclusion
This work provides a bridge between physically accurate noise modeling and scalable stabilizer simulation. By exploiting the $T_2 \le T_1$ regime, researchers can now simulate "digital twins" of real quantum hardware without the inaccuracies of PTA or the overhead of traditional exact methods. Future work will focus on incorporating qubit-specific calibration data and extending this unified decomposition approach to other non-Clifford errors like coherent gate noise.
Summary updated