Real-Time Stabilization of Spectral Diffusion in Superconducting Qubits

This repository contains the simulation source code and reproduction data for the paper:
"Real-Time Stabilization of Spectral Diffusion in Superconducting Qubits via Syndrome-Density Feedback and Adaptive Decoding"

📄 Abstract

Superconducting quantum processors suffer from time-varying noise, specifically gate error drifts driven by two-level system (TLS) defects. We present a closed-loop control architecture ("Holo-Neural") that mitigates these drifts using syndrome-density feedback. Simulating a distance-11 rotated surface code under realistic circuit-level noise, we demonstrate the recovery of logical fidelity in two distinct regimes: a "Death Zone" stress test and a "Realistic" maintenance scenario.

📊 Key Results

1. The "Death Zone" Stress Test (High Noise)

Parameters: 5% Measurement Error, Drift to 3.5% Gate Error.

Metric	Standard Qubit	Stabilized Qubit (Ours)
Logical Error Rate	46.6% (Catastrophic)	0.98% (Stable)
Suppression Factor	1.0	47.2x

2. The "Realistic" Regime (Daily Driver)

Parameters: 1% Measurement Error, Drift to 1.5% Gate Error.

Metric	Standard Qubit	Stabilized Qubit (Ours)
Logical Error Rate	13.1%	0.11%
Suppression Factor	1.0	120x

🛠️ Installation

Requires Python 3.10+ and high-performance quantum simulation libraries:

pip install stim pymatching numpy matplotlib

🚀 Reproduction

Experiment 1: Extreme Stress Test (Generates Figure 1)

python3 simulation_study.py

Experiment 2: Realistic Regime (Generates Figure 2)

python3 realistic_simulation.py

📂 Repository Structure

• simulation_study.py: Main kernel (High-Noise Stress Test).
• realistic_simulation.py: Secondary kernel (Low-Noise Realistic Regime).
• final_proof.png: Plot for Experiment 1.
• realistic_proof.png: Plot for Experiment 2.
• README.md: Documentation.

🔗 Citation

If you use this code in your research, please cite the ArXiv preprint (in review):

@article{HoloNeural2026,
  title={Real-Time Stabilization of Spectral Diffusion in Superconducting Qubits},
  author={Justin Arndt},
  journal={arXiv preprint arXiv:2601.XXXXX},
  year={2026}
}

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.