Learning Non-Markovian Quantum Noise from Moiré-Enhanced Swap Spectroscopy with Deep Evolutionary Algorithm

Murphy Yuezhen Niu, Vadim Smelyanskyi, Paul Klimov, Sergio Boixo, Rami Barends, Julian Kelly, Yu Chen, Kunal Arya, Brian Burkett, Dave Bacon, Zijun Chen, Benjamin Chiaro, Roberto Collins, Andrew Dunsworth, Brooks Foxen, Austin G. Fowler, Craig Gidney, Marissa Giustina, Rob Graff, Trent Huang, Evan Jeffrey, David Landhuis, Erik Lucero, Anthony Megrant, Josh Mutus, Xiao Mi, Ofer Naaman, Matthew Neeley, Charles J. Neill, Chris Quintana, Pedram Roushan, John M. Martinis, Hartmut Neven

Abstract

Characterizing non-Markovian noise in superconducting qubits is critical for building fault-tolerant quantum computers. Microscopic Two-Level System (TLS) defects cause gate-dependent errors that traditional Markovian models fail to capture. Researchers at Google and UCSB have developed a novel framework combining a non-Markovian physical model, Moiré-enhanced spectroscopy, and a Deep Neural Network trained via Evolutionary Algorithms (DNN-EA). This approach achieves a 100x increase in characterization accuracy and a 10,000x speedup in optimization compared to standard techniques, enabling high-precision sensing of the qubit environment.

Introduction

Superconducting qubits are highly sensitive to their material environment. A primary source of decoherence is the interaction with Two-Level System (TLS) defects—microscopic imperfections that absorb energy from the qubit. Standard noise characterization often assumes Markovian (memoryless) behavior. However, during logic gates, the interaction between a frequency-tunable qubit and a TLS is inherently non-Markovian, meaning the error depends on the specific control trajectory. Efficiently extracting TLS parameters from experimental data is a significant challenge due to measurement noise and the complexity of the underlying dynamics.

Methodology

The researchers developed a three-pronged strategy to address these challenges:

1. Non-Markovian Physical Modeling

The team derived a model for the dynamics between a TLS and a frequency-tunable Transmon qubit during SWAP-like operations. This model accounts for both weak and strong coupling regimes, providing a rigorous basis for understanding gate-dependent errors.

2. Moiré-Enhanced Spectroscopy

To amplify the "fingerprints" of TLS defects, the study employed non-uniform temporal sampling in Two-Dimensional Swap Spectroscopy (TSSD). This technique leverages the Moiré effect—interference patterns created by sampling periodic signals at mismatched intervals—to capture high-frequency information with significantly fewer data points.

3. Deep Evolutionary Algorithm (DNN-EA)

A Deep Neural Network (DNN) was designed to map TSSD experimental data directly to physical TLS parameters, such as coupling strength ($\lambda$), decoherence rate ($\Gamma_2$), and TLS frequency ($\omega_{TLS}$). To avoid the local minima common in gradient-based optimization of noisy data, the network was trained using an Evolutionary Algorithm (EA).

Schematic of the DNN-EA framework mapping Moiré-enhanced spectroscopic data to physical TLS parameters.

Results and Discussion

The DNN-EA approach was benchmarked against traditional optimization methods like COBYLA and exhaustive grid searches using data from a Google quantum processor.

Performance and Efficiency

The method demonstrated a 10,000x speedup in optimization time compared to exhaustive search methods. Furthermore, the accuracy of the characterization improved by two orders of magnitude (100x) over standard techniques. The gradient-free nature of the Evolutionary Algorithm proved essential for navigating the complex, noisy landscape of experimental data.

Performance measure comparison between COBYLA and the Evolutionary Algorithm, highlighting the significant speedup and lower error rates.

Robustness and Fidelity Prediction

The model successfully reproduced experimental data across multiple frequency regimes, showing resilience against background fluctuations. A key insight from the model is the "Coherence Paradox": more coherent TLS defects can actually cause larger gate errors because they maintain their state longer, leading to stronger non-Markovian interference. The framework allows for predicting drops in two-qubit gate fidelity, which is vital for developing noise-aware quantum compilers.

Conclusion

By treating qubits as high-precision sensors and utilizing advanced machine learning, this work provides a robust path for characterizing the non-Markovian noise environment of quantum processors. Future efforts will focus on scaling this characterization to entire chips and implementing real-time "online" calibration to mitigate the temporal drift of TLS defects.

Summary updated

Abstract

Two-level-system (TLS) defects in amorphous dielectrics are a major source of noise and decoherence in solid-state qubits. Gate-dependent non-Markovian errors caused by TLS-qubit coupling are detrimental to fault-tolerant quantum computation and have not been rigorously treated in the existing literature. In this work, we derive the non-Markovian dynamics between TLS and qubits during a SWAP-like two-qubit gate and the associated average gate fidelity for frequency-tunable Transmon qubits. This gate dependent error model facilitates using qubits as sensors to simultaneously learn practical imperfections in both the qubit's environment and control waveforms. We combine the-state-of-art machine learning algorithm with Moiré-enhanced swap spectroscopy to achieve robust learning using noisy experimental data. Deep neural networks are used to represent the functional map from experimental data to TLS parameters and are trained through an evolutionary algorithm. Our method achieves the highest learning efficiency and robustness against experimental imperfections to-date, representing an important step towards in-situ quantum control optimization over environmental and control defects.
BibTeX
@misc{niu2019learning,
  author = {Murphy Yuezhen Niu and Vadim Smelyanskyi and Paul Klimov and Sergio Boixo and Rami Barends and Julian Kelly and Yu Chen and Kunal Arya and Brian Burkett and Dave Bacon and Zijun Chen and Benjamin Chiaro and Roberto Collins and Andrew Dunsworth and Brooks Foxen and Austin G. Fowler and Craig Gidney and Marissa Giustina and Rob Graff and Trent Huang and Evan Jeffrey and David Landhuis and Erik Lucero and Anthony Megrant and Josh Mutus and Xiao Mi and Ofer Naaman and Matthew Neeley and Charles J. Neill and Chris Quintana and Pedram Roushan and John M. Martinis and Hartmut Neven},
  title = {{Learning Non-Markovian Quantum Noise from Moiré-Enhanced Swap Spectroscopy with Deep Evolutionary Algorithm}},
  year = {2019},
  eprint = {1912.04368},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1912.04368},
}

Loading related content...

Figure