Reconstruction of Noise from Dynamical Mode Decomposition in Quantum Systems

Abstract: Noise fundamentally limits the performance of quantum systems by degrading coherence and reducing control fidelity. Characterizing this noise is essential for improving measurement accuracy, understanding environmental interactions, and designing effective control strategies. However, extracting its spectral characteristics and associated dephasing times from limited experimental data remains a key challenge. Here, we introduce a data-driven framework based on Dynamical Mode Decomposition (DMD) to analyze the dynamics of quantum systems under stochastic noise. We reinterpret DMD modes as statistical weights over an ensemble of quantum trajectories and use this mapping to construct a noise power spectral density (PSD) via a nonlinear transformation. This enables identification of dominant frequency contributions in both white and $1/f$ noise environments and directly yields the pure dephasing time $T_2^*$ from DMD eigenvalues. To overcome instability in standard DMD reconstruction, we develop a constrained extrapolation scheme using the extracted $T_2^*$ as a physical bound and the learned PSD as mode weights. We demonstrate our method on simulated Ramsey-type dynamics and validate its ability to recover spectral features and coherence decay. This approach offers a robust, data-driven tool for diagnostic and predictive noise analysis in quantum systems.