Tomographic reconstruction of non-Markovian Gaussian noise in continuous-variable photonic systems
Memory effects in continuous-variable (CV) photonic systems can make standard Markovian Gaussian noise models inaccurate, even when the average noise strength is correctly calibrated. We propose a theoretical and simulation-based tomography framework for reconstructing such memory effects directly from time-resolved homodyne data. The method estimates an effective time-local Gaussian generator, represented by drift and diffusion matrices, using a regularized maximumlikelihood procedure constrained by complete positivity (CP) of the finite-time Gaussian channel. This description covers any Gaussian-preserving noise dynamics for which a well-defined time-local generator exists, rather than being restricted to particular noise mechanisms. Non-Markovianity is then diagnosed at the generator level through temporary violations of local CP-divisibility, rather than from time dependence alone. This distinction is important because ordinary nonstationary noise can remain CP-divisible, whereas genuine memory can produce a non-CP-divisible reduced generator while the accumulated channel remains physical. Numerical benchmarks covering memorydependent loss, nonstationary thermal excitation, and pseudomode-mediated memory show that the reconstruction distinguishes CP-divisible time-dependent noise from genuine non-CP-divisible dynamics and captures the corresponding signatures in the Gaussian covariance evolution. The framework therefore provides a practical route for diagnosing non-Markovian Gaussian noise in CV photonic quantum systems.
Authors
- Hamid Reza Naeij (ORCID: https://orcid.org/0000-0002-5673-501X)
Publication Details
- Journal
- arXiv (Cornell University)
- Published
- 2026-11-17
- DOI
- https://doi.org/10.48550/arxiv.2611.22693
- Primary Topic
- Quantum Information and Cryptography
- Type
- preprint