Physica Scripta
Benchmarking Non-Markovianity Measures in Digital Quantum Simulation of Open Quantum Systems
S. M. Yousuf Iqbal Tomal
Abstract
Abstract Digital quantum simulation of open quantum systems has produced several methods for reproducing non-Markovian dynamics, among them pseudomode embeddings, the hierarchical equations of motion, and collision models. These methods are validated against populations, coherences or spectra, not against the measures that quantify memory. We study how the Breuer-Laine-Piilo (BLP) measure behaves under the approximations of a digital simulation. The circuits run on classical density-matrix simulators, where the exact value of the measure is computable. First, a two-level encoding of the pseudomode misrepresents the measure with a regime-dependent sign. It overstates the memory by a factor of 2.3 at strong coupling, loses most of it at moderate temperature, and both signs occupy large regions of the coupling-temperature plane. A variance-aware rule for the required Fock dimension, ⌈⟨n⟩ + cσₙ⌉ + 1, accounts for the number statistics: at equal mean occupation, Fock, coherent and thermal states need 4, 8 and 16 levels. We calibrate c for trajectory- and measure-level targets and cross-validate it across three families. Second, the raw finite-shot estimate overstates the true value by up to 89% under realistic noise. A significance-threshold estimator reduces this bias to −0.004 at the strongest noise, and a Markovian control with a percentile rule keeps false positives at or below 0.3%, with measured miss rates. Third, slow non-Gaussian device noise creates backflow in a memoryless channel, so the measure characterizes the implemented channel and not the source of its memory. The Rivas-Huelga-Plenio measure, computed from reconstructed maps, does not converge on these channels and serves only as a diagnostic. At equal two-qubit gate count, a pseudomode and a two-site collision chain cross over in accuracy. Every result is validated against an exact reference and reproducible from recorded data. We make no claim of quantum advantage.