High-fidelity simulation and reduced-order modeling of optical pumping in quantum devices
Abstract
Spatially resolved modeling of optically pumped alkali-metal vapors requires the solution of a high-dimensional nonlinear diffusive quantum master equation (DQME) coupled to state-dependent light propagation. We develop a structure-preserving computational framework for this coupled three-dimensional system in the high-pressure limit. The high-fidelity solver combines second-order finite differences, explicit pseudo-time marching, and an exponential-integrator discretization of the light-propagation equation. The scheme preserves the density-matrix trace, while a derived global pseudo-time-step restriction guarantees nonnegative, bound-preserving population updates and frozen-coefficient $1$-norm stability. In the absence of a closed-form solution for the general coupled problem, the solver is verified against complementary analytical reductions, pumping-rate asymptotics, and a light-propagation invariant, all of which are accurately recovered. Tensorized GPU implementation reduces the per-iteration cost by more than two orders of magnitude relative to a single-core CPU. For repeated parameter queries, we further construct an intrusive POD--Galerkin reduced-order model. The quadratic spin-exchange term is treated by an exact nonlinear quadratic expansion, whereas the nonpolynomial and nonlocal optical-pumping term is hyper-reduced using the discrete empirical interpolation method. At out-of-sample parameter points, the reduced model closely reproduces the full-order population and polarization fields while achieving speedup of more than 7×103 relative to the single-core CPU full-order solver. The resulting hierarchy connects verified high-fidelity simulation with efficient reduced-order prediction for spatially resolved alkali-metal spin dynamics.
Related
- Reduced-Space Multi-Fidelity Bayesian Optimization of Process Simulation Models
- Q-Edge: Symmetry-Reduced Quantum Simulation of Structured Extreme Dependence
- Greedy Orthogonal Polynomial Manifolds with Lipschitz Control for Dimensionality Reduction and Reduced Modeling
- Matrix-Free Moment-Matching Method for Reduced-Order Modeling of Quadratic-Bilinear Descriptor Systems with Multiple Inputs
- Combined High-Dimensional and Reduced-Order Modeling Based on an Iterative Domain Decomposition Method
BibTeX
@misc{xu2026highfidelity,
title = {{High-fidelity simulation and reduced-order modeling of optical pumping in quantum devices}},
author = {Keyu Xu and Kezheng Yan and Jinbo Hu and Yanqing Shi and Xiaorui Xu and Nan Zhao and Qian Wang},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.41778},
eprint = {2610.41778},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.41778},
}
RIS
TY - UNPB
TI - High-fidelity simulation and reduced-order modeling of optical pumping in quantum devices
AU - Keyu Xu
AU - Kezheng Yan
AU - Jinbo Hu
AU - Yanqing Shi
AU - Xiaorui Xu
AU - Nan Zhao
AU - Qian Wang
PY - 2026
DA - 2026/10/30
PB - arXiv
DO - 10.48550/arxiv.2610.41778
UR - https://arxiv.org/abs/2610.41778
AB - Spatially resolved modeling of optically pumped alkali-metal vapors requires the solution of a high-dimensional nonlinear diffusive quantum master equation (DQME) coupled to state-dependent light propagation. We develop a structure-preserving computational framework for this coupled three-dimensional system in the high-pressure limit. The high-fidelity solver combines second-order finite differences, explicit pseudo-time marching, and an exponential-integrator discretization of the light-propagation equation. The scheme preserves the density-matrix trace, while a derived global pseudo-time-step restriction guarantees nonnegative, bound-preserving population updates and frozen-coefficient $1$-norm stability. In the absence of a closed-form solution for the general coupled problem, the solver is verified against complementary analytical reductions, pumping-rate asymptotics, and a light-propagation invariant, all of which are accurately recovered. Tensorized GPU implementation reduces the per-iteration cost by more than two orders of magnitude relative to a single-core CPU. For repeated parameter queries, we further construct an intrusive POD--Galerkin reduced-order model. The quadratic spin-exchange term is treated by an exact nonlinear quadratic expansion, whereas the nonpolynomial and nonlocal optical-pumping term is hyper-reduced using the discrete empirical interpolation method. At out-of-sample parameter points, the reduced model closely reproduces the full-order population and polarization fields while achieving speedup of more than 7×10^3 relative to the single-core CPU full-order solver. The resulting hierarchy connects verified high-fidelity simulation with efficient reduced-order prediction for spatially resolved alkali-metal spin dynamics.
ER -