In a new study published in Defence Technology, a team of researchers from China introduced a framework that integrates finite-element simulation with neural networks to isolate density disturbances from signal measurements.
"Atomic density fluctuations act as a primary source of systematic error in Faraday optical rotation angle (FORA) measurements, inherently limiting the performance of atomic gyroscopes," explains corresponding author Associate Professor Haoying Pang. "We moved beyond traditional suppression techniques to propose a computational approach that directly estimates and compensates for these disturbances in real-time."
The core of this innovation is the three-dimensional atomic density (TDAD) model. "Unlike the traditional one-dimensional Lambert-Beer law, this advanced model captures complex spatial light absorption characteristics, accounting for polarization distribution, density-dependent relaxation, and wall-induced effects," says Pang. "This multidimensional approach provides a more physically accurate representation of the alkali metal cell's internal environment."
To implement this model, the team leveraged a data-driven pipeline. By utilizing COMSOL Multiphysics for finite-element simulations, the researchers generated a comprehensive dataset of absorption-density mappings. These data were then used to train a feedforward neural network, creating a high-precision estimator for atomic density fluctuations. A specific decoupling equation was subsequently constructed to mathematically separate the density contributions from the System output signal.
"Experimental validation on a K-Rb-21Nesystem demonstrated that this method achieves superior long-term stability compared to traditional platinum-resistance temperature control methods," shares Pang. "This framework is designed to be generalizable, paving the way for applying similar decoupling strategies to other optical pumping-based sensors, such as optically pumped magnetometers, where incident laser power density and spatial atomic polarization distribution are critical factors."
