A Deep Learning Approach for Dynamic Modeling of Stimulated Raman Scattering in Chalcogenide Microstructured Optical...

Area of Science:

• Nonlinear optics
• Computational photonics
• Machine learning in physics

Background:

• Stimulated Raman scattering (SRS) is crucial for optical communications and sensing.
• Traditional modeling methods like the split-step Fourier method (SSFM) are computationally intensive.

Purpose of the Study:

• To develop a computationally efficient deep learning framework for modeling nonlinear pulse propagation and SRS.
• To accelerate the prediction of SRS behavior compared to conventional simulation techniques.

Main Methods:

• A hybrid neural network architecture was designed to capture spatio-temporal dependencies in nonlinear pulse propagation.
• Experiments were conducted using chalcogenide microstructured optical fibers (MOFs) for real-world data generation.
• The deep learning model's performance was validated against the SSFM.

Main Results:

• The hybrid neural network achieved 116x speedup on GPU and 44x on CPU compared to SSFM.
• The model demonstrated accurate and generalizable predictions of SRS.
• Successful generation of mid-infrared (MIR) SRS in a 2-$\mu$m direct-pumped As$_{2}$S$_{3}$ MOF was achieved.

Conclusions:

• The proposed deep learning framework offers a significant acceleration for SRS modeling.
• This advancement enables real-time analysis and inverse design of nonlinear photonic systems.
• The use of chalcogenide MOFs provides a valuable platform for experimental validation and data generation.