KAN-PINN-Based Simulation of DFB Lasers

To address the challenges encountered in the simulation and parameter extraction of distributed-feedback (DFB) lasers, this work starts from the rate equations of the DFB laser to derive its steady-state formulation and small-signal model, and constructs steady-state and transient KAN-based physics-informed neural-network (KAN-PINN) architectures. In the steady state, accurate L-I/I–V curves and −3 dB bandwidth results are obtained. In addition, the parameters of the DFB laser are systematically organized, and a KAN-PINN inverse mode, an adaptive moment estimation (Adam) optimizer-based physical inversion, and a hybrid strategy combining the two are proposed. The extracted parameters show small deviations from the true values, and the simulation results agree well with the actual data. The proposed methodology can be extended to other lasers and even to a broader class of optoelectronic devices for parameter extraction and simulation.