Gradient Probabilistic Algorithm for Compact Lithium Niobate Integrated Photonic Devices

Compact photonic devices are highly desired in photonic integrated circuits. In this work, we use an efficient inverse design method to design a 50/50 beam splitter in lithium niobate integrated platforms. We employ the Gradient Probability Algorithm (GPA), which is built upon traditional gradient algorithms. The GPA utilizes the adjoint method for the comprehensive calculation of the electric field across the entire design area in a single iteration, thereby deriving the gradient of the design area. This enhancement significantly accelerates the algorithm's execution speed. The simulation results show that an ultracompact beam splitter with a footprint of 13 mu m x 4.5 mu m can be achieved in lithium niobate integrated platforms, where the insertion loss falls below 0.5 dB within the 1500 nm to 1700 nm range, thus reaching its lowest point of 0.15 dB at 1550 nm.