Silicon-based dual-channel optical switches based on GSST phase change materials

The deep coupling of optical switches, optical computing, and AI is reshaping the arithmetic infrastructure landscape. Optical switches, as the dynamic reconfiguration core of optical computing systems, realize low-power optical path switching through non-volatile phase-change materials. It can meet the challenging needs of AI for low power consumption, high parallelism and high speed. In this paper, we propose a 1 × 2 reconfigurable optical switch with heterogeneous integration on silicon substrate, which consists of a silicon strip waveguide and a GSST-Bragg (Ge₂Sb₂Se₄Te₁ ) grating waveguide. The results show that the extinction ratio of two-channel switching can reach more than 21dB; for the straight-through and branch ports, the losses at the center wavelength are about 0.24dB and 0.66dB, which are superior to the traditional electro-optical modulation schemes. Combined with a photonic neural network architecture, the optical switch can support dynamic configuration of a wavelength routing matrix.