Search

Optical phase change materials (OPCMs) hold significant promise in photonic integrated circuits (PICs) due to their non‐volatile modulation capabilities through phase transition. Recently, integrating OPCMs with silicon devices has been leveraged to achieve multi‐level modulation in neural network computing. However, the conventional OPCM Ge2Sb2Te5 (GST), faces challenges in large‐scale PICs due to its large absorption. In this work, an optimized low‐loss GST superlattice is demonstrated, which significantly lowers crystalline attenuation coefficients from 2.69 to 0.539 dB µm−1, with negligible loss in amorphous state, while retaining excellent phase modulation capability. Additionally, 15 distinguishable extinction ratio states are achieved by manipulating irradiation energy for a hybrid Si microring resonator (MRR) embedded with GST superlattice. To further explore the potential application of GST superlattice in optical computing, an all‐optical tunable temporal differentiator is empirically demonstrated with a single hybrid MRR, achieving a broad dynamic differentiation order ranging from 0.7 to 1.3. The non‐volatile nature of OPCM facilitates low energy consumption and effectively optimizes chip density. The proposed structural design strategy to reduce the absorption loss is universal for any promising OPCMs feasible for large‐scale artificial intelligence‐driven PICs, including optical analog computing, optical switching and optical neural networks. [ABSTRACT FROM AUTHOR]