Segmentation-Engineered Ge4Sb6Te7 Switch on SOI Platform for Multilevel Nonvolatile Photonic Neural Inference

Abstract Phase-change materials (PCMs) have emerged as key enablers of nonvolatile, ultracompact photonic switches for energy-efficient deep neural network (DNN) applications. In this work, we computationally investigate the recently discovered Ge4Sb6Te7 (GST-467) as a high-contrast optical PCM and demonstrate its suitability for multilevel photonic computing. We experimentally measure the complex refractive indices of amorphous and crystalline GST-467 and use them to numerically propose a segmented silicon-on-insulator photonic switch optimized at 1550 nm. Three-dimensional (3D) finite-difference time-domain simulations show that segmentation significantly enhances the extinction ratio while maintaining low insertion loss, yielding a design figure of merit more than 7-fold higher than that of an unsegmented design. Multiphysics thermo-optical simulations further demonstrate efficient, reversible laser-induced switching with sub-nJ energy requirements for crystallization and amorphization. Compared with established GST, GSST, and GSS compositions, GST-467 provides the largest transmittance contrast and supports up to 48 resolvable optical states. When deployed as multilevel weights in photonic DNN architectures, the GST-467 switch achieves superior classification accuracy on EMNIST and Fashion-MNIST benchmarks. These results position GST-467 as a highly promising PCM for scalable, low-energy photonic computing and neuromorphic hardware.