Low-loss phase-change material based programmable mode converter for photonic computing

Phase-change materials (PCMs)-based integrated photonic memory offers a viable pathway for the development of neuromorphic computing chips. The sizable optical contrast in the telecom band between amorphous and crystalline phases of PCM, in particular, Ge2Sb2Te5 (GST), is used for multilevel programming. However, the high extinction coefficient k of crystalline GST leads to high optical loss, posing a serious challenge for scaling up the device array for practical use. In this work, we focus on the atomic understanding and application of the so-called low-loss PCM, Sb2Se3, through multiscale simulations. First, we elucidate the bonding origin of the wavelength-dependent optical properties of amorphous and crystalline Sb2Se3 via ab initio calculations. Given the suppressed k in the telecom band, we design a programmable mode converter (PMC) waveguide device that utilizes only the contrast in refractive index n between amorphous and crystalline Sb2Se3 to encode multiple optical levels per waveguide device. The finite-difference time-domain simulations show that a single PMC device can achieve 5-bit programming precision (32 levels) via direct laser writing, and the photonic tensor core formed by the PMC array could possibly be scaled to 128 × 128. Finally, a thorough comparison between low-loss PCM and conventional PCM is provided. The increasing demand for data storage and processing necessitates the development of low-power, non-volatile memory and neuromorphic computing devices. This study explores the potential of Sb2Se3, a low-loss phase-change material, for photonic computing applications. The authors employed ab initio simulations to investigate the bonding and optical properties of Sb2Se3, revealing its low-loss characteristics in the telecom band. The study predicts a programmable mode converter device using Sb2Se3, achieving 32 distinct states with an insertion loss of just −0.65 dB per node. This low-loss feature could allow for larger photonic array sizes, supporting matrix sizes greater than 128 × 128, which is crucial for efficient matrix-vector multiplication operations. The findings highlight Sb2Se3’s potential in scalable photonic computing, with future directions focusing on experimental validation and addressing material limitations such as cycling endurance and crystallization speed. Integrated photonic memory based on phase-change materials (PCM) offers a promising approach for neuromorphic computing. Here, we employ a multiscale simulation scheme, which combines density functional theory and finite-difference time-domain simulations, to investigate the low-loss origin of Sb2Se3, and design a programmable mode converter array with 5-bit precision and −0.65 dB insertion loss per node. This Sb2Se3-based photonic tensor core is predicted to be scalable to a large matrix size of 128 × 128, and simulated image recognition accuracy could be comparable to the pure software-based predictions. Our work is expected to stimulate the exploration of low-loss PCM for photonic computing.