Neuromorphic computing, drawing inspiration from the brain, stands out for its high energy efficiency in executing complex tasks. Memristive device-based neuromorphic computing has demonstrated ultrahigh efficiency. While there are numerous review papers in this field, the majority concentrate on the device level, bypassing the connections among the performance metrics of memristive devices and those of neuromorphic chips. In this review, we investigate the recent progress in neuromorphic computing from the fundamental memristive devices to the intricate neuromorphic chips, highlighting their links and challenges.

Coupled subwavelength silicon Mie resonators have shown the low loss waveguiding and slow light behavior.

Tunable devices based on bilayer metasurfaces have attracted researchers’ attention in recent years for their accurate tuning abilities and high integration. In tunable devices such as tunable beam splitters and Alvarez metalenses, opposite quadratic or cubic target phase profiles are imparted on both layers, and a varying total phase profile arises through the relatively lateral displacement between the two layers. However, there is a lack of a generalized target phase profile design method to design these tunable devices. In this study, a generalized phase profile design method named Integral of Total Phase Profile Difference (ITPD) is proposed to calculate the target phase profiles of both layers. Multiple integral equations describe the relationship between the target phase profiles and the total phase profiles. Based on this method, a tunable beam splitter and an Alvarez metalens are redesigned respectively. Moreover, a new tunable device that can be converted from a beam splitter to a metalens is designed by the ITPD method. The ITPD design method is promising for designing tunable devices with arbitrary total phase profiles in dynamic or multifunctional optical systems.

We propose a flat retroreflector that can efficiently reflect the electromagnetic waves back along its incident direction in a wide continuous range of angles. This retroreflector consists of a quadratic metalens and a flat metallic reflector at the focal plane of the former. The quadratic metalens is a dielectric pillar array encoded with a quadratic phase profile and it is embedded in the top side of the substrate. The flat reflector is on the bottom side of the substrate. The designed retroreflector has a diameter of 40 mm, a thickness of 15 mm, and a working frequency of 77 GHz. Through meta-units optimization, a retroreflection efficiency of 38.51% at ± 60° incidence and an average retroreflection efficiency of 46.39% for the incident angles from 0° to 60° can be numerically demonstrated. This flat retroreflector is easy for integration, which is promising for potential applications in the miniature wireless communication systems.

We present a beam homogenizer utilizing on a random metalens array (RMA) for enhanced beam shaping. The RMA, comprising 28 × 28 individual metalenses with random sizes, is designed using an improved Gerchberg-Saxton algorithm. As a laser beam traverses the RMA, it is divided into 28 × 28 beams, each exhibiting a unique speckle pattern. These beams overlap in the far field, significantly enhancing beam uniformity. The RMA demonstrates adaptability to various incident light modes and misalignment variations. The numerical simulations indicate over 90% uniformity in the shaped beams, while in experimental results show a reshaped line beam achieving 95.9% uniformity. This beam homogenizer has potential applications in minimizing the size of LiDAR systems or laser engraving systems, offering broader utility across these fields.