The terahertz frequency band, which is typically defined as ranging from 0.3 to 10 THz, possesses unique properties, such as the ability to penetrate the optically opaque dielectric materials and relatively short wavelength. Combined with the fact that the molecular vibrations of many substances occur within the terahertz band, these properties make the terahertz a promising candidate for applications in imaging, sensing, spectroscopy, and broadband communications. Despite its significant potential, the practical use of the terahertz band in real-world scenarios has been limited. This limitation can be attributed to three primary factors: the absence of compact high-power terahertz transmitters, the lack of compact highly sensitive terahertz receivers, and the challenges posed by high free space loss and atmospheric absorption affecting terahertz signals. The deployment of beam steering techniques can overcome these challenges. Among these techniques, the reconfigurable reflectarray stands out as a viable solution, offering numerous advantages, including the elimination of the need for a waveguide-based feeding network. Furthermore, the reflectarray operates as a standalone device that is compatible with diverse free-space terahertz systems.

 In this work, I present a design for a reconfigurable 1D terahertz reflectarray. The reflectarray comprises 80 reflective elements, each with dimensions of 300 µm x 5000 µm. The reflective elements are arranged next to each other on their longer side and individually driven by 5-bit microelectromechanical systems (MEMS) actuators, allowing a maximum vertical displacement of 600 µm. To calculate the radiation pattern of the reflectarray, a novel mathematical model based on the Huygens-Fresnel principle is proposed.

Two configurations of the reflectarray are explored here. The first, an approximate blazed grating, achieves a maximum steering angle of ±56.4° from the reflectarray normal at 0.3 THz. However, the steering range decreases with increasing frequency, reaching a maximum steering angle of only ±11.5° at 1 THz. Despite discrete achievable steering angles and the appearance of grating lobes, this configuration maintains a high grating efficiency (>0.8) for most steering angles, enabling the reflectarray to attain a directivity comparable to an equivalent uniform linear array of isotropic radiators. To overcome the limitations of the approximate blazed grating, a genetic algorithm (GA) is employed to customize the reflectarray’s radiation pattern. The GA optimizes three key features: maximizing the directivity in an arbitrary direction, achieving 16.82 dBi at 0.3 THz; minimizing the sidelobe level, achieving -18.4 dB (5 dB lower than the innermost sidelobe level of -13.5 dB from the equivalent linear array); inserting nulls at specific angles while maintaining the main beam direction. Subsequently, a set of optimal solutions is obtained from a multi-objective GA optimization to navigate the trade-off behavior between the objectives of maximizing the directivity and minimizing the sidelobe level. Moreover, I conduct a comparative analysis between the approximate blazed grating and the GA-optimized reflectarray, focusing on their respective height profiles. This examination reveals that the GA-optimized reflectarray operates as a concealed grating structure with non-perfect periodicity, leading to an absence of grating lobes which are the common issues encountered in traditional grating structures. This unexpected outcome shows one of the advantages of the GA-based configuration.

Four reflectarray phantoms, i.e., reflectarrays without MEMS actuation systems are fabricated, and their planar and 3D radiation patterns are measured. Two of them are designed to approximate blazed gratings with two distinct diffraction angles, while the remaining two are GA-optimized to share the same main beam direction as the blazed gratings. The measured radiation patterns for all four phantoms closely align with the mathematical model predictions, validating the accuracy and reliability of the proposed model. Furthermore, the 3D radiation patterns of the GA-optimized phantoms exhibit an absence of grating lobes, in contrast to the approximate blazed grating phantoms that display such lobes.

Both configurations are applicable for broadband systems that do not require simultaneous bandwidth (e.g. FMCW-Radar). However, in applications requiring simultaneous bandwidth (e.g. communication), the achievable displacement of the reflector becomes the limiting factor. The proposed reflectarray offers complete elevation coverage within a 3 dB gain bandwidth of 10 GHz. As the bandwidth expands to 20 GHz, the steering range decreases to ±30°. In the context of bandwidth enhancement, a large actuation displacement proves beneficial only when the reflectarray emulates a mirror. Nevertheless, the constraints posed by the technological feasibility of actuation systems make achieving a substantial improvement in displacement of the reflector, thereby enabling a practical steering range with a mirror configuration, extremely challenging.