Wavelength-Multiplexed 2D Beam Steering via a Passive Diffractive Network

Problem
This work addresses the limitations of conventional optical systems that rely on single-layer dispersive elements for beam steering, which are restricted to one-dimensional linear mappings. The authors highlight the need for a more versatile approach to beam steering that can utilize multiple wavelengths for high-dimensional control. This paper is a preprint and has not undergone peer review, indicating that the findings should be interpreted with caution.

Method
The proposed architecture consists of a passive diffractive network composed of cascaded spatially optimized diffractive layers. These layers are designed using deep learning techniques to map distinct wavelengths to specific output angles. The network operates across 625 wavelength channels within the 400-750 nm range, achieving a 25 x 25 array of independently addressable beam positions. The design leverages complex wavefront transformations to facilitate nonlocal mappings, allowing for arbitrary wavelength-to-angle transformations without the need for mechanical scanning or electronic phase control. The authors validate their approach through numerical simulations and experimental setups in both terahertz and visible spectral regimes, employing 3D fabricated diffractive layers and phase-only spatial light modulators.

Results
The proposed diffractive network demonstrates subwavelength positioning accuracy and high channel fidelity across the specified wavelength range. The experimental results confirm the capability of the system to achieve wavelength-multiplexed beam steering, with the ability to control 625 distinct beam positions. The performance is quantitatively superior to traditional gratings, which are limited to linear trajectories, showcasing the advantages of the nonlocal wavefront transformation approach. Specific performance metrics, such as the exact positioning accuracy and fidelity measures, are not detailed in the abstract but are expected to be elaborated upon in the full paper.

Limitations
The authors acknowledge that the current implementation is limited to specific spectral ranges and may require further optimization for broader applications. Additionally, the reliance on deep learning for the design process introduces potential challenges related to generalization and robustness in varying operational conditions. The scalability of the architecture in real-world applications, particularly in dynamic environments, is also a concern that warrants further investigation.

Why it matters
This work establishes a new paradigm for high-speed programmable beam steering, with significant implications for various fields, including optical communications, imaging, and sensing. The ability to control beam positions using wavelength as an intrinsic parameter opens avenues for advanced photonic information-processing systems. The findings contribute to the growing body of research on diffractive optics and may inspire further innovations in passive optical devices, as published in arXiv cs.NE.