Search for: All records

Flat lenses with focal length tunability can enable the development of highly integrated imaging systems. This work explores machine learning to inverse design a multifocal multilevel diffractive lens (MMDL) by wavelength multiplexing. The MMDL output is multiplexed in three color channels, red (650 nm), green (550 nm), and blue (450 nm), to achieve varied focal lengths of 4 mm, 20 mm, and 40 mm at these three color channels, respectively. The focal lengths of the MMDL scale significantly with the wavelength in contrast to conventional diffractive lenses. The MMDL consists of concentric rings with equal widths and varied heights. The machine learning method is utilized to optimize the height of each concentric ring to obtain the desired phase distribution so as to achieve varied focal lengths multiplexed by wavelengths. The designed MMDL is fabricated through a direct-write laser lithography system with gray-scale exposure. The demonstrated singlet lens is miniature and polarization insensitive, and thus can potentially be applied in integrated optical imaging systems to achieve zooming functions.

 

Multilayer diffractive optical neural networks (DONNs) can perform machine learning (ML) tasks at the speed of light with low energy consumption. Decreasing the number of diffractive layers can reduce inevitable material and diffraction losses to improve system performance, and incorporating compact devices can reduce the system footprint. However, current analytical DONN models cannot accurately describe such physical systems. Here we show the ever-ignored effects of interlayer reflection and interpixel interaction on the deployment performance of DONNs through full-wave electromagnetic simulations and terahertz (THz) experiments. We demonstrate that the drop of handwritten digit classification accuracy due to reflection is negligible with conventional low-index THz polymer materials, while it can be substantial with high-index materials. We further show that one- and few-layer DONN systems can achieve high classification accuracy, but there is a trade-off between accuracy and model-system matching rate because of the fast-varying spatial distribution of optical responses in diffractive masks. Deep DONNs can break down such a trade-off because of reduced mask spatial complexity. Our results suggest that new accurate and trainable DONN models are needed to advance the development and deployment of compact DONN systems for sophisticated ML tasks.

 

Machine learning can empower the design of cascaded diffractive optical elements (DOEs) at terahertz frequencies enabling the realization of holograms with a tailored multi‐degree‐of‐freedom reconfigurable operation when altering either the number, spacing, rotational alignment, and/or order of the elements. This unprecedented control over the spatial terahertz light distribution can profoundly impact multiple terahertz applications such as signal multiplexing, imaging, and displays. This work demonstrates this multi‐degree‐of‐freedom control in structures fabricated through 3D printing employing low‐loss materials. The designs are validated through 3D finite‐difference time‐domain (FDTD) simulations and experimental measurements, showing that, in all cases, the desired diffraction patterns are generated.

 

In this paper, we discuss flat programmable multi-level diffractive lenses (PMDL) enabled by phase change materials working in the near-infrared and visible ranges. The high real part refractive index contrast (Δn ∼ 0.6) of Sb 2 S 3 between amorphous and crystalline states, and extremely low losses in the near-infrared, enable the PMDL to effectively shift the lens focus when the phase of the material is altered between its crystalline and amorphous states. In the visible band, although losses can become significant as the wavelength is reduced, the lenses can still provide good performance as a result of their relatively small thickness (∼ 1.5λ to 3λ). The PMDL consists of Sb 2 S 3 concentric rings with equal width and varying heights embedded in a glass substrate. The height of each concentric ring was optimized by a modified direct binary search algorithm. The proposed designs show the possibility of realizing programmable lenses at design wavelengths from the near-infrared (850 nm) up to the blue (450 nm) through engineering PMDLs with Sb 2 S 3 . Operation at these short wavelengths, to the best of our knowledge, has not been studied so far in reconfigurable lenses with phase-change materials. Therefore, our results open a wider range of applications for phase-change materials, and show the prospect of Sb 2 S 3 for such applications. The proposed lenses are polarization insensitive and can have the potential to be applied in dual-functionality devices, optical imaging, and biomedical science. 

This work discusses the design and fabrication of a dual-plane terahertz (THz) hologram and an extended-depth-of-focus THz diffractive lens. The dual-plane THz hologram consists of 50 × 50 diffractive optical elements with identical element pixel size 1×1 mm, and the extended-depth-of-focus THz diffractive lens is designed with 25 concentric rings with identical ring width of 1 mm, resulting in same device dimension 50 mm × 50 mm. The height of the hologram pixels and concentric rings of the diffractive lens are optimized by nonlinear optimization algorithms with scalar diffraction theory based on Ray-Sommerfeld diffraction equation. Finite-Difference Time-Domain (FDTD) simulation results agree with optimization results obtained from the scalar diffraction theory for both the THz hologram and the THz diffractive lens. The demonstrated experimental results show that the proposed THz hologram and THz diffractive lens can generate the desired diffraction patterns. These diffractive structures have the potential to be applied in areas such as THz imaging, data storage, and displays.

 

Abstract Deep neural networks (DNNs) have substantial computational requirements, which greatly limit their performance in resource-constrained environments. Recently, there are increasing efforts on optical neural networks and optical computing based DNNs hardware, which bring significant advantages for deep learning systems in terms of their power efficiency, parallelism and computational speed. Among them, free-space diffractive deep neural networks (D 2 NNs) based on the light diffraction, feature millions of neurons in each layer interconnected with neurons in neighboring layers. However, due to the challenge of implementing reconfigurability, deploying different DNNs algorithms requires re-building and duplicating the physical diffractive systems, which significantly degrades the hardware efficiency in practical application scenarios. Thus, this work proposes a novel hardware-software co-design method that enables first-of-its-like real-time multi-task learning in D 2 2NNs that automatically recognizes which task is being deployed in real-time. Our experimental results demonstrate significant improvements in versatility, hardware efficiency, and also demonstrate and quantify the robustness of proposed multi-task D 2 NN architecture under wide noise ranges of all system components. In addition, we propose a domain-specific regularization algorithm for training the proposed multi-task architecture, which can be used to flexibly adjust the desired performance for each task. 

In this work, we explore inverse designed reconfigurable digital metamaterial structures based on phase change material Sb₂Se₃for efficient and compact integrated nanophotonics. An exemplary design of a 1 × 2 optical switch consisting of a 3 µm x 3 µm pixelated domain is demonstrated. We show that: (i) direct optimization of a domain containing only Si and Sb₂Se₃pixels does not lead to a high extinction ratio between output ports in the amorphous state, which is owed to the small index contrast between Si and Sb₂Se₃in such a state. As a result, (ii) topology optimization, e.g., the addition of air pixels, is required to provide an initial asymmetry that aids the amorphous state's response. Furthermore, (iii) the combination of low loss and high refractive index change in Sb₂Se₃, which is unique among all phase change materials in the telecommunications 1550 nm band, translates into an excellent projected performance; the optimized device structure exhibits a low insertion loss (∼1.5 dB) and high extinction ratio (>18 dB) for both phase states.

 

While terahertz spectroscopy can provide valuable information regarding the charge transport properties in semiconductors, its application for the characterization of low-conductive two-dimensional layers, i.e., σ_s <  < 1 mS, remains elusive. This is primarily due to the low sensitivity of direct transmission measurements to such small sheet conductivity levels. In this work, we discuss harnessing the extraordinary optical transmission through gratings consisting of metallic stripes to characterize such low-conductive two-dimensional layers. We analyze the geometric tradeoffs in these structures and provide physical insights, ultimately leading to general design guidelines for experiments enabling non-contact, non-destructive, highly sensitive characterization of such layers.