Photonic neural networks have been developed as a hardware platform to accelerate machine-learning inference. Digital micromirror devices (DMDs) have been playing a critical role in developing a variety of photonic neural networks for their ability to manipulate millions of optical spatial modes in a 2D plane at >10 kHz frame rate. DMDs have not only enabled high-throughput machine-learning inference but also made hardware-in-the-loop training possible with photonic neural networks. In this talk, we will review the functions of DMDs in a plethora of photonic-neural-network architectures and discuss how MEMS-based technologies can enable novel photonic neural networks in future.

Digital micromirror devices are versatile, high performance photonic components that combine high configurability with a large number of programmable parameters and high bandwidth. These are essential features in photonic neural networks. A DMD's mirrors can optically encode information to be injected into a photonics neural network, or they can even provide configurable connections between photonic neurons of the neural network itself. Their easy programmability makes them highly attractive, as through this feature DMDs act as the interface between the analogue world of the photonic neural network, and the digital world of programming languages as well as information processing. I will introduce several of such DMD-based operations in photonic AI will sketch future possibilities for the development of the field.

Two-photon photopolymerization (TPP) has emerged as a popular method for three-dimensional (3D) printing of micro-/nano-structures. On the other hand, optical diffraction tomography (ODT) has become attractive for label-free cell imaging by reconstructing the 3D refractive index (RI) distributions. We propose a high-speed ODT method to fully characterize the TPP-printed structures. Compared with traditional methods like scanning electron microscopy and atomic force microscopy, our ODT-based characterization method offers many advantages, such as revealing both the internal and external morphological features, mapping the 3D RI distributions, and quantifying the surface and side-wall roughness.

Maskless digital photolithography is a relatively new lithographic technique in which micrometer-scale pattern layouts can be configured conveniently by an electrically-addressable digital-micromirror-device. The authors’ group has been systematically looking into the possibility of generating submicron patterns using the technique, starting with photonic crystal (PhC) laser structures. Pattern tilting and grayscale exposure techniques are employed to fine-tune the PhC lattice constant and also to generate PhC patterns other than square-lattice structure. Diffraction effects are also taken into account carefully to generate non-periodic PhC patterns. The resultant PhC lasers exhibit laser performance comparable to those fabricated by electron-beam lithography.

In this talk, we will present our observations of printing kinetics in light-based tomographic additive manufacturing using optical scattering tomography. In particular we report a feature-size dependence on polymerization time that contributes significantly to errors in the printed object: small features tend to polymerize more slowly than large features. Therefore, prints are either missing small features or large features are overexposed. We investigate the cause of this feature size polymerization time dependence and present techniques to correct for these errors.

In this talk, we present a new methodology for computing projections in tomographic additive manufacturing. Currently, tomographic printing systems require that light-rays in the printing volume are parallel, and have low etendue. In this work, we show that accurate modeling of the light rays through the print volume enables improved printing in systems with diverging beams. We also demonstrate that ray-tracing can compensate for non-parallel projection in 3D. We anticipate that our ray-tracing methodology will relax the hardware requirements necessary in the conventional Radon-based approach, and enable a broader range of tomographic printing configurations.

Texas-Instruments phase-light-modulator (TI-PLM) is composed of a 2D array of micro-mirrors that move in a piston motion by a fraction of wavelength to modulate the optical path length of a beam. Such light phase modulation, which is required for holography, could be a promising alternative to the liquid crystal-based spatial-light-modulators (SLMs). In comparison, TI-PLM has better diffraction efficiency, enlarged reflection angle, reduced pixel size, and faster refresh speed which expands the spatial bandwidth. All these properties benefit the recording of complex holograms. We demonstrate the successful use of TI-PLM to record a variety of such holograms.