Performing optical logic operations by a diffractive neural network

Abstract

Optical logic operations lie at the heart of optical computing, and they enable many applications such as ultrahighspeed information processing. However, the reported optical logic gates rely heavily on the precise control of input light signals, including their phase difference, polarization, and intensity and the size of the incident beams. Due to the complexity and difficulty in these precise controls, the two output optical logic states may suffer from an inherent instability and a low contrast ratio of intensity. Moreover, the miniaturization of optical logic gates becomes difficult if the extra bulky apparatus for these controls is considered. As such, it is desirable to get rid of these complicated controls and to achieve full logic functionality in a compact photonic system. Such a goal remains challenging. Here, we introduce a simple yet universal design strategy, capable of using plane waves as the incident signal, to perform optical logic operations via a diffractive neural network. Physically, the incident plane wave is first spatially encoded by a specific logic operation at the input layer and further decoded through the hidden layers, namely, a compound Huygens' metasurface. That is, the judiciously designed metasurface scatters the encoded light into one of two small designated areas at the output layer, which provides the information of output logic states. Importantly, after training of the diffractive neural network, all seven basic types of optical logic operations can be realized by the same metasurface. As a conceptual illustration, three logic operations (NOT, OR, and AND) are experimentally demonstrated at microwave frequencies.

Spectrally encoded single-pixel machine vision using diffractive networks

“…Unlike conventional optical components used in machine vision systems, we use diffractive layers that are composed of two-dimensional (2D) arrays of passive pixels, where the complex-valued transmission or reflection coefficients of individual pixels are independent learnable parameters that are optimized using a computer through deep learning and error backpropagation (24). The use of deep learning in optical information processing systems has emerged in various exciting directions including integrated photonics solutions (25)(26)(27)(28)(29)(30)(31)(32) and free-space optical platforms (33)(34)(35)(36)(37)(38)(39)(40)(41)(42) involving, e.g., the use of diffraction (21,(43)(44)(45)(46). In this work, we harnessed the native dispersion properties of matter and trained a set of diffractive layers using deep learning to all-optically process a continuum of wavelengths to transform the spatial features of different object classes into a set of unique wavelengths, each representing one data class.…”

Optoelectronic convolutional neural networks based on time-stretch method

“…While researchers have paid much their attention on further improving performances of electronic implemented CNNs, the technology of optoelectronics may provide another scheme to breakthrough today's bottleneck of implementing CNNs and other photonic based AI structures with higher computing speed and lower energy consumption [8][9][10][11][12][13][14][15][16]. Optical devices are not able to record data, so electronic counterparts are adopted to store feedback results and better control the whole system.…”

All-Dielectric Meta-optics for High-Efficiency Independent Amplitude and Phase Manipulation

“…Optimizing metasurface unit cell parameters using the adjoint optimization method 26 , optimizing the plane wave spectrum 27 , and optimizing equivalent current distributions 28 , have been used to design sequential metasurface systems. Additionally, a variety of applications have been demonstrated with sequential metasurface devices: aberration correction 29 , optical retro-reflection 30 , full-color holography 31 , and optical diffractive neural networks 32,33 . The compound meta-optic is formed by two lossless phase-only metasurfaces separated by a physically short distance of homogeneous dielectric, as shown in Figure 1(a).…”

“…Adding additional metasurfaces would allow more control over the optical field (e.g. multi-wavelength performance 6,22,26 , or diffractive neural networks 26,32,33 for multi-input multi-output applications), but only two metasurfaces are necessary for amplitude and phase control at a single wavelength. As a proof of concept, we report experimental demonstrations of meta-optics that combine beam-forming and splitting, and produce high-quality, threedimensional holograms.…”