These authors contributed equally to this work.Artificial Neural Networks are computational network models inspired by signal processing in the brain. These models have dramatically improved the performance of many learning tasks, including speech and object recognition. However, today's computing hardware is inefficient at implementing neural networks, in large part because much of it was designed for von Neumann computing schemes. Significant effort has been made to develop electronic architectures tuned to implement artificial neural networks that improve upon both computational speed and energy efficiency. Here, we propose a new architecture for a fully-optical neural network that, using unique advantages of optics, promises a computational speed enhancement of at least two orders of magnitude over the state-of-the-art and three orders of magnitude in power efficiency for conventional learning tasks. We experimentally demonstrate essential parts of our architecture using a programmable nanophotonic processor.

These authors contributed equally to this work.Artificial Neural Networks are computational network models inspired by signal processing in the brain. These models have dramatically improved the performance of many learning tasks, including speech and object recognition. However, today's computing hardware is inefficient at implementing neural networks, in large part because much of it was designed for von Neumann computing schemes. Significant effort has been made to develop electronic architectures tuned to implement artificial neural networks that improve upon both computational speed and energy efficiency. Here, we propose a new architecture for a fully-optical neural network that, using unique advantages of optics, promises a computational speed enhancement of at least two orders of magnitude over the state-of-the-art and three orders of magnitude in power efficiency for conventional learning tasks. We experimentally demonstrate essential parts of our architecture using a programmable nanophotonic processor.Modern computers based on the von Neumann architecture are far more power-hungry and less effective than their biological counterparts -central nervous systemsfor a wide range of tasks including perception, communication, learning, and decision making. With the increasing data volume associated with processing big data, developing computers that learn, combine, and analyze vast amounts of information quickly and efficiently is becoming increasingly important. For example, speech recognition software (e.g., Apple's Siri) is typically executed in the cloud since these computations are too taxing for mobile hardware; realtime image processing is an even more demanding task [1]. To address the shortcomings of von Neumann computing architectures for neural networks, much recent work has focused on increasing artificial neural network computing speed and power efficiency by developing electronic architectures (such as ASIC and FPGA chips) specifically tailored to a task [2][3][4][5]. Recent demonstrations of electronic neuromorphic hardware architectures have reported improved computational performance [6]. Hybrid optical-electronic systems that implement spike processing [7][8][9] and reservoir computing [10,11] have also been investigated recently. However, the computational speed and power efficiency achieved with these hardware architectures are still limited by electronic clock rates and ohmic losses.Fully-optical neural networks offer a promising alternative approach to microelectronic and hybrid optical-electronic implementations. Linear transformations (and certain nonlinear transformations) can be performed at the speed of light and detected at rates exceeding 100 GHz [12] in photonic networks, and in some cases, with minimal power consumption [13]. For example, it is well known that a common lens performs Fourier transform without any power consumption, and that certain matrix operations can also be performed optically without consuming power. However, implementing such transformations with bulk opti...

New deep learning techniques may hold the key to solving intractable photonics problems.

A fundamental building block for nanophotonics is the ability to achieve negative refraction of polaritons, because this could enable the demonstration of many unique nanoscale applications such as deep-subwavelength imaging, superlens, and novel guiding. However, to achieve negative refraction of highly squeezed polaritons, such as plasmon polaritons in graphene and phonon polaritons in boron nitride (BN) with their wavelengths squeezed by a factor over 100, requires the ability to flip the sign of their group velocity at will, which is challenging. Here we reveal that the strong coupling between plasmon and phonon polaritons in graphene-BN heterostructures can be used to flip the sign of the group velocity of the resulting hybrid (plasmon-phonon-polariton) modes. We predict allangle negative refraction between plasmon and phonon polaritons and, even more surprisingly, between hybrid graphene plasmons and between hybrid phonon polaritons. Graphene-BN heterostructures thus provide a versatile platform for the design of nanometasurfaces and nanoimaging elements.negative refraction | plasmon polariton | phonon polariton | graphene-boron nitride heterostructure P olaritons with high spatial confinement, such as plasmon polaritons in graphene (1-5) and phonon polaritons in a thin hexagonal boron nitride (BN) slab (6-15), enable control over the propagation of light at the extreme nanoscale, due to their in-plane polaritonic wavelength that can be squeezed by a factor over 100. Henceforth we use the term squeezing factor (or confinement factor) to define the ratio between the wavelength in free space and the in-plane polaritonic wavelength. The combination of tunability, low losses, and ultraconfinement (1,2,8,10,11,15) makes them superior alternatives to conventional metal plasmons and highly appealing for nanophotonic applications (3-5, 10-13, 15). Their extreme spatial confinement, however, also limits our ability to tailor their dispersion relations.Unlike the case of 2D plasmons, the coupling between metal plasmons in a metal-dielectric-metal structure dramatically changes their dispersion relation and can even flip the sign of their group velocities (16,17). This has led to exciting applications by tailoring the in-plane plasmonic refraction, giving flexibility in controlling the energy flow of light. Specifically, by flipping the sign of the group velocity of metal plasmons, plasmonic negative refraction has been predicted (16) and demonstrated (17). The negative refraction has also been extensively explored in metamaterials, metasurfaces, and photonic crystals (18-26), but they become experimentally very challenging to realize when dealing with polaritons with high squeezing factors. In contrast to metal plasmons, the group velocity of graphene plasmons (2,11,27) and all other 2D plasmons (28-32) is always positive, including that in graphene-based multilayer structures (33). This has made the in-plane negative refraction for highly squeezed 2D plasmon polaritons seem impossible to achieve.Contrary to 2D plas...

Abstract:Light selection based purely on the angle of propagation is a long-standing scientific challenge. In angularly selective systems, however, the transmission of light usually also depends on the light frequency. We tailored the overlap of the bandgaps of multiple one-dimensional photonic crystals, each with a different periodicity, in such a way as to preserve the characteristic Brewster modes across a broadband spectrum. We provide theory as well as experimental realization with an all-visible-spectrum, p-polarized angularly selective material system. Our method enables transparency throughout the visible spectrum at one angle, the generalized Brewster angle, and reflection at every other viewing angle. 1The ability to control light has long been a major scientific and technological goal. In electromagnetic theory, a monochromatic electromagnetic plane wave is characterized (apart from its phase and amplitude) by three fundamental properties: its frequency, its polarization, and its propagation direction. The ability to select light according to each of these separate properties would be an essential step in achieving control over light (Fig. 1).Tremendous progress has been made toward both frequency selectivity and polarization selectivity. Frequency selectivity (Fig. 1A) can be obtained, for example, by taking advantage of photonic bandgaps in photonic crystals (1-5). Polarization selectivity (Fig. 1B) is accomplished for example by means of a "wire grid" polarizer (6) or by exploiting birefringent materials (7,8).Methods based on interference and resonance effects have been explored for angular selectivity, but they have limited applications because they are sensitive to frequency.An angularly selective material-system should ideally work over a broadband spectrum.Such a system could potentially play a crucial role in many applications, such as high efficiency solar energy conversion (9, 10), privacy protection (11), and detectors with high signal-to-noise ratios. Some progress has been made towards achieving broadband angular selectivity by means of metallic extraordinary transmission (12, 13), anisotropic metamaterials (14), combined use of polarizers and birefringent films (11), or geometrical optics at micrometer scale (15). The first two of these mechanism are difficult to realize in the optical regime; the other two work only as angularly selective absorbers.Here, we introduce a basic principle to achieve optical broadband angular selectivity. Our result rests on i) the fact that polarized light transmits without any reflection at the Brewster angle, ii) the existence in photonic crystals of band gaps that prevent light propagation for given frequency ranges, and iii) the bandgap broadening effect of heterostructures. First, we prove our fundamental idea theoretically for a single polarization and oblique incident angles, and also for both polarizations and normal angle of incidence. Second, we experimentally demonstrate the concept in the case of all-visible spectrum, p-polarized light. The dem...

Structural coloration is an interference phenomenon where colors emerge when visible light interacts with nanoscopically structured material, and has recently become a most interesting scientific and engineering topic. However, current structural color generation mechanisms either require thick (compared to the wavelength) structures or lack dynamic tunability. This report proposes a new structural color generation mechanism, that produces colors by the Fano resonance effect on thin photonic crystal slab. We experimentally realize the proposed idea by fabricating the samples that show resonance-induced colors with weak dependence on the viewing angle. Finally, we show that the resonance-induced colors can be dynamically tuned by stretching the photonic crystal slab fabricated on an elastic substrate

Demonstrations of passive daytime radiative cooling have primarily relied on complex and costly spectrally selective nanophotonic structures with high emissivity in the transparent atmospheric spectral window and high reflectivity in the solar spectrum. Here, we show a directional approach to passive radiative cooling that exploits the angular confinement of solar irradiation in the sky to achieve sub-ambient cooling during the day regardless of the emitter properties in the solar spectrum. We experimentally demonstrate this approach using a setup comprising a polished aluminum disk that reflects direct solar irradiation and a white infrared-transparent polyethylene convection cover that minimizes diffuse solar irradiation. Measurements performed around solar noon show a minimum temperature of 6 °C below ambient temperature and maximum cooling power of 45 W m–2. Our passive cooling approach, realized using commonly available low-cost materials, could improve the performance of existing cooling systems and enable next-generation thermal management and refrigeration solutions.

Artificial spin ice has been recently implemented in two-dimensional arrays of mesoscopic magnetic wires. We propose a theoretical model of magnetization dynamics in artificial spin ice under the action of an applied magnetic field. Magnetization reversal is mediated by domain walls carrying two units of magnetic charge. They are emitted by lattice junctions when the local field exceeds a critical value Hc required to pull apart magnetic charges of opposite sign. Positive feedback from Coulomb interactions between magnetic charges induces avalanches in magnetization reversal.