Search

Multiply-accumulation (MAC) is a crucial computing operation in signal processing, numerical simulations, and machine learning. This work presents a scalable, programmable, frequency-domain parallel computing leveraging gigahertz (GHz)-frequency acoustic-wave nonlinearities. By encoding data in the frequency domain, a single nonlinear acoustic-wave device can perform a billion arithmetic operations simultaneously. A single device with a footprint of 0.03 mm$^2$ on lithium niobate (LN) achieves 0.0144 tera floating-point operations per second (TFLOPS), leading to a computing area density of 0.48 TFLOPS/mm$^2$ and a core power efficiency of 0.14 TFLOPS/Watt. As applications, we demonstrate multiplications of two 16-by-16 matrices and convolutional imaging processing of 128-by-128-pixel photos. Our technology could find versatile applications in near-sensor signal processing and edge computing.

Low-phase-noise microwave-frequency integrated oscillators provide compact solutions for various applications in signal processing, communications, and sensing. Surface acoustic waves (SAW), featuring orders-of-magnitude shorter wavelength than electromagnetic waves at the same frequency, enable integrated microwave-frequency systems with much smaller footprint on chip. SAW devices also allow higher quality (Q) factors than electronic components at room temperature. Here, we demonstrate a low-phase-noise gigahertz-frequency SAW oscillator on 128{\deg}Y-cut lithium niobate, where the SAW resonator occupies a footprint of 0.05 mm$^2$. Leveraging phononic crystal bandgap-edge modes to balance between Q factors and insertion losses, our 1-GHz SAW oscillator features a low phase noise of -132.5 dBc/Hz at a 10 kHz offset frequency and an overlapping Hadamard deviation of $6.5\times10^{-10}$ at an analysis time of 64 ms. The SAW resonator-based oscillator holds high potential in developing low-noise sensors and acousto-optic integrated circuits.

Deep neural networks (DNNs) have heavily relied on traditional computational units like CPUs and GPUs. However, this conventional approach brings significant computational burdens, latency issues, and high power consumption, limiting their effectiveness. This has sparked the need for lightweight networks like ExtremeC3Net. On the other hand, there have been notable advancements in optical computational units, particularly with metamaterials, offering the exciting prospect of energy-efficient neural networks operating at the speed of light. Yet, the digital design of metamaterial neural networks (MNNs) faces challenges such as precision, noise, and bandwidth, limiting their application to intuitive tasks and low-resolution images. In this paper, we propose a large kernel lightweight segmentation model, ExtremeMETA. Based on the ExtremeC3Net, the ExtremeMETA maximizes the ability of the first convolution layer by exploring a larger convolution kernel and multiple processing paths. With the proposed large kernel convolution model, we extend the optic neural network application boundary to the segmentation task. To further lighten the computation burden of the digital processing part, a set of model compression methods is applied to improve model efficiency in the inference stage. The experimental results on three publicly available datasets demonstrate that the optimized efficient design improved segmentation performance from 92.45 to 95.97 on mIoU while reducing computational FLOPs from 461.07 MMacs to 166.03 MMacs. The proposed the large kernel lightweight model ExtremeMETA showcases the hybrid design's ability on complex tasks.

Deep neural networks (DNNs) utilized recently are physically deployed with computational units (e.g., CPUs and GPUs). Such a design might lead to a heavy computational burden, significant latency, and intensive power consumption, which are critical limitations in applications such as the Internet of Things (IoT), edge computing, and the usage of drones. Recent advances in optical computational units (e.g., metamaterial) have shed light on energy-free and light-speed neural networks. However, the digital design of the metamaterial neural network (MNN) is fundamentally limited by its physical limitations, such as precision, noise, and bandwidth during fabrication. Moreover, the unique advantages of MNN's (e.g., light-speed computation) are not fully explored via standard 3x3 convolution kernels. In this paper, we propose a novel large kernel metamaterial neural network (LMNN) that maximizes the digital capacity of the state-of-the-art (SOTA) MNN with model re-parametrization and network compression, while also considering the optical limitation explicitly. The new digital learning scheme can maximize the learning capacity of MNN while modeling the physical restrictions of meta-optic. With the proposed LMNN, the computation cost of the convolutional front-end can be offloaded into fabricated optical hardware. The experimental results on two publicly available datasets demonstrate that the optimized hybrid design improved classification accuracy while reducing computational latency. The development of the proposed LMNN is a promising step towards the ultimate goal of energy-free and light-speed AI.

Rapid developments in machine vision have led to advances in a variety of industries, from medical image analysis to autonomous systems. These achievements, however, typically necessitate digital neural networks with heavy computational requirements, which are limited by high energy consumption and further hinder real-time decision-making when computation resources are not accessible. Here, we demonstrate an intelligent meta-imager that is designed to work in concert with a digital back-end to off-load computationally expensive convolution operations into high-speed and low-power optics. In this architecture, metasurfaces enable both angle and polarization multiplexing to create multiple information channels that perform positive and negatively valued convolution operations in a single shot. The meta-imager is employed for object classification, experimentally achieving 98.6% accurate classification of handwritten digits and 88.8% accuracy in classifying fashion images. With compactness, high speed, and low power consumption, this approach could find a wide range of applications in artificial intelligence and machine vision applications., Comment: 15 pages, 5 figures

Optical nonreciprocity is manifested as a difference in the transmission of light for the opposite directions of excitation. Nonreciprocal optics is traditionally realized with relatively bulky components such as optical isolators based on the Faraday rotation, hindering the miniaturization and integration of optical systems. Here we demonstrate free-space nonreciprocal transmission through a metasurface comprised of a two-dimensional array of nanoresonators made of silicon hybridized with vanadium dioxide (VO2). This effect arises from the magneto-electric coupling between Mie modes supported by the resonator. Nonreciprocal response of the nanoresonators occurs without the need for external bias; instead, reciprocity is broken by the incident light triggering the VO2 phase transition for only one direction of incidence. Nonreciprocal transmission is broadband covering over 100 nm in the telecommunication range in the vicinity of {\lambda}=1.5 um. Each nanoresonator unit cell occupies only about 0.1 {\lambda}^3 in volume, with the metasurface thickness measuring about half-a-micron. Our self-biased nanoresonators exhibit nonreciprocity down to very low levels of intensity on the order of 150 W/cm^2 or a uW per nanoresonator. We estimate picosecond-scale transmission fall times and sub-microsecond scale transmission rise. Our demonstration brings low-power, broadband and bias-free optical nonreciprocity to the nanoscale.

We study the resonant enhancement of linear and nonlinear chiroptical effects in asymmetric silicon metasurfaces supporting multipolar Mie resonances and quasi-bound states in the continuum (quasi-BICs). We demonstrate theoretically and observe in experiment the pronounced linear circular dichroism at the quasi-BIC resonances. We further find that both local field enhancement and third-harmonic signal are large for Mie resonances and some quasi-BIC modes. We explain the selectivity of the nonlinear enhancement by employing the concept of critical coupling being more favorable for the modes with moderately large radiative quality factors ($Q$ factors). We demonstrate experimentally strong nonlinear chiroptical response associated with high efficiency of the third-harmonic generation and large nonlinear circular dichroism varying from $+0.918\pm0.049$ to $-0.771\pm0.004$ for the samples with different asymmetries. We believe our results suggest a general strategy for engineering nonlinear chiroptical response in dielectric resonant metasurfaces., Comment: 23 pages, 5 figures

Rapid advances in deep learning have led to paradigm shifts in a number of fields, from medical image analysis to autonomous systems. These advances, however, have resulted in digital neural networks with large computational requirements, resulting in high energy consumption and limitations in real-time decision making when computation resources are limited. Here, we demonstrate a meta-optic based neural network accelerator that can off-load computationally expensive convolution operations into high-speed and low-power optics. In this architecture, metasurfaces enable both spatial multiplexing and additional information channels, such as polarization, in object classification. End-to-end design is used to co-optimize the optical and digital systems resulting in a robust classifier that achieves 95% accurate classification of handwriting digits and 94% accuracy in classifying both the digit and its polarization state. This approach could enable compact, high-speed, and low-power image and information processing systems for a wide range of applications in machine-vision and artificial intelligence., Comment: 16 pages, 5 figures

Terahertz waves offer a profound platform for next-generation sensing, imaging, and information communications. However, all conventional terahertz components and systems suffer from a bulky design, sensitivity to imperfections, and transmission losses. Here, we propose and experimentally demonstrate on-chip integration and miniaturization of topological devices which may address many existing drawbacks of the terahertz technology. We design and fabricate topological devices based on valley-Hall photonic structures that can be employed for various integrated components of on-chip terahertz systems. More specifically, we demonstrate the valley-locked asymmetric energy flow and mode conversion with topological straight waveguide, multi-port couplers, wave division, and whispering gallery mode resonators. Our devices are based on topological membrane metasurfaces which are of great importance for developing on-chip photonics and bringing many novel features into terahertz devices.

Metasurfaces, composed of subwavelength scattering elements, have demonstrated remarkable control over the transmitted amplitude, phase, and polarization of light. However, manipulating the amplitude upon transmission has required loss if a single metasurface is used. Here, we describe high-efficiency independent manipulation of the amplitude and phase of a beam using two lossless phase-only metasurfaces separated by a distance. With this configuration, we experimentally demonstrate optical components such as combined beam-forming and splitting devices, as well as those for forming complex-valued, three-dimensional holograms. The compound meta-optic platform provides a promising approach for achieving high performance optical holographic displays and compact optical components, while exhibiting a high overall efficiency.

Rare-earth doped nanocrystals are emerging light sources used for many applications in nanotechnology enabled by human ability to control their various optical properties with chemistry and material science. However, one important optical problem -- polarisation of photoluminescence -- remains largely out of control by chemistry methods. Control over photoluminescence polarisation can be gained via coupling of emitters to resonant nanostructures such as optical antennas and metasurfaces. However, the resulting polarization is typically sensitive to position disorder of emitters, which is difficult to mitigate. Recently, new classes of disorder-immune optical systems have been explored within the framework of topological photonics. Here we explore disorder-robust topological arrays of Mie-resonant nanoparticles for polarisation control of photoluminescence of nanocrystals. We demonstrate polarized emission from rare-earth-doped nanocrystals governed by photonic topological edge states supported by zigzag arrays of dielectric resonators. We verify the topological origin of polarised photoluminescence by comparing emission from nanoparticles coupled to topologically trivial and nontrivial arrays of nanoresonators.

With the advent of increasingly elaborate experimental techniques in physics, chemistry and materials sciences, measured data are becoming bigger and more complex. The observables are typically a function of several stimuli resulting in multidimensional data sets spanning a range of experimental parameters. As an example, a common approach to study ferroelectric switching is to observe effects of applied electric field, but switching can also be enacted by pressure and is influenced by strain fields, material composition, temperature, time, etc. Moreover, the parameters are usually interdependent, so that their decoupling toward univariate measurements or analysis may not be straightforward. On the other hand, both explicit and hidden parameters provide an opportunity to gain deeper insight into the measured properties, provided there exists a well-defined path to capture and analyze such data. Here, we introduce a new, two-dimensional approach to represent hysteretic response of a material system to applied electric field. Utilizing ferroelectric polarization as a model hysteretic property, we demonstrate how explicit consideration of electromechanical response to two rather than one control voltages enables significantly more transparent and robust interpretation of observed hysteresis, such as differentiating between charge trapping and ferroelectricity. Furthermore, we demonstrate how the new data representation readily fits into a variety of machinelearning methodologies, from unsupervised classification of the origins of hysteretic response via linear clustering algorithms to neural-network-based inference of the sample temperature based on the specific morphology of hysteresis.

Image processing has become a critical technology in a variety of science and engineering disciplines. While most image processing is performed digitally, optical analog processing has the advantages of being low-power and high-speed though it requires a large volume. Here, we demonstrate optical analog imaging processing using a flat optic for direct image differentiation allowing one to significantly shrink the required optical system size. We first demonstrate how the image differentiator can be combined with traditional imaging systems such as a commercial optical microscope and camera sensor for edge detection. Second, we demonstrate how the entire analog processing system can be realized as a monolithic compound flat optic by integrating the differentiator with a metalens. The compound nanophotonic system manifests the advantage of thin form factor optics as well as the ability to implement complex transfer functions and could open new opportunities in applications such as biological imaging and machine vision.

Metasurfaces based on resonant nanophotonic structures have enabled novel types of flat-optics devices often outperforming the capabilities of bulk components, yet these advances remain largely unexplored for quantum applications. We show that non-classical multi-photon interferences can be achieved at the subwavelength scale in all-dielectric metasurfaces. We simultaneously image multiple projections of quantum states with a single metasurface, enabling a robust reconstruction of amplitude, phase, coherence, and entanglement of multi-photon polarization-encoded states. One- and two-photon states are reconstructed through nonlocal photon correlation measurements with polarization-insensitive click-detectors positioned after the metasurface, and the scalability to higher photon numbers is established theoretically. Our work illustrates the feasibility of ultra-thin quantum metadevices for the manipulation and measurement of multi-photon quantum states with applications in free-space quantum imaging and communications., Comment: 19 pages, 14 figures

Expanding the use of physical degrees of freedom to employ spatial multiplexing of data in optical communication is considered the most disruptive and effective solution to meet the capacity demand of the growing information traffic. Development of space-division-multiplexing methods stimulated research on spatial modulation, detection and processing of data, attracting interest from various fields of science. A passive all-dielectric metasurface with near-unity transmission is used to engineer spatial mode profiles, potentially of arbitrary complexity. The broadband response of the metasurface covers S, C, and L bands of fibre communications. Unlike conventional phase plates, the metasurface allows for both phase and polarization conversion, providing full flexibility for mode engineering. We employ the metasurface for both mode modulation and mode multiplexing in free-space optical communication, and demonstrate that it is capable of mode multiplexing with an extinction ratio in excess of 20 dB over the whole C-band with negligible penalty even for 100 Gb/s DP-QPSK signals. These results merge two seemingly different fields, optical communication and metamaterials, and suggest a novel approach for ultimate miniaturisation of mode multiplexers and advanced LiFi technologies.

Semiconductor fabrication is a mainstay of modern civilization, enabling the myriad applications and technologies that underpin everyday life. However, while sub-10 nanometer devices are already entering the mainstream, the end of the Moore's Law roadmap still lacks tools capable of bulk semiconductor fabrication on sub-nanometer and atomic levels, with probe-based manipulation being explored as the only known pathway. Here we demonstrate that the atomic-sized focused beam of a scanning transmission electron microscope can be used to manipulate semiconductors such as Si on the atomic level, inducing growth of crystalline Si from the amorphous phase, reentrant amorphization, milling, and dopant-front motion. These phenomena are visualized in real time with atomic resolution. We further implement active feedback control based on real-time image analytics to control the e-beam motion, enabling shape control and providing a pathway for atom-by-atom correction of fabricated structures in the near future. These observations open a new epoch for atom-by-atom manufacturing in bulk, the long-held dream of nanotechnology.

Atomic Force Microscopy (AFM) methods utilizing resonant mechanical vibrations of cantilevers in contact with a sample surface have shown sensitivities as high as few picometers for detecting surface displacements. Such a high sensitivity is harnessed in several AFM imaging modes. Here, we demonstrate a cantilever-resonance-based method to quantify electrostatic forces on a probe arising in the presence of a surface potential or when a bias voltage is applied to the AFM probe. We find that the electrostatic forces acting on the probe tip apex can strongly dominate cantilever response in electromechanical measurements and produce signals equivalent to few pm of surface displacement. In combination with modeling, the measurements of the force were used to determine the strength of the electrical field at the probe tip apex in contact with a sample. We find an evidence that the electric field strength in the junctions is limited by about 0.5 V/nm. This field can be sufficiently strong to significantly influence material states and kinetic processes through charge injection, Maxwell stress, shifts of phase equilibria, and reduction of energy barriers for activated processes. Besides, the results provide a baseline for accounting for the effects of local electrostatic forces in electromechanical AFM measurements as well as offer additional means to probe ionic mobility and field-induced phenomena in solids., Comment: Corrected misprints in units of Table 1

Metadevices based on dielectric nanostructured surfaces with both electric and magnetic Mie-type resonances have resulted in the best efficiency to date for functional flat optics with only one disadvantage: a narrow operational bandwidth. Here we experimentally demonstrate broadband transparent all-dielectric metasurfaces for highly efficient polarization manipulation. We utilize the generalized Huygens principle, with a superposition of the scattering contributions from several electric and magnetic multipolar modes of the constituent meta-atoms, to achieve destructive interference in reflection over a large spectral bandwidth. By employing this novel concept, we demonstrate reflectionless (~90% transmission) half-wave plates, quarter-wave plates, and vector beam q-plates that can operate across multiple telecom bands with ~99% polarization conversion efficiency.

Metasurface analogues of electromagnetically induced transparency (EIT) have been a major focus of the nanophotonics field over the past several years due their ability to produce high quality factor (Q-factor) resonances for applications such as low-loss slow light devices and highly sensitive optical sensors. However, Ohmic loss limits the achievable Q-factors in conventional plasmonic EIT metasurfaces to values less than 10, significantly hampering device performance. Here, we report experimental demonstration of a classical analogue of EIT using all-dielectric silicon-based metasurfaces. Due to extremely low absorption loss and coherent interaction of neighboring meta-atoms, a record-high Q-factor of 483 is experimentally observed, leading to a refractive index sensor with a figure-of-merit (FOM) of 103. Furthermore, we show that the dielectric metasurfaces can be engineered to confine the optical field in either the silicon resonator or the environment, allowing one to tailor light-matter interaction at the nanoscale., Comment: Two errors have been corrected: 1. In the simulations, cross-polarized fields were mistakenly not included leading to artificially inflated Q-factors. 2. The measured arrays were later found to have a partially connected gap, preventing cross-polarized light scattering and preserving a high Q-factor. The new manuscript corrects these errors and examines fully connected ring-based resonators

Metamaterials offer unprecedented flexibility for manipulating the optical properties of matter, including the ability to access negative index, ultra-high index and chiral optical properties. Recently, metamaterials with near-zero refractive index have drawn much attention. Light inside such materials experiences no spatial phase change and extremely large phase velocity, properties that can be applied for realizing directional emission, tunneling waveguides, large area single mode devices, and electromagnetic cloaks. However, at optical frequencies previously demonstrated zero- or negative-refractive index metamaterials require the use of metallic inclusions, leading to large ohmic loss, a serious impediment to device applications. Here, we experimentally demonstrate an impedance matched zero-index metamaterial at optical frequencies based on purely dielectric constituents. Formed from stacked silicon rod unit cells, the metamaterial possesses a nearly isotropic low-index response leading to angular selectivity of transmission and directive emission from quantum dots placed within the material., Comment: 15 pages, 5 figures

An experimental study was conducted on controlling the growth mode of La0.7Sr0.3MnO3 thin films on SrTiO3 substrates using pulsed laser deposition (PLD) by tuning growth temperature, pressure and laser fluence. Different thin film morphology, crystallinity and stoichiometry have been observed depending on growth parameters. To understand the microscopic origin, the adatom nucleation, step advance processes and their relationship to film growth were theoretically analyzed and a growth diagram was constructed. Three boundaries between highly and poorly crystallized growth, 2D and 3D growth, stoichiometric and non-stoichiometric growth were identified in the growth diagram. A good fit of our experimental observation with the growth diagram was found. This case study demonstrates that a more comprehensive understanding of the growth mode in PLD is possible.

We have designed and built the first band-limited coronagraphic mask used for ground-based high-contrast imaging observations. The mask resides in the focal plane of the near-infrared camera PHARO at the Palomar Hale telescope and receives a well-corrected beam from an extreme adaptive optics system. Its performance on-sky with single stars is comparable to current state-of-the-art instruments: contrast levels of $\sim10^{-5}$ or better at 0.8" in $K_s$ after post-processing, depending on how well non-common-path errors are calibrated. However, given the mask's linear geometry, we are able to conduct additional unique science observations. Since the mask does not suffer from pointing errors down its long axis, it can suppress the light from two different stars simultaneously, such as the individual components of a spatially resolved binary star system, and search for faint tertiary companions. In this paper, we present the design of the mask, the science motivation for targeting binary stars, and our preliminary results, including the detection of a candidate M-dwarf tertiary companion orbiting the visual binary star HIP 48337, which we are continuing to monitor with astrometry to determine its association., Comment: Accepted to ApJ