End-to-End Diverse Metasurface Design and Evaluation Using an Invertible Neural Network.

Employing deep learning models to design high-performance metasurfaces has garnered significant attention due to its potential benefits in terms of accuracy and efficiency. A deep learning-based metasurface design framework typically comprises a forward prediction path for predicting optical responses and a backward retrieval path for generating geometrical configurations. In the forward design path, a specific geometrical configuration corresponds to a unique optical response. However, in the inverse design path, a single performance metric can correspond to multiple potential designs. This one-to-many mapping poses a significant challenge for deep learning models and can potentially impede their performance. Although representing the inverse path as a probabilistic distribution is a widely adopted method for tackling this problem, accurately capturing the posterior distribution to encompass all potential solutions remains an ongoing challenge. Furthermore, in most pioneering works, the forward and backward paths are captured using separate models. However, the knowledge acquired from the forward path does not contribute to the training of the backward model. This separation of models adds complexity to the system and can hinder the overall efficiency and effectiveness of the design framework. Here, we utilized an invertible neural network (INN) to simultaneously model both the forward and inverse process. Unlike other frameworks, INN focuses on the forward process and implicitly captures a probabilistic model for the inverse process. Given a specific optical response, the INN enables the recovery of the complete posterior over the parameter space. This capability allows for the generation of novel designs that are not present in the training data. Through the integration of the INN with the angular spectrum method, we have developed an efficient and automated end-to-end metasurface design and evaluation framework. This novel approach eliminates the need for human intervention and significantly speeds up the design process. Utilizing this advanced framework, we have effectively designed high-efficiency metalenses and dual-polarization metasurface holograms. This approach extends beyond dielectric metasurface design, serving as a general method for modeling optical inverse design problems in diverse optical fields.

Plasmon-Enhanced Photocatalytic CO2 Reduction for Higher-Order Hydrocarbon Generation Using Plasmonic Nano-Finger Arrays.

The carbon dioxide reduction reaction (CO2RR) is a promising method to both reduce greenhouse gas carbon dioxide (CO2) concentrations and provide an alternative to fossil fuel by converting water and CO2 into high-energy-density chemicals. Nevertheless, the CO2RR suffers from high chemical reaction barriers and low selectivity. Here we demonstrate that 4 nm gap plasmonic nano-finger arrays provide a reliable and repeatable plasmon-resonant photocatalyst for multiple-electrons reactions: the CO2RR to generate higher-order hydrocarbons. Electromagnetics simulation shows that hot spots with 10,000 light intensity enhancement can be achieved using nano-gap fingers under a resonant wavelength of 638 nm. From cryogenic 1H-NMR spectra, formic acid and acetic acid productions are observed with a nano-fingers array sample. After 1 h laser irradiation, we only observe the generation of formic acid in the liquid solution. While increasing the laser irradiation period, we observe both formic and acetic acid in the liquid solution. We also observe that laser irradiation at different wavelengths significantly affected the generation of formic acid and acetic acid. The ratio, 2.29, of the product concentration generated at the resonant wavelength 638 nm and the non-resonant wavelength 405 nm is close to the ratio, 4.93, of the generated hot electrons inside the TiO2 layer at different wavelengths from the electromagnetics simulation. This shows that product generation is related to the strength of localized electric fields.

Ultrafast Early Warning of Heart Attacks through Plasmon-Enhanced Raman Spectroscopy using Collapsible Nanofingers and Machine Learning.

As the leading cause of death, heart attacks result in millions of deaths annually, with no end in sight. Early intervention is the only strategy for rescuing lives threatened by heart disease. However, the detection time of the fastest heart-attack detection system is >15 min, which is too long considering the rapid passage of life. In this study, a machine learning (ML)-driven system with a simple process, low-cost, short detection time (only 10 s), and high precision is developed. By utilizing a functionalized nanofinger structure, even a trace amount of biomarker leaked before a heart attack can be captured. Additionally, enhanced Raman profiles are constructed for predictive analytics. Five ML models are developed to harness the useful characteristics of each Raman spectrum and provide early warnings of heart attacks with >98% accuracy. Through the strategic combination of nanofingers and ML algorithms, the proposed warning system accurately provides alerts on silent heart-attack attempts seconds ahead of actual attacks.

Hot Electron-Driven Photocatalysis Using Sub-5 nm Gap Plasmonic Nanofinger Arrays.

Semiconductor photocatalysis has received increasing attention because of its potential to address problems related to the energy crisis and environmental issues. However, conventional semiconductor photocatalysts, such as TiO2 and ZnO, can only be activated by ultraviolet light due to their wide band gap. To extend the light absorption into the visible range, the localized surface plasmon resonance (LSPR) effect of noble metal nanoparticles (NPs) has been widely used. Noble metal NPs can couple incident visible light energy to strong LSPR, and the nonradiative decay of LSPR generates nonthermal hot carriers that can be injected into adjacent semiconductor material to enhance its photocatalytic activity. Here we demonstrate that nanoimprint-defined gap plasmonic nanofinger arrays can function as visible light-driven plasmonic photocatalysts. The sub-5 nm gaps between pairs of collapsed nanofingers can support ultra-strong plasmon resonance and thus boost the population of hot carriers. The semiconductor material is exactly placed at the hot spots, providing an efficient pathway for hot carrier injection from plasmonic metal to catalytic materials. This nanostructure thus exhibits high plasmon-enhanced photocatalytic activity under visible light. The hot carrier injection mechanism of this platform was systematically investigated. The plasmonic enhancement factor was calculated using the finite-difference time-domain (FDTD) method and was consistent with the measured improvement of the photocatalytic activity. This platform, benefiting from the precise controllable geometry, provides a deeper understanding of the mechanism of plasmonic photocatalysis.

A memristor-based hybrid analog-digital computing platform for mobile robotics.

Algorithms for mobile robotic systems are generally implemented on purely digital computing platforms. Developing alternative computational platforms may lead to more energy-efficient and responsive mobile robotics. Here, we report a hybrid analog-digital computing platform enabled by memristors on a mobile inverted pendulum robot. Our mobile robotic system can tune the conductance states of memristors adaptively using a model-free optimization method to achieve optimal control performance. We implement sensor fusion and the motion control algorithms on our hybrid analog-digital computing platform and demonstrate more than one order of magnitude enhancement of speed and energy efficiency over traditional digital platforms.

Detection of Fake Alcoholic Beverages Using Electrolyte-Free Nanogap Electrochemical Cells.

Because of the similarity of odor, appearance, and chemical structure of methanol and ethanol, measuring the low concentration of methanol in an alcoholic beverage is difficult to perform in a quick, quantitative, and repeatable fashion. However, it is important for people to monitor the content of methanol in a liquor because a high amount of methanol absorbed will result in blindness, coma, and death. In response to this need, we have developed electrolyte-free methanol electrolysis and ethanol electrolysis based on the nanogap electrochemical cells for the methanol and ethanol sensing. Upon applying a voltage, a high electric field across the nanogap cell enhances the solution ionization and the ion transport rate. Moreover, the nanoscale distance between the electrodes provides a shorter path for electrolysis to easily occur. The nanogap electrochemical cells not only make the direct electrolyte-free organic solvent electrolysis possible but also enhance the sensitivity of the chemical of interest in low-concentration solutions without the influence of the added electrolyte. The nanogap electrochemical cells have been demonstrated having high sensitivity to detect 0.15% methanol volume concentration in deionized water solutions without adding any electrolyte, and its ability for the fake alcoholic beverages' detection has successfully demonstrated.