<!-- *** Custom HTML *** --> Automated detection of orbital angular momentum (OAM) can tremendously contribute to quantum optical experiments. We develop convolutional neural networks to identify and classify noisy images of Laguerre–Gaussian (LG) modes collected from two different experimental set ups. We investigate the classification performance measures of the predictive classification models for experimental conditions. The results demonstrate accuracy and specificity above 90% in classifying 16 LG modes for both experimental set ups. However, the F-score, sensitivity, and precision of the classification range from 57% to 92%, depending on the number of imperfections in the images obtained from the experiments. This research could enhance the application of OAM light in telecommunications, sensing, and high-resolution imaging systems.

This paper presents a comparative study of sampling methods within the FedHome framework, designed for personalized in-home health monitoring. FedHome leverages federated learning (FL) and generative convolutional autoencoders (GCAE) to train models on decentralized edge devices while prioritizing data privacy. A notable challenge in this domain is the class imbalance in health data, where critical events such as falls are underrepresented, adversely affecting model performance. To address this, the research evaluates six oversampling techniques using Stratified K-fold cross-validation: SMOTE, Borderline-SMOTE, Random OverSampler, SMOTE-Tomek, SVM-SMOTE, and SMOTE-ENN. These methods are tested on FedHome's public implementation over 200 training rounds with and without stratified K-fold cross-validation. The findings indicate that SMOTE-ENN achieves the most consistent test accuracy, with a standard deviation range of 0.0167-0.0176, demonstrating stable performance compared to other samplers. In contrast, SMOTE and SVM-SMOTE exhibit higher variability in performance, as reflected by their wider standard deviation ranges of 0.0157-0.0180 and 0.0155-0.0180, respectively. Similarly, the Random OverSampler method shows a significant deviation range of 0.0155-0.0176. SMOTE-Tomek, with a deviation range of 0.0160-0.0175, also shows greater stability but not as much as SMOTE-ENN. This finding highlights the potential of SMOTE-ENN to enhance the reliability and accuracy of personalized health monitoring systems within the FedHome framework.

Graphene attracts enormous interest for photonic applications as it provides a degree of freedom to manipulate electromagnetic waves. In this paper, we present new graphene-based aperiodic multilayer structures as selective, tunable, and switchable thermal emitters at infrared frequencies. For these optimized aperiodic thermal emitters, we investigate the effect of the chemical potential and number of graphene layers on the range of selectivity, tunability, and switchability of thermal emittance. We find that the proposed thermal emitters show about an order of magnitude narrower thermal band, e.g., improved selectivity. The tunability of thermal power emitted from the structure with 32 graphene layers is ∼3.5 times larger than that of the structure with eight graphene layers, changing from $\lambda =3.34\mu \mathrm{m}\mathrm{to}2.85\mu \mathrm{m}$ by increasing the chemical potential from 0.0 eV to 1.0 eV. We demonstrate that the arrangement with 32 graphene layers can decrease by ∼83% of the power emitted for $\lambda =3.34\mu \mathrm{m}$ , providing ∼4.5 times stronger switchability than for the structure with eight graphene layers. The electrically dynamic control of the proposed graphene-based aperiodic multilayer structures can pave the way for a new class of in situ wavelength selective, tunable, and switchable thermal sources.

This paper delves into the intricate world of whispering gallery mode (WGM) resonators within complex microsphere configurations, exploring their optical properties and behavior. Integrated with optical sensing and processing technology, WGM resonators offer compact size, high sensitivity, rapid response, and tunability. The study investigates the impact of configuration, size, excitation, polarization, and coupling effects on WGM properties. Notable findings include enhanced sensitivity in single microsphere resonators, influence of unequal sphere sizes and excitation locations on WGM modes, and higher quality factors (Q‐factors) in triangular three‐microsphere resonator configurations. Circular polarization was found to elevate Q‐factors, while the nine‐microsphere resonator configuration exhibited increased intensity of dominant WGM peaks with higher laser power, suppressing other peaks. These insights guide the design and optimization of microsphere resonator systems, positioning them for applications in sensing and optical information processing.

This paper presents an all-optical two-bit binary-to-gray code converter based on linear photonic crystal (PhC) waveguides aimed at improving efficiency and integration within binary neural network hardware accelerators. The converter leverages interference effects within 2D photonic crystal waveguides to achieve reliable binary-to-gray conversion with minimal power requirements and a compact design footprint. Simulation results exhibit exceptional extinction ratios of 9.49 dB and 10.11 dB for output bits G0 and G1, respectively, within a 10 μm × 10 μm footprint. Additionally, the device achieves a delay of 0.52 μs, representing a significant reduction in delay and energy consumption compared to conventional approaches. The proposed converter is compatible with photonic integrated circuits and offers dual functionality, supporting reversible Feynman logic for signal duplication and integration within binary neural networks. This compact, low-power converter marks a step forward in developing high-efficiency optical components for telecommunications, data processing, and neuromorphic computing applications.

We present a class of inverse-designed, aperiodic multilayer graphene-based perfect absorbers operating in the mid-infrared spectrum (3-5 μm), a range vital for atmospheric transparency and advanced sensing. Our design leverages a fixed material sequence-graphene, PPSU dielectric spacers, and PbSe layers on a gold substrate-while achieving precise spectral tunability solely through layer thickness variation, enabling absorption peak control in 0.25 μm steps without any change in material composition. This physical tunability allows scalable fabrication of wavelength-specific devices using a single manufacturing process. We further demonstrate electrical switchability by dynamically modulating graphene's chemical potential (µc from 0 eV to 1 eV), enabling absorption amplitude control and wavelength redshifting without structural alteration. The proposed absorber achieves > 99.9% efficiency using only five graphene layers in a compact ~ 2 μm stack, offering significant advantages in size, weight, power, and cost. Our hybrid micro-genetic inverse design algorithm enables this performance while preserving > 90% absorption at incidence angles up to 52°, supporting broad angular robustness. Extensive simulation and field distribution analyses confirm the role of plasmonic confinement and impedance matching. Additionally, we validate the design's fabrication tolerance and benchmark its performance against recent state-of-the-art absorbers. By combining advanced inverse design with nanophotonic structures, our work advances the field of mid-infrared absorbers, providing a scalable and efficient platform for next-generation optical devices.