Researchers are exploring six core performance dimensions to create next-generation optoelectronic computing chips. These chips aim to improve computing power, accuracy, and energy efficiency by leveraging photons' strengths. The development of these chips is crucial for advancing artificial intelligence and high-performance computing.
Recent progress in terahertz meta-device design is fueled by machine learning and deep learning approaches, promising improved performance and multifunctionality. However, challenges persist, including computational pressure, interpretability, and fabrication constraints.
A new framework, DeepMoCo, uses Graph Neural Networks to correct motion artifacts in in vivo imaging, achieving both global alignment and local deformation recovery simultaneously. This enables high-fidelity monitoring of brain activity in freely behaving animals.
Researchers developed a multimode band-engineering framework in 1D grating-based photonic crystal waveguides to meet growing demand for scalable and flexible filtering. Devices achieved TE-1-, TE-2-, and TE-3-pass filtering with measured bandwidths exceeding 200 nm, 123 nm, and 70 nm, respectively.
Deep learning enables clearer images, faster biological dynamics, and intelligent analysis while reducing labeling demands and preserving biological structure. The review highlights the need for reliable and physics-informed reconstruction, including evaluation metrics and standardized protocols.
Optical pre-sensor computing reduces redundant data acquisition and transmission by extracting task-relevant features before sensor detection, easing communication bandwidth demands and processing latency. This approach enables efficient visual perception for robots and other physical AI systems.
Researchers developed a thin-film lithium niobate integrated photonic receiver for high-resolution dual-band radar, achieving a 12 GHz equivalent bandwidth and 1.27 cm ranging resolution. The device integrates electro-optic modulation and microwave demultiplexing on a single chip, reducing system size and weight.
Recent advances in photobiomodulation and nanomedicine integration highlight the therapeutic potential of non-photoactive nanoparticles. The review proposes possible mechanisms underlying their interactions, including regulation of reactive oxygen species, mitochondrial function, and cellular signaling networks.
The study demonstrates the robustness of valley photonic molecular crystals (VPMCs) to strong rotational disorder, opening new avenues for integrated photonic devices. Experimental validation on silicon chips confirmed the existence of topological edge states without relying on structural periodicity.
Researchers developed a compact 0.36-mm² photonic chip for industrial swept-source OCT inspection, achieving high-precision 3D imaging and contactless measurement. The system's sensitivity reached 87 dB and successfully resolved internal structures, with an absolute error of less than 2 μm in thickness measurements.
Researchers have developed a compact Kerr microcomb that generates low-noise terahertz frequency combs and supports high-speed wireless communication. The platform combines compact packaging, low-noise terahertz synthesis, and high-order wireless modulation, enabling a scalable approach for future terahertz systems.
Scientists created a high-bandwidth optical interconnect architecture to address AI computing's growing demands. The O-band silicon photonic interconnect combines wavelength and port multiplexing to enable terabit-scale optical interconnects.
Advances in photonic instrumentation are reshaping Brillouin imaging capabilities, enabling dynamic imaging, three-dimensional measurements, and complex biological specimens. However, translating optical measurements into reliable biological interpretation requires rigorous interpretation of Brillouin observables.
A feasible scheme is proposed to realize self-healing skin modes in photonic Floquet lattices via local potential at remote end. The researchers demonstrated the skin mode tunability (SMT) mechanism, which can spectrally isolate a specific skin mode and turn it into a self-healing state.
Scientists create EUV spatiotemporal skyrmions using HHG, which enables control of their topological structure and polarization-sensitive interaction with matter. This breakthrough opens new avenues for attosecond science and EUV optics.
A novel ultrasensitive fluorescent voltage probe, HB mito Crimson, detects fast and localized electrical signals in mitochondria. This probe enables monitoring of near-field potential signals at the inner mitochondrial membrane interface, revealing a class of transient miniature mitochondrial electric flickers.
Researchers have introduced topological couplers into coupled resonator optical waveguides, decoupling free spectral range and finesse to alleviate the bandwidth tradeoff. This innovation leads to a significant improvement in bandwidth, enabling applications in wavelength division multiplexing and nonlinear optical processing.
Light-field microscopy (LFM) enables snapshot volumetric acquisition, allowing for faster synchronous three-dimensional information. Recent advances have made LFM practical for high-speed neuroimaging, particularly in calcium imaging and voltage imaging.
Single-molecule localization microscopy is vulnerable to thermal drift, mechanical vibrations, and optical fluctuations. Established drift correction methods are categorized and hybrid strategies combining multiple techniques are discussed.
Researchers developed a label-free imaging strategy that reveals cellular states and dynamic heterogeneity without added labels. Spectral exponent analysis of stochastic fluctuations in iSCAT image sequences distinguishes cell types and tracks dynamic transitions during mitotic progression.
Raman microscopy generates molecular fingerprints from biological samples without stains, revealing cellular composition and real-time changes. The technology has advanced significantly, with applications in precision medicine, single-molecule analysis, and multidimensional analysis.
Researchers developed DUET, a motion-robust dual-color microscope that overcomes challenges in neuroscience imaging. The system achieves stable dual-color imaging with minimal crosstalk, enabling cell-type-specific activity patterns to be studied during natural behaviors.
A new AI chip using spectral convolutional neural networks (SCNN) analyzes meibomian gland tissue for diagnosis of dry eye disease, outperforming traditional methods and capturing molecular information in a single measurement.
Researchers developed a compact silicon photonic 'Vernier Caliper' spectrometer with near-uniform resonator responses across a broad wavelength range. The device resolves fine spectral peaks as little as 0.74 pm, outperforming commercial benchtop spectrometers in accuracy.
A global research team developed an 'All-Surface' 3D laser patterning technique to directly draw highly conductive carbon circuitry on transparent glass. This technology overcomes conventional 2D lithography limits, enabling the fabrication of complex 3D interconnected electrodes and redistribution layers.
Researchers developed Adaptive-SN2N, a risk-aware adaptive deep learning framework that resolves critical background artifact challenges. The framework integrates risk-aware adaptive normalization, self-inspired learning, and Gaussian-weighted overlap inference to suppress artifacts while improving photon efficiency.
Researchers develop an event-based framework for functional biological imaging, capturing vascular and neuronal dynamics at high temporal resolution. An unsupervised reconstruction algorithm maps sparse event data into smooth functional images.
Researchers developed LightELF to detect phase singularities and encode information in optical topological knots. The system uses event-driven mechanism for high-speed detection and decoding of knot-based optical signals.
Researchers develop a sub-terahertz biosensor that leverages band folding to distinguish cancer cells based on their unique dielectric properties, enabling rapid and non-invasive screening. The sensor detects distinct shifts in transmission spectra as malignancy increases, bridging physics and biology.
The Janus meta-imager uses diffractive deep neural networks and metasurfaces to perform asymmetric imaging tasks in both forward and backward directions. It enables all-optical encryption and high-density optical data storage with minimal energy consumption.
Researchers report a novel intersection among quantum optics, atomic physics, and precision metrology with the achievement of sub-shot-noise optical readout in a Rydberg atomic medium. The study utilizes velocity-selective Rydberg EIT to enable low-loss optical interface and coherent propagation of squeezed light.
Researchers have developed a hybrid-phase cooperative dispersion-engineering strategy to unlock independent, dual-spin achromatic wavefront control in a single-layer metasurface. This breakthrough enables truly independent dual-spin control of phase and group delay within a compact platform.
A compact sensor achieves multispecies gas detection with high selectivity and sensitivity, utilizing nanomaterial-functionalized fibers and Kerr soliton dual-microcombs. The system successfully identified 12 distinct components in complex gas mixtures, achieving record-low detection limits.
Researchers propose a general method to grow multi-layer scintillators with high uniformity and radiation stability. The technique allows for efficient coupling of scintillators with specific energy X-ray response characteristics, enabling high-resolution multi-energy X-ray imaging for various applications.
Researchers have developed a compact atomic magnetometer that uses a phase-gradient metasurface to detect ultra-weak magnetic fields with high sensitivity. The new design achieves a sensitivity of 2.67 pT/Hz 1/2 under an external magnetic field of approximately 10000 nT.
Researchers developed a method for efficient atomic migration under ambient conditions using molecular monolayers and localized surface plasmons. The process regulates atom movement through dipole-dipole interactions, enabling scalable atomic patterning.
Researchers developed a multilayer device that decouples visible and infrared control, achieving record-wide thermal modulation for extreme environments. The breakthrough allows objects to adapt to various temperatures, making it suitable for applications like defense, stealth, smart windows, and wearable photonics.
Researchers have introduced a transparent, colorless, and unidirectional solar concentrator that can be directly coated onto standard window glass. The device achieves broadband polarization-selective diffraction and waveguiding without compromising clarity.
A new multi-objective optimization algorithm, PNSGA-II, was introduced to optimize the optical path length and ratio of optical path length to volume of multi-pass cells. The designed cell achieved an optical path length exceeding 80m, significantly improving LITES sensor sensitivity.
The study develops a novel polarimetric binocular three-dimensional imaging method leveraging multi-feature self-supervised learning. It simultaneously improves image clarity and enables accurate depth estimation in turbid water, a critical advancement for marine exploration.
A Shanghai Jiao Tong University research team developed non-volatile, programmable micro-ring transceivers using low-loss phase-change material Sb2Se3 on silicon MRR PN junctions. The breakthrough chip achieved 400 Gbps data rate with minimal impact on modulation and detection performance.
Researchers develop a novel application of lithium niobate film modulators for precise and high-speed control of light field phases. The proposed strategy enables tailoring of optical coherence in random light fields, paving the way for practical applications in optical imaging, encryption, and information transmission.
A deep learning-powered approach improves image quality without requiring clean reference images, suitable for high-speed biological dynamics. The TeD network selectively utilizes relevant spatiotemporal features to enable robust denoising even with fast-moving structures.
The AI-driven Distributed Acoustic Sensing (DAS) system leverages narrow-linewidth lasers for real-time acoustic wave propagation information. Data acquisition, preprocessing, and machine learning model construction are key stages in this technology.
The research team developed a solar-powered floating robot that combines photocatalysis, evaporation, and autonomous navigation for efficient water purification. The system uses a hybrid material and responds efficiently to sunlight across a broad spectrum.
A new photonic accelerator based on a nonlinear optoelectronic oscillator (NOEO) has been proposed to speed up reinforcement learning in AI. The accelerator outperformed existing methods, achieving high speeds and accuracy in solving complex problems like the multi-armed bandit problem and Tic Tac Toe game.
A team from Tsinghua University has shattered the paradigm of traditional waveguide QED by achieving a total decay rate below γ₀ through energy quantum confinement effect. This mechanism, operating in non-Markovian regime, dynamically traps energy quanta within the waveguide, converting energy loss into temporary storage.
Researchers propose a non-local metasurface that combines the generalized Kerker effect with non-local collective interactions to generate highly efficient and directional vortex beams. The design leverages intrinsic dipole singularities, non-local coupling, and Bragg scattering to achieve high-performance results.
Researchers have developed a metasurface differentiator that achieves broadband operation and high numerical aperture for image edge detection. The device outperforms existing metasurfaces with improved polarization independence, operating bandwidth, and resolution.
Researchers have proposed a color holographic display system with a customized achromatic liquid crystal grating that realizes a nine-time enlargement of viewing angle and a five-time increase in brightness. The system uses a composite loss function for high-brightness hologram encoding, enhancing energy utilization efficiency while ma...