View by Category
Multi-task large-scale integrated optical vision processor using ultra-fast parallel nanofabrication
Published
, Published online: 02 July 2026
, doi: 10.37188/lam.2026.096
Optical neural networks (ONNs) promise ultra-fast low-power machine vision; however, visible-wavelength implementations are constrained by limited neuron density and accuracy. Although random projections provide efficient untrained feature encoding, we advance ONN performance using a high-throughput randomised multi-focus two-photon lithography (TPL) platform that fabricates millions of 500 nm neurons at the millimetre scale within 15 min. The resulting platform achieves ≥97% classification accuracy in multiple image classification and keypoint detection tasks using minimal digital parameters that outperform other devices of comparable neuron densities while enabling compact integration with camera systems through its transparent design. Our results indicate that ONNs can serve as scalable and practical solutions for high-performance multi-task machine vision.
Published
, Published online: 12 May 2026
, doi: 10.37188/lam.2026.049
As semiconductor manufacturing advances towards finer feature sizes, mask optimization (MO) has become increasingly critical in optical lithography to ensure pattern fidelity. In extreme ultraviolet (EUV) lithography, full-chip MO encounters significant challenges in terms of computational accuracy and efficiency, which are exacerbated when employing curvilinear patterns. Herein, we propose a full-chip curvilinear MO framework for EUV lithography that integrates deep-learning-enabled forward modelling with gradient-based inverse optimization. We represent the forward model using a tuneable U-net trained on data generated by an accurate and efficient modified Born series method. This model achieves a significantly lower complexity by describing the 3D mask effect through amplitude and phase perturbations. For inverse optimization, gradients are calculated via the adjoint method using slices of the 3D mask field as input—a significantly more efficient approach than utilising the entire 3D field. Evaluated under typical scenarios, the proposed framework demonstrates a four-order-of-magnitude speedup compared with MO based on the finite-difference time-domain method without compromising accuracy. Leveraging this framework, the MO for a 1 mm2 wafer area with 19.41 nm critical dimensions can be completed in 31.7 h using 1,000 GPUs, highlighting its potential for full-chip EUV curvilinear mask optimization.
Published
, Published online: 06 May 2026
, doi: 10.37188/lam.2026.052
The weak scattering and overwhelming background of periodic structures fundamentally hinder the inspection of subwavelength defects embedded in dense nanopatterns. Herein, we introduce an actively tunable photothermal modulation scheme that leverages the temperature-dependent resonance shifts of silicon nanostructures to engineer their far-field scattering signatures. Localised optical heating induces a redshift in the underlying resonances, producing a strongly nonlinear change in both the defect and background scattering. This modification amplifies defect-induced perturbations and suppresses background contributions, substantially enhancing the inspection sensitivity for deep-subwavelength defects. A coupled optical-thermal model quantitatively captures the temperature rise and transient thermal evolution and predicts the resonance modulation achievable under the given pump conditions. This study establishes reversible, non-destructive photothermal resonance modulation as a general mechanism for dynamically engineering optical contrast in patterned media, offering a pathway towards high-sensitivity wafer inspection and tunable nanophotonic sensing.
Published
, Published online: 06 May 2026
, doi: 10.37188/lam.2026.036
Direct-write multi-photon laser lithography (MPL) combines highest resolution on the nanoscale with essentially unlimited 3D design freedom. The groundbreaking potential of this technique has been demonstrated in various application fields, including micromechanics, material sciences, microfluidics, life sciences, as well as photonics, where in-situ printed optical coupling elements offer new perspectives for package-level system integration. However, millimeter-wave (mmW) and terahertz (THz) devices did not yet leverage the unique strengths of MPL, even though the underlying devices and structures could also greatly benefit from 3D freeform microfabrication. A key challenge is that functional mmW and THz structures require materials with high electrical conductivity and low dielectric losses, which are not amenable to structuring by multi-photon polymerization. In this work, we introduce and experimentally demonstrate a novel approach that leverages MPL for fabricating high-performance mmW and THz structures with hitherto unachieved functionalities. Our concept exploits in-situ printed polymer templates that are selectively coated through highly directive metal deposition techniques in combination with precisely aligned 3D-printed shadowing structures. The resulting metal-coated freeform structures (MCFS) offer high surface quality, low dielectric losses, and conductivities comparable to bulk material values, while lending themselves to in-situ fabrication on planar mmW and THz circuits. We experimentally show the viability of our concept by demonstrating functional THz structures such as ultra-broadband chip-chip interconnects, THz probe tips, and suspended THz antennas. We believe that our approach offers disruptive potential in the field of mmW and THz technology and may unlock an entirely new application field for laser-based 3D manufacturing.
Published
, Published online: 11 May 2026
, doi: 10.37188/lam.2026.060
The exceptional mechanical and thermal properties of silicon carbide (SiC) make it vital for advanced optics; however, its hardness and brittleness cause subsurface defects (SSDs) during machining that impair performance and longevity. Current detection methods remain destructive and inefficient, whereas conventional optical coherence tomography (OCT) struggles with limited penetration, surface scattering interference, and poor defect contrast in this highly scattering material. We propose a non-destructive off-axis bright- and dark-field synchronous OCT (BADF-OCT) method that captures complementary scattered signals at dual angles to enhance weak subsurface feature detection. The broadband 1100–1500 nm near-infrared spectral-domain OCT system provides high axial resolution with adequate SiC penetration. Experimental validation on reaction-bonded SiC demonstrates clear discrimination between surface fracture and subsurface crack layers, providing reliable detection of micrometre-scale defects at depths up to ~200 μm. Three-dimensional volumetric imaging combined with bright/dark-field data fusion effectively distinguishes true SSDs from surface contaminants, significantly improving the recognition accuracy. This study is expected to contribute to the development of high-energy lasers, large-scale scientific facilities for light sources, and advanced optical manufacturing.
Published
, Published online: 29 April 2026
, doi: 10.37188/lam.2026.053
Characterizing the optical homogeneity of side-polished cylindrical transparent materials remains challenging. To address this challenge, a four-step absolute measurement method based on a Fizeau interferometer is proposed for cylindrical transparent materials. The refractive index distribution is derived from wavefront data obtained through four sequential measurements: empty-cavity interference, transmission interference, front-surface interference, and back-surface interference. A homogeneity error of 1 × 10−5 was introduced in MATLAB simulations, yielding a result of 9.9999 × 10−6 with a residual error of 8.0319 × 10−11, confirming the method’s validity. Two repeated measurements performed at different times yielded homogeneity values of hom1 = 9.5802 × 10−6 and hom2 = 9.3331 × 10−6 (2.7% deviation), demonstrating good robustness. The uncertainties of the two measurements were 1.0997 × 10−6 and 0.8767 × 10−6, respectively, and the expanded uncertainties were 2.1994 × 10−6 and 1.7534 × 10−6, respectively. This method effectively isolates surface errors from material homogeneity, providing a practical approach for the accurate characterization of cylindrical optical components.
Published
, Published online: 29 April 2026
, doi: 10.37188/lam.2026.033
Precise high-speed motion tracking and velocimetry are critical underlying technologies in various areas such as advanced manufacturing, robotics, and modern physics. Mature detection methods such as Doppler velocimetry and dual-comb interferometry measurements cannot achieve directional unambiguity detection without a trade-off between speed and precision due to the inherent limitations of their system mechanisms. In this study, we propose an innovative high-speed and precise motion detection method based on chirped power oscillation waves (CPOW) generated by dispersion-controlled dual-swept lasers. Our results demonstrate that the displacement and velocity of the target can be directly identified via the power oscillation of the zero-frequency point on the interference signals after carefully controlling for the difference in group delay dispersion between the two beams of the swept lasers. We achieved sub-micrometer-scale displacement measurement accuracy with an update frame rate on the order of MHz and a relative velocity measurement error of better than 0.1%. Furthermore, the proposed method can also be used to reveal the unpredictable operational states of motion-control equipment influenced by mechanical vibrations. This dynamic displacement measurement and velocimetry method based on CPOW opens the door to advanced, fast, and high-precision ranging systems.
Published
, Published online: 28 April 2026
, doi: 10.37188/lam.2026.050
Multi-core fiber (MCF) imaging is essential for minimally invasive endoscopy in medicine and industrial inspection. However, the bulky distal optics increase the diameter and invasiveness, causing tissue damage. Its applications are further constrained by low spatial resolution and prominent honeycomb artifacts. We present a lensless MCF imaging approach based on Spectral-Guided Artifact Removal Network (SGARNet). In this framework, a physics-informed prior is embedded in a lightweight SpectralGate module to suppress lattice-frequency artifacts in the feature domain. The experimental results show a 12.12 dB improvement in PSNR and 0.4064 increase in SSIM, indicating superior performance over previous methods. The robustness and generalizability are confirmed by successful reconstructions across diverse textural complexities and biological tissue samples. These results demonstrate potential for practical deployment in compact and safer biomedical endoscopes.
Published
, Published online: 22 April 2026
, doi: 10.37188/lam.2026.045
Overcoming the trade-off between a wide field of view (FOV) and compactness remains a central challenge for integrating near-infrared (NIR) imaging into smartphones and AR glasses. Existing refractive NIR optics cannot simultaneously support ultrawide angles above 100° and ultrathin total track lengths (TTL) below 5 mm, fundamentally limiting their integration into portable devices. Herein, we present a wafer-level-manufactured meta-aspheric lens (MAL) that simultaneously achieves a 101.5° FOV, 3.39 mm TTL, and F/1.64 aperture within a compact volume of 0.02 cm3. Unlike previous hybrid systems that rely on separate refractive and diffractive components, the proposed MAL introduces a fully integrated architecture that provides a compact form factor. This integration also simplifies fabrication by enabling high-throughput production via micrometre-level precision alignment and bonding on a single wafer, which requires only one dicing step and no additional mechanical fixtures. Furthermore, the design process incorporates manufacturability and enables metalens dispersion modelling, ensuring that the experimental performance matches simulation results. We validated the MAL method using both direct and computational imaging experiments. Despite its small form factor, our scalable MAL demonstrated strong NIR imaging performance in eye tracking, blood vessel imaging, and computational pixel super-resolution tasks. This scalable MAL technology establishes a new benchmark for high-performance miniaturised NIR imaging, and opens the door for next-generation smartphones and AR optical systems.
Published
, Published online: 23 April 2026
, doi: 10.37188/lam.2026.027
Surface-enhanced Raman scattering (SERS) is widely used for trace detection and compositional analysis of biochemical samples. Constructing multidimensionally ordered hotspots with high densities and intensities is crucial for achieving superior SERS substrate performance. Here, we propose a multilevel SERS substrate based on curvature and structural optimisation strategies. We fabricated microlenses with various curvatures via modification and etching using a temporally-shaped femtosecond laser. These lenses were decorated with wrinkles and Ag nanoparticles (AgNPs) via sequential pre-strain application and chemical deposition. Experimental and simulation results demonstrated that the coupling of the wide-field electric field induced by the microlens with the localised plasmonic hot spots on the AgNPs and wrinkles enhanced the localised surface electric field. Curvature-optimised microlenses can increase the wide-field electric fields. The fabricated SERS substrates achieved a low minimum detection limit of 10−11 M and an enhancement factor of approximately 1.22 × 107. These substrates can be employed to detect thiram fungicide on crops using two different methods (in situ detection and solution-assisted detection), demonstrating potential for operating efficiently under different usage conditions.
- First
- Prev
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- Next
- Last
- Total:22
- To
- Go
Email
RSS