High-power femtosecond lasers have demonstrated remarkable potential for a wide range of atmospheric applications. Extending these capabilities to the ultra-high-power regime in air is highly desirable; however, severe spatiotemporal distortions induced during propagation impose significant limitations on both pulse generation and control. Here, we demonstrate that ultra-high-power laser pulses with peak powers above 20 TW and intensities exceeding 1 TW/cm 2 can be delivered into air using a helium-filled compressor. Operating at atmospheric pressure, the helium-filled compressor effectively suppresses nonlinear phase accumulation prior to beam injection into air, enabling an output beam with no observable spatiotemporal distortion relative to a low-power reference and providing sufficient space for beam control after compression. This approach establishes a practical platform for high-power, high-intensity laser operation in ambient air and opens new opportunities for atmospheric strong-field physics and more accessible extreme-light sources.
We establish the theoretical framework and investigate the fundamental limits for an integrated optical chip design containing a rotation sensing ring resonator coupled to an inverse weak value amplified Sagnac interferometer that amplifies the signal with phase information. We achieve, for conservative parameter choices, a minimum detectable angular rotation rate of ∼0.1 °/hr and an Allan deviation of∼0.08 °/hr assuming shot noise limited performance. For matched detected optical power, our design can improve the signal-to-noise ratio, reduce the minimum detectable angular rotation rate, and Allan deviation by more than ten times compared to an ideal standard Sagnac interferometer with balanced detection. We also show the robustness of our design against backscattering and multimode directional coupler imperfections.
In this paper, we present a reliability risk of using a virtual guard ring (VGR) in single-photon avalanche diodes (SPADs) and propose an optimized SPAD to mitigate this risk and consequently improve the device performance. Utilizing a lower-doped deep junction to implement a SPAD is an appropriate approach to prevent band-to-band and trap-assisted tunneling and improve the near-infrared efficiency in advanced CMOS technology. To realize the deep-junction-based SPAD, a VGR based on the retrograde doping profile is typically applied to prevent premature edge breakdown. However, because the VGR does not include a physical guard-ring structure at the side of the junction, carriers generated close to the cathode region can be more susceptible to triggering avalanche events in the VGR region, thereby degrading device stability and reliability. To investigate this issue, we fabricated two different VGR-based SPADs based on TCAD simulations and performed comparative analysis in terms of I-V characteristics, dark count rates, waveforms, and lifetimes. The default SPAD operates for only about 1 hour or less at the excess bias voltage of 2 V under dark and illuminated conditions, whereas the optimized SPAD operates stably for over 24 hours at the excess bias voltage of 5 V regardless of illumination. These results clearly reveal the previously underexplored reliability limitation of VGR-based SPADs and demonstrate that cathode-region optimization is an effective strategy for robust and high-performance deep-junction SPADs and high-density SPAD arrays.
Soliton-crystal microcombs host strongly squeezed vacuum in their below-threshold modes, but a critically coupled resonator releases at most 3 dB of it. We present a complete, openly reproducible design pipeline from material data to detected quantum noise. Full-vector finite-element modeling of a 4H-silicon-carbide-on-insulator waveguide yields the resonator dispersion, a Lugiato–Lefever solver provides the two-soliton crystal, and a linearized Heisenberg–Langevin model gives the multimode squeezing at the output port. An auxiliary ring with twice the free spectral range Purcell-extracts the odd-mode quadrature lattice, optimally near ten cavity linewidths at a fixed 8.3 mW pump. An explicit two-ring model with the computed dispersion of both rings predicts 7.9 dB of detectable squeezing over 1.81 GHz, with two dominant supermodes and an entangled odd-mode lattice. Third-order dispersion shifts the crystal repetition rate measurably yet leaves the squeezing unchanged until it destroys the crystal. All code, data, and scripts are released openly.
In the non-paraxial regime, the electric field must be treated as a three-dimensional vector, with all components contributing to its three-dimensional degree of polarization. Such 3D behavior arises naturally in nanoscale light-matter interactions and can also be engineered through tightly focusing structured vectorial light fields. In both cases, understanding the relationship between the focal non-paraxial field properties and the corresponding paraxial back-focal-plane distributions is essential for advancing fundamental studies and enabling applications in nanoscale imaging and sensing. Here, we formulate this problem in a cylindrical basis and investigate the focal (3 × 3) coherence matrix as well as the 3D degree of polarization as functions of the input correlations between the radial and azimuthal components of generalized cylindrical vector beams (GCVBs), numerical aperture, and propagation distance.
Rydberg-atom electrometry based on electromagnetically induced transparency (EIT) offers an optical approach for electric-field readout, but reliable readout of weak transient responses remains challenging under low signal-to-noise ratio (SNR) conditions. Here, we demonstrate machine-learning-enhanced readout of partial-discharge-like (PD-like) transient responses in a Cs 73 D 5/2 Rydberg-EIT vapor-cell system. Matched filtering (MF) and normalized cross-correlation (NCC) are used for candidate generation, followed by principal component analysis (PCA)/logistic-regression-based waveform-morphology discrimination and a cross-cycle temporal-consistency refinement. Across experimentally acquired low-SNR cycles, the PCA/logistic morphology selector achieved a strict localization accuracy of 63.25%, which increased to 86.32% after incorporating cross-cycle temporal consistency. The results indicate that weak Rydberg-EIT transient responses can retain discriminative local waveform morphology even when conventional template-peak ranking becomes unreliable. Overall, these results demonstrate that morphology-based machine-learning discrimination, complemented by temporal-consistency refinement, can improve the reliability of weak Rydberg-EIT transient readout under low-SNR conditions.
We propose and experimentally demonstrate a multifunctional liquid prism based on a hydrophilic film. By introducing a hydrophilic sidewall into the proposed liquid prism, an initially tilted liquid-liquid (L-L) interface is passively established through asymmetric surface wettability. Under electrowetting actuation, the L-L interface is continuously steered by applying 40-80 V to the hydrophobic sidewalls, resulting in a maximum interface tilt angle of 25° and a corresponding optical beam deflection of 1.9°. In addition, selective actuation of a single hydrophobic sidewall at voltages above 80 V transforms the device into a cylindrical lens, enabling multifunction operation within the same platform. The proposed device features a compact structure, straightforward electrical control, and multifunctional optical performance, making it well-suited for applications in optical image stabilization and light-sheet microscopy.
Light-field microscopy (LFM) captures volumetric fluorescence in a single exposure but is constrained by the spatial–angular sampling trade-off of a detector with a limited pixel count, leading to two major reconstruction limitations: z-ambiguity, which results in low and depth-dependent axial resolution, and severe reconstruction degradation near the nominal focal plane (NFP). Here, we present modulated light-sheet light-field microscopy (MoLiS-LFM), which acquires multiple z-sparse light-field images per volume using mutually independent, z-modulated light-sheet excitation. This approach effectively guarantees sparsity in the reconstructed volume and thereby improves axial resolving and localization capabilities. In addition, piezo-driven objective displacement shifts the NFP relative to the stationary specimen for a complementary MoLiS-LFM acquisition, leveraging LFM’s high-fidelity off-focus reconstruction to reduce near-NFP degradation. Fluorescent-bead experiments demonstrate improved axial resolving and localization capabilities, and whole-body C. elegans muscle imaging shows the recovery of structures near the NFP. To demonstrate the feasibility of MoLiS-LFM for in vivo imaging, we applied it to whole-brain imaging in larval zebrafish and achieved cellular-scale volumetric imaging over a 550 × 550 × 200~ μ m 3 volume at 9.06 volumes per second.
For lightweight and compact star-sensor payloads, reducing the complexity of the optical front end helps satisfy size and mass constraints, but may also result in insufficient off-axis aberration correction, spatially variant star-spot morphology, and unstable segmentation of overlapping stars. This study proposes an extraction and localization method for spatially variant complex star spots in lightweight star sensors. First, a radial point spread function (PSF) sampling and rotational-extension strategy is adopted to construct a compact full-field PSF representation. A photometry-dominated hybrid topographic watershed is then introduced to improve the separation of overlapping and non-convex star spots. Normalized cross-correlation (NCC)-based matching and residual-error map compensation are further combined to achieve subpixel star-spot localization. Simulation results show that the radial-rotation PSF model reconstructs full-field PSFs with a mean NCC of 0.9861. In the edge-field overlap-transition range of 4.5–8 px, compared with the conventional distance-field watershed, the proposed hybrid topographic watershed increases the two-star topology recovery rate from 0.400 to 0.630 and reduces the corresponding pair error by approximately 21.3%. Under representative and high-signal static conditions, the valid-measurement RMSE values (RMSE valid ) of the proposed method are 0.2896 px and 0.2115 px, respectively. In dynamic simulations, the RMSE valid values are 0.2639 px and 0.2794 px for rotations about the X and Y axes, respectively, and 0.4154 px under roll motion. Digital micromirror device (DMD)-based optical-bench experiments further demonstrate the applicability of the method to measured PSF reconstruction, overlapping-star segmentation, and relative localization, with an average relative localization RMSE of approximately 0.256 px.
Large-field-of-regard (FOR) aspherical window infrared systems exhibit strong dynamic aberrations and spatially nonuniform degradation, and stepwise design cannot jointly optimize image quality and manufacturability. We present an end-to-end optical–digital joint optimization framework for large-FOR aspherical window infrared imaging that unifies differentiable fast ray tracing, differentiable PSF-based degradation modeling, and neural restoration in a single co-design loop. Under angle-dependent aberrations and spatially variant blur induced by the aspherical window across a wide field of regard, optical parameters and the restoration network are co-optimized to jointly improve image quality and manufacturability of the minimalist optical system. To improve efficiency and accuracy, we accelerate ray–surface intersection and frequency-domain PSF computation and impose physical and fabrication constraints. Experimental results demonstrate that the proposed method provides higher ray-tracing accuracy and superior image restoration performance compared with competing methods. Joint optimization reduces system length by 19.5% and lens count from three to one, while improving corrector manufacturability.
When tip-tilt and piston perturbations occur simultaneously in laser arrays, learning to estimate them from single intensity images becomes substantially more challenging. Conventional network training on purely random samples is poorly suited for this task. In this study, we propose a joint variable-density bin sampling strategy to generate training samples with rich diversity, yielding remarkable gains in establishing a reliable mapping between far-field intensity patterns and the corresponding tip-tilt and piston aberrations. Additionally, we integrate a squeeze-and-excitation (SE) channel-attention module into the network to further boost prediction accuracy. Simulation results demonstrate that our method can achieve stable joint correction of tip-tilt and piston aberrations across 37- and 61-channel beam arrays under dynamic noise disturbances, with the normalized power-in-the-bucket (PIB norm ) measured at approximately 0.97 and 0.93, respectively.
