Accurate waveform-level modeling of quantum-well semiconductor optical amplifiers (QW-SOAs) is important for the design and performance evaluation of optical amplification, optical signal processing, and high-capacity communication systems. Traveling-wave (TW) models are widely regarded as high-fidelity tools for QW-SOA simulation because they resolve the distributed interaction between the optical field and carrier densities along the device. However, this accuracy comes at a high computational cost, particularly for long-sequence waveform simulations. Lumped QW-SOA models provide a faster alternative by removing the spatial dimension, yet their accuracy degrades in nonlinear regimes where gain saturation and carrier induced phase dynamics become significant. This paper proposes a physics-enhanced residual learning (PERL) model that trains a neural network to learn the residual gain and phase discrepancies between a lumped model and a high-fidelity TW model, rather than the complete time-domain waveform response of the QW-SOA. By exploiting the physics-based prior information embedded in the lumped model, the proposed approach reduces the learning complexity and enables a lightweight PERL network to achieve waveform-level accuracy comparable to the TW model while retaining the computational efficiency of the lumped model. For 16-ary quadrature amplitude modulation signals, the proposed PERL model reduces the mean symbol-level normalized standard deviation (NSD) from 0.20 to 0.0095. Runtime comparisons further show a more than 20-fold speedup over TW simulation at comparable accuracy.
Despite the growing demand for reliable long-range underwater acoustic communication in complex marine environments, the low signal-to-noise ratio (SNR) of acoustic sensors remains a major bottleneck to extending communication range. A wedge-shaped gradient-index metamaterial (WGIM) is integrated with distributed-feedback (DFB) fiber-laser hydrophones to construct a multi-band underwater acoustic receiver. The WGIM provides passive acoustic-field compression, frequency-selective pressure enhancement, and spatial separation of multiple sub-bands, while the DFB fiber-laser hydrophones enable high-sensitivity detection at the corresponding pressure maxima. Simulations and tank measurements show in-band enhancement gains of 13.0–15.1 dB over three 500 Hz bands centered at 20.21, 23.00, and 24.83 kHz, respectively. In a coastal sea trial, the WGIM-DFB composite receiver supports Quadrature Phase Shift Keying (QPSK) communication over approximately 10 km. The three sub-bands each achieve a bit rate of 0.5 kbit/s, giving an aggregate bit rate of 1.5 kbit/s, with measured bit-error rates of 4.16×10−4, 8.33×10−4, and 8.33×10−4, respectively. These results demonstrate that passive metamaterial assisted acoustic-field enhancement combined with high sensitivity fiber-optic hydrophones can improve the receive-side link margin and reliability of long-range multi-band underwater acoustic communication.
Silicon waveguide distributed Bragg reflectors (DBRs) were designed and fabricated using a 248 nm Deep Ultra-Violet (UV) projection lithography process within a photonic foundry platform. The influence of key structural parameters, including grating period, fill factor, number of periods, and corrugation width, on the spectral response of the device was systematically investigated. The measured Bragg wavelength, extinction ratio, and 3 dB bandwidth were statistically evaluated across ten DBR chips fabricated within a multi-project wafer (MPW) run. For operation near 1550 nm, a corrugation width of 250 nm with a 0.60 μm grating period, 0.5 fill factor, and 100 grating periods, provides a mean Bragg wavelength of 1551.35 nm, with yields of 90%, 70%, and 60% within ±5, ±3, and ±2 nm, respectively. The corresponding extinction ratio yields are 100%, 90%, and 90% within ±5, ±3, and ±2 dB, while the 3 dB bandwidth achieves 100% yield for all three tolerances. This parameter combination results in a compact grating length of only 60.30 μm, demonstrating that near 1550 nm operation can be achieved with a short DBR while maintaining high fabrication yield across the investigated performance metrics. These results demonstrate that the spectral characteristics of silicon DBRs can be effectively engineered within the design rules of a 248 nm Deep UV fabrication platform, providing practical guidelines for implementing wavelength selective components in photonic integrated circuits (PICs), particularly within MPW based silicon photonics platforms.
The choice of initial structure greatly influences the design and optimization of single-photon LiDAR and imaging composite optical systems. However, current approaches do not accommodate the need for focal length distribution in multi-path systems, nor facilitate efficient initial structure searches through aberration correction. Herein, a high-quality initial structure construction and optimized design method based on engineering-applicable optical configurations and excellent axial image quality is proposed. This method derives the optical focal distribution and paraxial aberration models for the reflecting ray path and the catadioptric ray path. By using a basic genetic optimization strategy of merely 22 ultra-short iterations, a high-quality initial structure with excellent image quality can be quickly obtained, without the need for complex algorithms or extensive iterations. The coaxial two-mirror system is extended to a coaxial three-mirror system to verify the scalability of structure construction method. Through a multi-objective genetic algorithm, a Pareto front solution set is generated to achieve the optimal trade-off between image quality and structural constraints. Based on the initial structure after search and optimization, a multi-path composite optical system with a common aperture for visible light, infrared light, and single-photon LiDAR is designed and analyzed. This work can provide an efficient and reliable design starting point for single-photon LiDAR and imaging composite optical systems, and further promote the transformation of the optical design patterns from relying on computational power for unpredictable optimization to targeted scientific optimization guided by optical configurations and aberration models.
We demonstrate a wavelength-tunable laser system based on a hybrid mode-locked thulium-doped fiber laser, delivering outputs at 1.9 μm and 970 nm. The seed oscillator employs a hybrid mode-locking scheme based on a semiconductor saturable absorber mirror and a nonlinear amplifying loop mirror, enabling stable pulse generation with continuous wavelength tuning from 1896 to 1969 nm at a repetition rate of 15.7 MHz and a signal-to-noise ratio exceeding 67 dB. Following cascaded all-fiber amplification, second-harmonic generation in a 5-mm BBO crystal produces tunable output from 962 to 974 nm with a maximum average power of 36.27 mW and pulse durations of 89-146 ps. Both spectral bands exhibited relative RMS power fluctuations below 1% over a 2-h measurement period. The proposed architecture provides a stable solution for applications requiring synchronized, wavelength-tunable excitation in the short-wave infrared and near-infrared bands.
