Feng Gao

Accurate modeling and parameter estimation of distributed feedback (DFB) lasers are limited by the strong nonlinearity of laser dynamics and small-sample constraints. Based on the traveling-wave model (TWM), a hybrid physics-informed neural network (PINN)-TWM framework is proposed to improve data efficiency and enable accurate DFB laser response modeling. The framework maps intrinsic laser parameters to light-current (P-I) curves and small-signal frequency responses. For parameter estimation, a cascaded forward-inverse PINN architecture is developed, in which the trained physics-informed forward surrogate provides response-level guidance for recovering intrinsic physical parameters. Simulation results show that the forward model consistently outperforms the convolutional neural network (CNN) baseline across the evaluated device responses, reducing the MSE by more than 40% under the same limited training set of 500 samples. Meanwhile, the cascaded framework achieves the lowest parameter estimation error among the compared inverse methods. This study demonstrates the potential of combining physics-based modeling with data-driven learning for DFB laser response prediction and parameter estimation under small-sample conditions.

The failures of semiconductor lasers are often linked to the emergence or growth of defects. However, most research mainly focuses on the postmortem failure analysis caused by defects, lacking dynamic process analysis of defect expansion. This limitation hinders the understanding of defect growth patterns and expansion. In this work, we establish a macroscopic phenomenological model based on the dynamic characteristics of defects in semiconductor lasers. The expansion of defects is regarded as the diffusion transfer process of lattice strain, and the diffusion-limited aggregation (DLA) model is used to describe the aggregation process of random morphology of defects. The effect of model parameters on the growth pattern is studied, and the phenomenological relation between model parameters and actual defect features is established. The model successfully replicated experimentally observed morphologies in a distributed feedback (DFB) laser under high-temperature and high-current excitation. It not only predicts the intermediate process of defect expansion but also reveals the accelerated process. This research provides a novel approach to describing the defect evolution process in semiconductor lasers, contributing to a deeper understanding of defect expansion modes and characteristics within semiconductor lasers. It holds significant guiding implications for improving device reliability.

Parameters used to calculate the operating characteristics of distributed feedback (DFB) lasers have to be chosen properly to ensure a perfect match between the calculated and measured results. We propose a multi-objective parameter extraction method based on the deep neural network (DNN), that can not only efficiently and accurately extract the realistic parameters of DFB lasers but also resolve coupling issues among parameters. Two forward DNN models are trained using traveling wave model (TWM) and transfer matrix method (TMM), respectively, to predict the light-current (L-I) characteristics, amplitude modulation (AM) response, and spectrum of DFB lasers. After fabricating and measuring lasers, the extracted parameters are iteratively optimized against the experimentally measured results utilizing the well-performing forward models and differential evolution algorithm. Through Pareto sorting, an optimal parameter set is extracted, which can accurately reproduce all target characteristics (the LI curve, AM response, and spectrum) at the same time. The proposed method greatly reduces the computational complexity and avoids the time-consuming situation of traditional TWM and TMM calculations. Using the extracted parameters, the large-signal modulation is theoretically calculated. And it fits very well with the measured results, which verifies the practicality and reliability of the proposed method.

We propose and realize a compact and temperature-insensitive wavelength detector, integrated with a widely tunable external-cavity laser based on the standard silicon photonic platform. The fabricated wavelength detector is comprised of a hybrid Mach-Zehnder Interferometer (MZI) with silicon and silicon-nitride waveguides. The minimum temperature sensitivity of -0.28pm/K with the extinction ratio up to 28 dB is achieved, benefitting from the large difference of thermal-optic coefficients between silicon and silicon-nitride waveguides. Moreover, a widely tunable external-cavity laser is designed and fabricated, adopting the hybrid MZI wavelength detector as an on-chip wavelength locker. The wavelength tuning range from 1546.6 nm to 1647.2 nm is measured with the side-mode suppression ratio larger than 45 dB. The picometer-level wavelength monitoring accuracy is experimentally verified through the feedback of photocurrents from Ge/Si photodetectors, which validates the accuracy and robustness of the device.

This study proposes an efficient and accurate method for parameter extraction of quantum well distributed feedback (DFB) lasers by combining the rate equation model, finite element method, transmission matrix method, and traveling wave model (TWM). By fabricating and measuring the companion Fabry-Perot (FP) lasers, material and structural parameters common with the target DFB laser are extracted efficiently. All the intrinsic parameters of the DFB laser are accurately extracted by integrating multiple mathematical models, and the possibility of multiple solutions is avoided. From the extracted parameters, the output characteristics of the DFB laser are simulated using the TWM. The simulation results agree closely with the experimental results, proving the feasibility and accuracy of the proposed method.

High-speed directly modulated distributed feedback (DFB) laser is crucial for high-speed optical communication systems. The modulation bandwidth and frequency chirp of such devices are primarily affected by the differential gain and linewidth enhancement factor (LEF) of the quantum wells (QWs) in the active region. In this work, the optical gain properties of long wavelength AlGaInAs-InP strained quantum wells (QWs) with different crystal orientations are numerically investigated using the multi-bands k.p theory, which considers both valence-band anisotropic and nonparabolicity. Compared to the QW laser grown on a conventional (001) substrate, a higher differential gain, and smaller LEF are observed for the laser grown on the (110) substrate with the (110) plane as the mirror facet. This is primarily due to the reduced effective mass of the valence band. The dynamic characteristics of (001)-and (110)-oriented DFB laser is theoretically studied using the one-dimensional traveling wave model (1D TWM) at 25°C and 95°C. The simulation results show that the lower chirp and wider modulation bandwidth can be achieved for the QW laser grown on (110) substrate.

We demonstrate a III-V/Si widely wavelength tunable laser covering the entire L-band for the optical communication systems. By carefully designing the silicon ring filter, low threshold current and high output power are expected. Using the standard silicon photonic process, the silicon filter chip is fabricated and compactly packaged with a reflective semiconductor optical amplifier. The low threshold current of 20 mA is achieved at 15 °C, and over 76 mW output power is obtained at 320 mA. The wavelength tuning range from 1565 nm to 1635 nm is realized with the side-mode suppression ratio larger than 50 dB. Furthermore, the intrinsic linewidth narrower than 25 kHz and relative intensity noise below −152 dB/Hz is achieved, which can support the coherent communications.

An artificial intelligent neural network improved particle swarm optimization algorithm is proposed for the inverse design of semiconductor optical amplifier. Seven input parameters, current-gain curve and saturation output power curve are selected to form the data set based on the physical model of semiconductor optical amplifier. The effectiveness of forecasting performance is improved by contrasting two back propagation neural network techniques (Scaled Conjugate Gradient and Levenberg-Marquardt) and operational settings (Central Processing Unit and Graphics Processing Unit). Higher accuracy is achieved through feedback analysis of neuron number optimization and test error. The addition of a unique backpropagation neural network can make the fitness of particle swarm algorithm mostly converge below $2\times 10^{-4}$ . The relative difference between original performances and inverse predictions is close to 0%, which proves the effectiveness of parameter extraction. This method can take advantage of neural networks to improve accuracy and speed of particle swarm optimization algorithms for efficient semiconductor optical amplifier inverse design and multi-solution analysis.

We propose a simple and cost-effective method, using a direct frequency modulation (FM) and noncoherent detection (NCD) scheme, to suppress the nonlinear optical effects in dense wavelength division multiplexed (DWDM) optical communication. The FM transmitter comprises a directly modulated distributed feedback laser and a saturable semiconductor optical amplifier. In the NCD receiver, an optical slope filter as the FM to intensity modulation (IM) signal convertor is placed before a conventional photodetector. Because the FM signal has more evenly distributed optical power, bit-pattern-dependent nonlinear effects are consequently suppressed. After analyzing the nonlinear effects in the FM-NCD system and traditional IM direct detection (IM-DD) system, we found that the minimum achievable BER of the proposed FM-NCD scheme is 40 dB smaller. Moreover, a 2 Tbps (10 Gb/s × 200 channels) capacity was achieved by the FM-NCD system in 100 km DWDM passive optical networks (PONs), which is twice the capacity of the IM-DD system (10 Gb/s × 100 channels) under the same condition. These results indicate that WDM-PONs with the cost-effective FM-NCD scheme are strong candidates for future broad access networks and show great potential for the combination of optical access and metro networks for future generations of PONs.