在QFN芯片封装缺陷检测中，增加图像分割环节可有效提高缺陷检测准确性与检测效率。针对图像分割中传统算法效率低、智能优化算法分割精度低稳定性差的问题，本文提出一种基于改进灰狼优化算法（IGWO）的图像多阈值分割方法。首先，改进原始灰狼优化算法非线性因子，平衡算法搜索效率与挖掘能力；其次，引入反向学习策略提高种群整体质量，引入正弦函数、调整头狼权重以改进灰狼更新策略，增强算法多样性与挖掘能力；然后，提出头狼靠拢与种群变异交替进行的位置更新策略，平衡算法收敛性能与跳出局部最优能力；最后，以Kapur熵为适应度函数，求解最优分割阈值。将本文提出的改进灰狼优化算法的多阈值图像分割方法，与灰狼优化算法（GWO）、基于翻筋斗觅食策略的灰狼优化算法（DSF-GWO）、基于莱维飞行的樽海鞘群优化算法（LSSA）、改进北方苍鹰算法（INGO）的图像分割方法进行实验对比，结果表明：本文方法在分割用时方面，约为DSF-GWO的1/2，INGO的1/4；在分割精度与稳定性方面，在进行QFN芯片缺陷图像的连续30次分割时，本文方法具有最大Kapur熵平均值、最小标准差与最短分割时间。因此本文方法可实现高精度、高稳定性与高效率的QFN芯片图像多阈值分割。

现如今处于第五代移动通信技术（5G）网络发展建设的重要阶段，前传网建设中带宽需求大、光模块成本高和铺设难度大等问题亟待解决。已知四通道小型可插拔（QSFP）封装具有4个通道，而传统的QSFP28 50 Gbit/s光模块设计方案只使用了其中的两个通道。文章选用比QSFP封装更小的小型可插拔（SFP）封装制成50 Gbit/s光模块，底面积减小了40.5%，体积减小了52.2%，功耗减少了42.1%，传输相同距离光模块成本减少了73.4%。使用4阶脉冲幅度调制（PAM4）技术实现25 Gbit/s光器件传输50 Gbit/s信号，达成了单波长速率翻倍。通过对比测试，常温发射机色散眼图闭合四相（TDECQ）减小了20.5%，灵敏度提高了0.6 dB，且其余各项参数均满足电气与电子工程师协会（IEEE）802.3cd中50 GE以太网的相关要求，能够稳定进行40 km信息传输。

A time-resolved high-power laser photometer, which measures the real-time variations of transmission, internal reflection, and scattering simultaneously with picosecond time resolution, was developed to investigate the material response sequence during high-power nanosecond laser irradiation in thick fused silica. It was found that the transient transmission decreased sharply, accompanied by an increase in internal reflection at the rising edge of the laser pulse. The transient transmission recovered, while laser damage did not occur, but it did not recover if the scattering increased, indicating the occurrence of laser damage. The reason for the sharp decrease of transmission and the relationship between the transmission drop and laser damage were discussed.

The nonlinear absorption (NLA) properties of potassium dideuterium phosphate crystals at 515 nm under different excitation laser intensities are investigated with the Z-scan technique. Two critical intensities are highlighted: the critical intensity for exciting the NLA and the critical intensity of the multiphoton absorption mechanism transition. Experimental results indicate the existence of defect states located in the band gap, which can be manipulated by varying laser intensity. A model based on the change of multiphoton absorption mechanism induced by the transformation of defect species is proposed to interpret the experiments. Modeling results are in good agreement with the experiment data.

To reveal the physical mechanism of laser ablation and establish the prediction model for figuring the surface of fused silica, a multi-physical transient numerical model coupled with heat transfer and fluid flow was developed under pulsed CO2 laser irradiation. The model employed various heat transfer and hydrodynamic boundary and thermomechanical properties for assisting the understanding of the contributions of Marangoni convention, gravitational force, vaporization recoil pressure, and capillary force in the process of laser ablation and better prediction of laser processing. Simulation results indicated that the vaporization recoil pressure dominated the formation of the final ablation profile. The ablation depth increased exponentially with pulse duration and linearly with laser energy after homogenous evaporation. The model was validated by experimental data of pulse CO2 laser ablation of fused silica. To further investigate laser beam figuring, local ablation by varying the overlap rate and laser energy was conducted, achieving down to 4 nm homogenous ablation depth.