Advanced electronic materials are the fundamental building blocks of integrated circuits (ICs). The microscale properties of electronic materials (e.g., crystal structures, defects, and chemical properties) can have a considerable impact on the performance of ICs. Comprehensive characterization and analysis of the material in real time with high-spatial resolution are indispensable. In situ transmission electron microscope (TEM) with atomic resolution and external field can be applied as a physical simulation platform to study the evolution of electronic material in working conditions. The high-speed camera of the in situ TEM generates a high frame rate video, resulting in a large dataset that is beyond the data processing ability of researchers using the traditional method. To overcome this challenge, many works on automated TEM analysis by using machine-learning algorithm have been proposed. In this review, we introduce the technical evolution of TEM data acquisition, including analysis, and we summarize the application of machine learning to TEM data analysis in the aspects of morphology, defect, structure, and spectra. Some of the challenges of automated TEM analysis are given in the conclusion.Advanced electronic materials are the fundamental building blocks of integrated circuits (ICs). The microscale properties of electronic materials (e.g., crystal structures, defects, and chemical properties) can have a considerable impact on the performance of ICs. Comprehensive characterization and analysis of the material in real time with high-spatial resolution are indispensable. In situ transmission electron microscope (TEM) with atomic resolution and external field can be applied as a physical simulation platform to study the evolution of electronic material in working conditions. The high-speed camera of the in situ TEM generates a high frame rate video, resulting in a large dataset that is beyond the data processing ability of researchers using the traditional method. To overcome this challenge, many works on automated TEM analysis by using machine-learning algorithm have been proposed. In this review, we introduce the technical evolution of TEM data acquisition, including analysis, and we summarize the application of machine learning to TEM data analysis in the aspects of morphology, defect, structure, and spectra. Some of the challenges of automated TEM analysis are given in the conclusion.

为克服单一卫星传感器成像的不足，提出了基于密集连接网络的合成孔径雷达（SAR）与多光谱影像的融合算法。首先分别对SAR影像与多光谱影像进行预处理，使用双三次插值法重采样到同一空间分辨率下，然后使用密集连接网络来分别提取影像的特征图，并采用区域能量最大的融合策略将深度特征进行融合，将融合图像输入到预训练的解码器中进行重构，获得最终融合影像。实验采用哨兵1号SAR影像、Landsat-8影像和高分1号卫星影像进行验证，并与基于成分替换、基于多尺度分解和基于卷积神经网络的代表性方法进行对比试验。实验结果表明，基于密集连接网络的融合算法在多尺度结构相似度指标上的精度高达0.930 7，在其他多种评价指标上也都优于其他融合算法，SAR影像的细节信息和多光谱影像的光谱信息得到很好的保留。

气溶胶光学厚度(AOD)是气溶胶浓度和大气浊度的重要表征参数。 通过遥感手段实现大气气溶胶光学厚度的反演是大气监测与治理过程中的重要方式, 其中遥感反演AOD的重点和难点是如何选择适合卫星传感器成像特点的方法和符合研究区域的气溶胶类型。 针对传统暗目标法无法直接应用于高分四号(GF-4)卫星多光谱遥感数据的问题, 通过研究得出了GF-4卫星多光谱数据中红、 蓝波段等效地表反射率的分布和两者之间的线性关系, 结合AOD反演原理改进暗目标法使其适用于GF-4卫星多光谱遥感数据; 分析6S辐射传输模型输入参数中气溶胶类型对AOD反演精度的影响, 结果表明气溶胶类型是影响AOD高精度反演的关键要素之一; 利用粒子群(PSO)聚类算法对京津冀地区气溶胶特性实测样本进行聚类分析, 通过分析各个气溶胶类型聚类结果的占比和半衰期变化情况, 最终确定聚类得到的C1、 C4型和6S模型内置的大陆型气溶胶类型进行京津冀地区的AOD反演。 为了验证不同气溶胶类型AOD反演结果的精度, 将反演结果与MODIS气溶胶产品和气溶胶自动观测网(AERONET)地基站点数据进行对比验证, 通过相关系数、 绝对误差等评价标准对不同气溶胶类型的适用性和特点进行评价。 实验结果表明, 以细粒子为主导的C4型气溶胶更满足京津冀地区夏秋两季的气溶胶特点, 与AERONET地基数据的一致性较好, 进一步证明了PSO聚类算法能够有效减小气溶胶类型的差异对AOD反演精度的影响。

Two-dimensional (2D) transition metal dichalcogenides alloys are potential materials in the application of photodetectors over a wide spectral range due to their composition-dependent bandgaps. The study of bandgap engineering is important for the application of 2D materials in devices. Here, we grow the Mo_1-xW_xSe₂ alloys on mica, sapphire and SiO₂/Si substrates by chemical vapor deposition (CVD) method. Mo_1-xW_xSe₂ alloys are grown on the mica substrates by CVD method for the first time. Photoluminescence (PL) spectroscopy is used to investigate the effects of substrates and interlayer coupling force on the optical bandgaps of as-grown Mo_1-xW_xSe₂ alloys. We find that the substrates used in this work have an ignorable effect on the optical bandgaps of as-grown Mo_1-xW_xSe₂. The interlayer coupling effect on the optical bandgaps of as-grown Mo_1-xW_xSe₂ is larger than the substrates effect. These findings provide a new way for the future study of the growth and physical properties of 2D alloy materials.