Research Of Micro-nano Structure Imaging Based On Microwave Near Field

Microscopic technology in the microscopic world is an effective way to observe the internal structure and surface characteristics of materials in modern science,and its technological progress is closely related to the development of physics,chemistry,microbiology and other fields.In recent years,the rapid development of integrated circuits and nanomaterials has increased the demand for microstructure size device detection technology.Traditional optical microscope and detection technology can no longer meet the microscopic requirements of some specific fields,especially highly integrated components,such as non-damage,electrical characteristic parameter measurement,subsurface detection,etc.Microwave near-field detection technology breaks through the Abbe diffraction limit and enables accurate analysis of substances below half a wavelength.At the same time,its working source is microwave,microwave has penetration,very sensitive to changes in the electromagnetic characteristics of the material,so microwave near-field microscope can achieve non-destructive detection of samples,electromagnetic parameter measurement,surface imaging,etc.This thesis builds a simulation environment for the microwave near-field detection system,and conducts research on its theoretical limit resolution and simulation imaging.The specific content includes(1)According to the experimental equipment and working conditions,the simulation software is used to simulate the microwave near-field working environment,which reduces the influence of the environment and instrument performance on the experiment.The structure of the resonant cavity is optimized by simulation,and the quality factor of the quarter-wavelength coaxial resonant cavity is improved.(2)At the same time,a one-dimensional simulation analysis was carried out on the sample,and the limit resolution of the microwave near-field microscope was obtained at about 80 nm.Afterwards,in the process of two-dimensional imaging of the sample,the boundary thinning scanning method was used to obtain a clear topography of the grid mask,and then the influence of the Z-axis direction on the imaging effect was discussed,and the three-dimensional imaging of the microwave near-field was carried out.Research.(3)In view of the situation that the imaging results of the microwave near-field experiment are not clear due to system errors,the scanning method of the probe is optimized.In the experiment,the imaging effects of two different scanning trajectories were compared,and the S-type scanning trajectory was selected for follow-up research.Secondly,different step length experiments were carried out on the single scan step length,and the imaging optimization results were obtained in which the scan step size was set to one-tenth of the minimum size of the sample.Finally,samples with different electromagnetic parameters were detected,and the imaging differences of resonant cavity parameters were compared and analyzed.(4)The method of machine learning is used to realize the identification and detection of common material categories in integrated circuits.First,the problem of how to select the model dataset is discussed,and then the redundant data of the height scan curve is processed.Finally,the three methods of support vector machine,random forest and BP neural network are used to learn and predict the measured height scan curve,and the prediction performance of different models and different parameters of resonator is compared and analyzed,and the results show that random forest and S12 have the best prediction effect,and the accuracy can be close to 100%.