Imaging Method Of High Speed Atomic Force Microscope

Nanotechnology has become one of the most significant technologies in thetwenty-first century. The development of nanotechnology relies heavily on the meansof nano-observation and nanomanipulation－Scanning Probe Microscope (SPM),among which anAtomic Force Microscope (AFM) is widely used in chemistry,materials, physics, biology and other nano-related scientific fields due to thefollowing two reasons:1) the usage of an AFM is not limited by the sampleconductivity;2) it supports the sample observation either in the atmosphere or liquidenvironments. Therefore, the development of the atomic force microscope technologyhas greatly promoted the progress in the field of nanotechnology research. However,with the fast development of nano-technology, researchers put forward higherrequirements on the AFM nano-measurement and nano operational performance, forwhich there presents significant difficulty for exisiting AFMs due to its inherentdefection. These difficulties are as follows:1) the scanning speed/efficiency is slow,and it is unable to meet the imaging requirements for real-time requirements of highlyactive samples;2) in high-speed scanning, complex non-linear characteristics ofatomic force microscopy will lead to large imaging errors of the sample topography;3) the operation of AFMs is very difficult, since there are vaious Scan Modes, and itusually takes long time to train operators for the adjustment of scanning parameters.These restrictions severely hamper the in-depth study in the field of nanotechnology.To address these problems, this dissertation implements in-depth study for theAFM imaging methods to further improve the performance of AFMs in thehigh-speed scanning tasks. Specifically, its main work includes the following fouraspects:First, for the control signal in the imaging system, two kinds of preprocessingtechniques are proposed based on the dynamics of the piezoelectric scanning tube andsample characteristics of the imaging signal. First, in order to alleviate the dynamiceffects, this dissertation proposes a model identification method to reduce the imaging errors in high-speed AFM scanning, based on which the control signal is transformedinto the actual stretching amount of the piezoelectric scanner. Meanwhile, it isdifficult to ensure that the sample is completely horizontal when the sample is placedon the loading platform, which causes the control signal to generate a scan slant. Inorder to facilitate further analysis of the subsequent chapters, the dissertationproposes a convenient online real-time removal preprocessing method to reduce theslant plane. Simulation and experimental results show that the proposedpreprocessing methodssignificantly inhibites the imaging errors due to the dynamiccharacteristics of the piezoelectric scanner and at the same time eliminates the slantoccuring in sample placing.Second, the dissertation designs an improved dynamic imaging method based onthe integration of the near point set. In high speed scanning, to suppress the negativeeffect of the strong nonlinear dynamics on the AFM image quality, this dissertationproposes a high-speed AFM imaging method based on the fact that the scanning pointhas similarity with the neighborhood. This imaging method takes into accunt bothsystem nonlinearity and the priori knowledge, and thus a nonlinear filter isconstructed with its coefficient being a function of the control error signal, based onwhich the imaging error is suppressed. Simulation and experimental results show thatthe imaging algorithm can dramatically improve the accuracy of high-speed scanningof the AFM imaging.Third, in order to improve the imaging accuracy of the tapping AFM, atopography estimation method is designed based on an adaptive unscented kalmanfilter (AUKF). The tapping-mode AFM imaging model is first analyzed, simplified,and the discrete-time model is derived. On this basis, the sample topography isintroduced into the imaging filter system as the process noise. While taking advantageof the large amount of dynamic data in tapping scanning mode, the nonlinearimagingsystem isfiltered with the parameter adaptive UKF. Thisdissertation uses the variationof the convariance over time to estimate the actual noise level of the current process,thus achievingself-adapting for the process noise covariance parameters.Fourth, for the convenient operation of AFM, a data-driven based control andimaging method is designedto accomplish the self tunning of control parameters. When sample, scanner, probe or scanning parameters are changed, the modelparameter of the AFM imaging system can be changed accordingly. Thus, adjustmentof the controller parameters should also be made repetitively. This dissertationproposes a data-driven based control and imaging method, which is capable ofadjusting the control parameters automatically so as to reduce the difficulty in AFMoperation. Specifically, CARIMA(Controlled Auto-Regressive and Moving-Average)model is introduced to describe the local dynamiclinearized system model, which isthen identified using data-dirven based methods. Subseqently, theproportional-integral(PI) parameters is calculated online based on global predictivecontrol (GPC) optimization method, and thus the automatoic tunning of PI prametersis accomplished. Simulation and experimental results are provided to show thatwhenthe scanning speed is changed or the PI parameters are chosen unappropriately,the proposed method still works well toimprove the imaging accuracy.