Reaearch On Ion Beam Polishing Removal Function Algorithm Based On RBF Neural Network

With the rapid development of science and technology, more and more ultra-smooth surfaces are required for the research on optics, microelectronics, astronomy and other areas. Those areas also have higher and higher requirements for the surface accuracy about critical optical elements, which is even to achieve sub-nanometer precision. It is because of the demand inducing the rapid development of ultra-precision processing technology, so that ion beam polishing techniques owning unique advantages emerges. It is an advanced technology based on atomic sputtering theory, and it can achieve polishing of stress-free and non-contact atomic scale for the material of the optical element surface. In this thesis, ion beam removal function of ion beam processing technology is mainly studied to make the processing precision improved, and theoretical model of removal function is established and corrected to make the model estimation accuracy improved. According to the established removal function model and ion beam polishing principle, dwell time of processed point can be calculated. From the viewpoint of ion beam used to compose the removal function, the Ar+ bombardment of quartz glass(Si O2) is treated as a research model in the thesis. And the main contents include:(1) Research on ion beam characters: Physical sputtering principle and sputtering yield were studied in the process of ion beam polishing in the thesis. Sputtering yield can be primarily affected the incident ion species, angle of incidence, and ion beam energy. Sputtering experiments were to be done to obtain the relevant data. And these obtained data was simulated by MATLAB software in order to discuss the relationship between sputtering yield and related process parameters.(2) Research on ion beam removal function mathematical model: Removal rate of the optical element surface was studied under the uniform ion beam bombardment. And the theoretical model of ion beam removal function was established. Through this model, we can know that ion beam removal function is affected by sputtering yield. In order to achieve removal function of the multi-angle controlling, we established the relationship between the angle of incidence and sputtering yield. The key parameter, namely sputtering yield, was calculated based on RBF neural network algorithm. Relevant parameters of the network were selected, and we can use the trained network to reckon the relationship between the angle of incidence and sputtering yield.(3) Solving and optimizing dwell time: SVD algorithm and LSQR algorithm were applied to solve dwell time of ion beam polishing. The simulations about polishing process were to be done and the arising edge effects of the simulations were to be analyzed and resolved.