Study On Surface Quality Of High Precision Micro Milling Of Monocrystalline Silicon

Monocrystalline silicon is a kind of semiconductor material,which is widely used in the field of microelectronics.Traditional processing methods are mainly lithography and chemical etching,which can only process 2D or2.5D structures.In contrast,micro-milling can produce 3D surfaces.Because of the special physical properties of monocrystalline silicon,cracks are easy to occur in processing,and the material is mainly removed by crushing,the surface quality of monocrystalline silicon is poor and difficult to predict,so it is very necessary to study the optimization method surface quality of monocrystalline silicon.In this paper,monocrystalline silicon is the object of study.Firstly,the structure of monocrystalline silicon is introduced,the process of micro-milling is analyzed,the cutting trajectory equation of micro-milling cutter is established,and the law of crack formation and extension is analyzed.The surface roughness of brittle materials is analyzed theoretically,and the evaluation method of surface integrity is determined.Secondly,a micro-milling experimental system is built,and the experiment of micro-milling of monocrystalline silicon are carried out.The influence of micro-milling process parameters on R_ais studied by orthogonal experiments,the result is f_z>n>a_p.Then,through the single factor experiment to study its law,the results show that R_aincreases with the increase of f_zand a_p,and increases first and then decreases with the increase of n;with the increase of f_zand a_p,the microgroove width K,the number of pits N and edge change B all increase,and with the increase of n,the microgroove width K and edge change B increase,the number of pits N increases first and then decreases.In the experimental range of f_z=0.05μm/z～0.8um/z,n=5000r/min～30000r/min,a_p=5μm～20μm,the optimal process parameters is f_z=0.05μm,n=5000r/min,a_p=5μm.Then,the Acoustic Emission(AE)signals are collected by changing the micro-milling process parameters,it is concluded in the time domain that the AE RMS and Energy increase with the increase of f_z,a_p,n,and the damage of microgrooves will lead to the sudden increase of RMS and Energy.In the frequency domain,it is concluded that the amplitude of the power spectrum in the 20～50k Hz(frequency band 1)and 80～180k Hz(frequency band 2)is higher,and the amplitude of frequency band 2 is higher than that frequency band 1;with the increase of f_z,the amplitudes of both bands increase;with the increase of n,the amplitudes of the two bands remain basically unchanged;with the increase of a_p,the amplitude of frequency band 1 increases gradually,and the amplitude of frequency band 2 remains basically unchanged,and microgrooves damage will lead to the amplitude increasing.The corresponding relationship between AE and R_ais established.R_aincreases with the increase of RMS value and R_aincreases with the increase of the amplitude of frequency band 1,and the change trend is consistent.In particular,based on a large number of experimental data,the surface roughness prediction model of monocrystalline silicon micro-milling is constructed by using multiple linear regression and BP neural network respectively.By calculating the correlation coefficient R~2,F test,confidence interval,MSE and training sample R value,it is proved that the two prediction models have high reliability and accuracy.Finally,the comparative analysis shows that the BP neural network method is better,and the optimal process parameters is verified by the experiment.This model can predict the surface roughness of monocrystalline silicon micro-milling and guide the selection of machining parameters.