Surface Quality Prediction Of Ultra-precision Ground Single-crystal Silicon Considering Elasto-plastic Deformation Characteristics

Single-crystal silicon wafers are widely used in the field of integrated circuit manufacturing.Ultra-thin wafer is a key enabling factor for the implementation of Through Silicon Via 3D(TSV-3D)packaging technology and achievement of high integration level.Ultra-precision grinding is a common process in silicon wafer back-thinning process due to its high efficiency,high quality,and low cost.As the target thickness of wafer thinning decreases,how to control the depth of subsurface damage introduced by the grinding and thinning process while ensuring the efficiency of the thinning process has become an important issue for the grinding process of TSV-3D IC wafers.Ground surface quality prediction model is the theoretical basis for the process optimization,however,there are few reports in the literature on the prediction of ground surface quality based on silicon grinding mechanism.The selection of wafer thinning process parameters on existing production lines still comes from empirical guidance from process trials provided by grinding wheel manufacturers.Once the wafer thinned thickness changes,costly process trials need to be redesigned to find the optimal process parameters.Moreover,it is difficult to optimize the process parameters in different grinding stages based on empirical parameters from the process trials.Therefore,the study of singlecrystal silicon ground surface quality prediction is of great significance for wafer thinning process and the semiconductor industry.To meet the needs and visions of industrial production and scientific research for understanding the nature of the grinding process and establishing grinding prediction models,this dissertation starts from the investigation of silicon grinding mechanism.First,a single-grit scratch test apparatus was developed with scratch conditions similar to those of silicon ultraprecision grinding.The deformation characteristics of single-crystal silicon were explored under high-speed scratch conditions with nano-cutting depth.Subsequently,this dissertation proposed a grinding wheel grit characterization method based on grinding marks analysis,and characterized the key characteristics of the grinding wheel grits.Finally,a prediction model for the surface quality of single-crystal silicon ultra-precision grinding at macroscopic scale was established in combination with grinding kinematic analysis,which is an important guideline for understanding the nature of ultra-precision grinding process and guiding the parameter optimization of grinding cut thinning.The main research works and conclusions of this dissertation are as follows.(1)A single grit nanoscale depth-of-cut high-speed scratching test apparatus was developed for analyzing the ultra-precision grinding mechanism of single-crystal silicon.The platform can simultaneously meet the following conditions or functions,including nanoscale scratching depth,m/s scratching speed,submicron tool tip radius,controlled scratching length and online monitoring of scratching load.(2)Based on the developed scratching apparatus,scratch tests were carried out and the elastic deformation characteristics of single-crystal silicon was quantified.The elastic recovery increases as the scratch speed decreases.A critical resolved shear stress criterion applicable to the prediction of single-crystal silicon dislocation layer depth was proposed:the resolved shear stress along the direction of dislocation loop propagation within the single-crystal silicon slip planes is important in predicting the single-crystal silicon dislocation layer depth.Meanwhile,based on the scratching test results,this dissertation analyzed the applicability of existing damage prediction models in high-speed scratching single-crystal silicon tests from three indicators,i.e.subsurface crack depth,phase transformation layer depth and surface plastic zone width.The results showed that the defect size of single-crystal silicon during scratching depends mainly on the scratching load,while the scratching speed has a significant effect on the residual geometric characteristics of the scratches.(3)A novel method was proposed to characterize the active grinding grits of two types of wafer grinding wheels,SD3000 and SD600,including the distribution of the active grinding grits in space and the distribution of the grit tip radius of the active grinding grits.The results showed that the number of active grinding grits was much smaller than that of grits exposed on the grinding wheel surface,with the number of active grinding grits per unit area being about 1.4/mm2 for SD3000 grinding wheels at 100 nm grinding depth and 0.51/mm2 for SD600 grinding wheels at 1000 nm grinding depth.Therefore,it is no longer reasonable to simplify the grit tip size as the average grit size in silicon ultra-precision grinding.The distribution of the grit tip radius needs to be considered in the simulation process.(4)Based on the elastic-plastic deformation characteristics of single-crystal silicon characterized in(2)and the active grinding grit characteristics of the grinding wheel characterized in(3),a single-crystal silicon ultra-precision grinding surface quality simulation a prediction model for the surface quality of single-crystal silicon ultra-precision grinding at macroscopic scale was established in combination with grinding kinematic analysis.The model was able to predict the workpiece grinding surface morphology and its corresponding subsurface damage profile under the given grinding wheel parameters,workpiece parameters and grinding parameters.Subsequently,grinding simulation and experiments with different grinding parameters were designed to validate the prediction model in terms of three evaluation indexes:namely,grinding surface roughness,grinding surface power spectrum density and grinding subsurface damage depth.