The Study On The Key Techniques For Process Control In Modern Wafer Manufacturing

Modern semiconductor manufacturing is one of the most complex processes in manufacturing industry.Modern semiconductor manufacturing processes involve more than 300 steps,and the processing cycle is usually more than two months,which makes quality control particularly important.In order to achieve high yield in chip manufacturing,Statistical Process Control(SPC)technology has been widely used in wafer manufacturing processes to monitor the entire manufacturing process.These processes can be subdivided into single-wafer and batch processes.Single-wafer processes are those that process one wafer at a time,while batch processes are those that process multiple wafers simultaneously.With the shrinking of the critical dimension(CD)and the advanced equipment,the new data characteristics such as nesting and fixed position effects appear.What's more,the yield,performance and reliability of chips are sensitive to particulate contamination,then higher level of cleanliness is achieved,causing new data characteristic of the measured particle count data.These new data characteristics make the existing statistical process control technology no longer applicable.In order to ensure the quality and reliability of chips,the new statistical process control methods are needed to be developed for monitoring the single-wafer processes,batch processes and the particle in the air.This dissertation focuses on the reasearch of SPC of modern wafer manufacturing and the number of particles in its corresponding environment.According to the data characteristics of nesting and fixed position effects in single-wafer and batch processes,excess zeros of the particle data in the air,the corresponding mathematical models are established.The corresponding control charts are developed based on theoretical analysis and computer simulation.Thus,the statistical process control for modern wafer manufacturing and the number of particles in its corresponding environment is realized.The research results of this dissertation can be widely used in semiconductor manufacturing,which can improve the quality and reduce the defects of products in semiconductor manufacturing and improve the market competitiveness of the enterprises.The main works of this dissertation include:(1)The one-level nested and fixed position effects model is introduced based on the data characteristics in single-wafer processes.Ten critical mathematical conclusions are obtained through theoretical derivation.Based on these conclusions,the test method of nesting and fixed position effects is given.The effects of the fixed position effects on the conventional s(standard deviation)control chart and R(range)control chart are discussed.It is proved that the application of s and R control chart would result in too wide gap between UCL and LCL while the fixed position effects exist,which means that s and R control charts are not sensitive enough to detect the occurrence of all unusual cases.Two SPC methods are presented in this dissertation.The first method includes two control charts,while the second method consists of three control charts,including the batch average,the within-wafer standard deviation and the within-wafer tendency control charts.They are used to monitor the between-batch variation,the within-wafer variation and the within-wafer fixed position effects.Note that,a new method for calculating the control limit of the standard deviation control chart is proposed for the within-wafer fixed position effects,and the within-wafer tendency control chart is first proposed in the SPC area.Via comparison of the two SPC methods,it is pointed out that the second method proposed in this paper can also quickly judge the out-of-control mode from the control charts,and it can also distinguish between "good" and "bad" out of control.These two advantages can help engineers to find out the cause of out-of control or to improve process in practical applications of single-wafer processes.Finally,the two SPC methods are verified and compared by practical application in the process of CMP.All works will lay down a good foundation on research of batch process.(2)The two-level nested and fixed position effects model is introduced based on the data characteristics in batch processes.Eight critical mathematical conclusions are obtained through theoretical derivation.Based on these conclusions,the test method of two-level nested and fixed position effects is given.Two SPC methods are presented for batch process in this dissertation.The first one includes three control charts,while the second method consists of five control charts,including the batch average,the between-wafer standard deviation,the between-wafer tendency,the within-wafer standard deviation and the within-wafer tendency control charts.They are used to monitor the between-batch variation,between-wafer variation,between-wafer fixed position effects,within-wafer variation and within-wafer fixed position effects.Via comparison of the two SPC methods,it is pointed out that the second method proposed in this dissertation can quickly judge the out-of-control mode from the control charts,and it can also distinguish between "good" and "bad" out of control.These two advantages can help engineers to find out the cause of out-of control or to improve in practical applications of batch process.Finally,the two SPC methods are verified and compared by practical application in the process of thermal oxidation.(3)The statistical process control for monitoring particles with excess zero counts in modern wafer manufacturing is studied.The threshold-Poisson model is proposed based on Poisson distribution and the cause of excess zero counts.A suitable parameter-estimation method is developed.Then,by using 15 groups of measured particle data with excess zeros,the determination coefficient,AIC and BIC all suggest that threshold-Poisson model has better performance than the other popular models such as Poisson,zero-inflated Poisson(ZIP),generalized zero-inflated Poisson(GZIP),Neyman and Gamma-Poisson models.After that,a control chart named the TC control chart is proposed and the control limits are derived.The minimum sample size needed to construct the TC control chart is discussed.The methods and key points for establishing the EWMA and CUSUM charts based on the threshold-Poisson model are also discussed.Finally,we use the threshold-c control chart to monitor the particle count,and the result is reasonable.Meanwhile,the determination coefficient is very close to 1,which illustrates the correctness of the model and the parameter estimation method.