Disturbance Rejection Run-to-run Control For Semiconductor Manufacturing Processes

The development of the semiconductor manufacturing industry is a reliable guarantee for enhancing the comprehensive strength of our country and guaranteeing national security.With the development of technology,Wafer sizes continue to increase and critical dimensions are gradually reduced.In the semiconductor manufacturing process,the phenomena of "random disturbance,model mismatch,multi-product,multi-stage" are becoming ever-worsening,which puts more stringent requirements on the precise control of the wafer production process.Run-to-run(RtR)control is one of the important means to improve the overall effect of the process and increase the Wafer yield.Under the funding of the national natural science foundation of China,the RtR control of the semiconductor manufacturing process is taken as the research object,the issues of random disturbances,carry-over effects,and multi-stage influences are considered,and the research on the disturbance estimation and RtR control algorithm design is carried out in the study.The main contents are as follows.Aiming at the problems of model mismatch,drift disturbance caused by using consumables and external disturbance caused by environmental changes in the semiconductor manufacturing process,the RtR control algorithm based on the extended state observer(ESO)is proposed.The external disturbances and model mismatches are considered as a total disturbance.Discrete linear and non-linear ESOs are designed and a RtR controller is designed based on that to estimate and restrain the total disturbance.Based on the idea of thread control,the algorithm is also transformed for mixed-product process control.Furthermore,the stability region is calculated and discussed.Finally,the validity of the proposed method is verified by numerical simulation and reverse engineering.In addition to complex determinant and stochastic disturbance,carryover effects are common in the semiconductor manufacturing process.In order to overcome the carryover effect and stabilize the process output,a discrete active disturbance rejection RtR control algorithm is proposed in this paper.First of all,the process dynamics are described by a discrete transfer function.Secondly,according to the principle of ADRC,a DADRC algorithm for SISO linear discrete systems is proposed.With an incremental state space considered as a canonical form,an ESO and a state error feedback control are constructed and a target filter is also introduced.Therefore,disturbance suppression and target tracking are achieved in the semiconductor manufacturing process.Furthermore,a model-based discrete auto disturbance rejection control algorithm is derived based on the offline identification process model.Then the parameter tuning method is designed and the stability is analyzed.Finally,numerical simulations verify that the disturbance rejection capability and target tracking performance of the proposed algorithms are excellent.Based on the research of control algorithms for a single process.In order to improve the overall Wafer yield of the multi-stage semiconductor manufacturing process,a multi-stage RtR control algorithm is proposed.The chemical vapor deposition(CVD)and chemical mechanical polishing(CMP)are taken as an example.Based on the analysis of the characteristics of the two processes,A virtual metrology(VM)model for the CVD process is designed based on the Gaussian process regression method to estimate the Wafer output quality and calculate the estimated credibility.And then the multi-stage RtR control algorithm is constructed by combining the Wafer quality estimate with the double exponential weighted moving average(d EWMA)RtR control of the CMP process.Besides,the performance of the algorithm is analyzed.Numerical simulation and reverse engineering verify the accuracy of the VM model and the effectiveness of the multi-stage RtR control algorithm.The issue of the performance declination with environmental changes in the semiconductor manufacturing process system is unavoidable in practical applications.In order to monitor the process status and diagnose the cause of the failure,the best achievable performance index(BAPI)and model evaluation index(MEI)are proposed.Firstly,based on the estimated state of the linear ESO,the process disturbance parameters are estimated by using the recursive extended least square method.Secondly,the BAPI of the process is established according to the minimum variance performance,which is used to monitor the change of RtR control performance of the process.In addition,the system MEI is established according to the internal model control principle to assistant the adjustment of RtR control parameters.Finally,it is verified through numerical simulation that the degradation of system control performance can be reflected by the BAPI;and the model-mismatch can be reflected by the MEI.The research results provide solutions for key control problems in the semiconductor manufacturing process,which will promote the practical application of RtR control methods in the semiconductor manufacturing industry,and alleviate the "stuck neck" problem in the IC industry.Therefore,the research is with high-level theoretical value and a great potential application perspective.