Study On The Intelligent Optimization Technology In Optimizing The CMP Process And The Polishing Slurry Of Copper

With the rapid development and advanced integration of the giant large scale integrated(GLSI)circuits,the feature sizes of the interconnecting wire become samaller,so the demand for the flatting becomes higer.In the pattern copper chemical mechanical polishing(CMP)process,the polishing slurry and the CMP process are two important factors which seem to directly affect the polishing rate and the planarization performance.Different CMP process and different proportion of each component in polishing fluid will result in different polishing rate and different effect for planarization performance.Thus,in order to improve the the CMP process and the slurry performance,how to establish the mapping relationship between the CMP process and the proportion of each component in the slurry with the polishing rate and the planarization performance is a very important issue.Aiming at this issue,the computer intelligent optimization technology was introducted in this paper to analysis the relationship between the CMP process and the copper alkaline polishing slurry with the polishing rate and the planarization performance.The copper alkaline polishing slurry mainly contains FA/O chelating agent developed by the Institute of Microelectronics of Hebei University of Technology.First,as an interdisciplinary research topic,on the basis of studying the CMP alkaline reaction mechanism,the computer intelligent optimization technology was studied further.Back propagation(BP)neural network was combined with the artificial bee colony(ABC)algorithm,namely BP-ABC algorithm,as a modeling method of existing.In view of the BP-ABC algorithm's poor ability to predict,the improved BP-ABC algorithm,namely Back Propagation Niche CrowdingArtificial Bee Colony(BP-NCABC)algorithm was proposed in this paper.Through the simulation experiments proved that the BP-NCABC algorithm had a better modeling ability and generalization ability than the BP-ABC algorithm.Then,the simulation experiment of CMP experimental data had been carried out by BP-NCABC algorithm and then established the relational model among the proportion of each component in the slurry and the polishing rate,which can predict the polishing rate.By the same way,aiming at the CMP experimental data of the 300 mm copper blanket wafer,the relationgship between the CMP process and the polishing slurry with the polishing rate and the With-In-Wafer-Non-Uniformity(WIWNU)was established,which can predict the polishing rate and the WIWNU at the same time.Thus the efficiency of optiminating the CMP process and the slurry was improved.Last,according to the two relational models mentioned aboved,the method of sensitivity analysis was adopted to quantize the influence of the CMP process and the each component in the slurry on the Cu polishing rate and the WIWNU during the CMP process.According to the results of the quantitative analysis,the actual experiment results showed that: when the pressure was 1.2psi,the speed was 87r/min,the flow rate was 300ml/min,the abrasive concentration was 7vol.%,the oxidizing agent concentration was 0.5vol.%,the FA/O type chelating agent concentration was 2.5vol.% and the surfactant concentration was 3vol.%,the polishing rate was 907.39nm/min,the WIWNU was 2.92%,the step height difference of before and after polishing was about 3100? and the surface roughness was 0.386 nm.And the surface defects of copper film reduced obviously after polishing.The planarization performance was good,meetting the requirements of the industrial development of the CMP process.