Study On Barrier Cobalt Chemical Mechanical Planarization Based On SVM

Integrated circuits(ICs)are the foundation of today's social development,and ICs have been filled in various industries.The manufacturing process of ICs is very complicated.One of the most critical technologies is chemical mechanical polishing(CMP).CMP is an indispensable prerequisite for ICs to achieve high integration.Cobalt(Co),with its excellent physicochemical properties,is a key material for the barriers of the technology nodes at 14 nm and below.However,as the feature size of the device decreases,the thickness of the barrier layer continuously decreases.In the chemical mechanical polishing process of the cobalt-based barrier layer,the removal rate and removal rate selectivity of cobalt and copper are difficult to control.At the same time,the chemical mechanical polishing process is very complicated.The traditional orthogonal experiment has the problems of long cycle,12-inch wafers and other consumables.The experiment can only be carried out for a limited number of times,relying on the comparative analysis of the experimental results to find the optimal slurry ratio.It is a difficult,time-consuming and costly task.This topic aims at the above problems by using the polishing liquid components: abrasive,active agent,chelating agent,hydrogen peroxide,BTA effect on Co/Cu CMP experiments,combining experimental data with emerging machine learning methods to train a predict the mathematical model of removal rate and removal rate selectivity,and derive the optimal cobalt and copper removal rate and the corresponding selection of the ratio of the polishing liquid components through this model.It is used to guide the CMP experiment and effectively shorten the research and development cycle to save costs.Research has important theoretical significance and practical value.This subject first studied the effect of pH,nano-silica sol abrasive,active agent and chelating agent on cobalt removal rate and static corrosion rate,and optimized a set of cobalt slurry with good performance through orthogonal test.Subsequently,hydrogen peroxide and benzotriazole were introduced to reduce the galvanic corrosion between copper and cobalt on the premise of ensuring the removal rate selectivity of cobalt and copper.The components of slurry(pH=10)were finally determined as: abrasive,active agent,chelating agent,hydrogen peroxide,and benzotriazole.Based on the above experimental results,cobalt and copper chemical mechanical polishing data were used to train a support vector machine(SVM)model.The genetic algorithm(GA)and particle swarm algorithm(PSO)were used to optimize the SVM parameters.Optimize the method and analyze the optimization results to establish the regression model.Finally,the trained SVM model was used to predict the optimal cobalt and copper removal rates and the corresponding slurry ratios to guide the CMP experiments and achieved significant results.