Identifying significant features for neural network-based particle contamination prediction in plasma etching process

Particle contamination on wafer surface during fabrication process has become a major issue in semiconductor manufacturing. Particles on wafer can cause the circuit to malfunction and wafer defect. This will lead to manufacturing costs and reducing productivity. Conventional method to control particles contamination managed to detect the particles contaminations only after the etching process is completed. Therefore, developing a system where an early detection of particle is crucial. In this study, methods for particle contamination prediction in plasma etching process using Artificial Neural Network (ANN) is proposed. The proposed method comprises the following steps: data collection, data pre-processing, feature selection and particle detection. First, data was collected from Statistical Process Control (SPC) database and Advanced Process Control (APC) database. Secondly, the raw data were pre-processed to remove missing values and unwanted attributes. Thirdly, the feature selection technique was studied to select the most relevant features or parameters indicating the number of particles contamination. Three feature selection techniques were used which are Minimum Redundancy Maximum Relevance (mRMR), Least-Square Feature Selection (LSFS) and Maximum Likelihood Feature Selection (MLFS). Finally, datasets with the selected features together with datasets without features selected were used as inputs for training and testing of ANN in particles contamination detection. Results for ANN showed that an equipment with feature selection method using MLFS result in the lowest error and highest R
²
value.