Effect of the principal component on the PCA-based neural network model for HFO2 thin film characteristics

Principal component analysis (PCA) based neural network models for the HfO2 thin film characteristics, such as the accumulation capacitance and the hysteresis index, grown by metal organic molecular beam epitaxy are presented. Considering the number of the principal components, the various input parameters are applied to the neural network modeling. In order to build the process model, the error back-propagation neural networks are carried out and the X-ray diffraction data are used to analyze the characteristic variation for the different process conditions and predict the response models for the characteristics. PCA is selected to reduce the dimension of the data sets. The compressed data are then used in the neural networks and those initial weights and biases are selected by Latin Hypercube sampling method. From this analysis, the effects of the principal components on the neural network models are examined.