Diagnosis of plasma processed X-ray photoelectron spectroscopy using principal component analysis and neural network

Plasmas are sensitive to a variation in process parameters or equipment components. To maintain device yield and equipment throughput, abnormal plasma states should be detected and diagnosed. An ex-situ diagnosis technique is presented. This was accomplished by predicting plasma-processed X-Ray photoelectron spectroscopy using neural network. The technique was evaluated with the experimental data collected during the plasma etching of silicon carbide thin films. A total of seventeen XPS patterns were collected. For this and principal component analysis (PCA)-reduced data, neural network recognition models were constructed and evaluated as a function of hidden neuron number. Both XPS and PCA-XPS models demonstrated a perfect recognition as well as comparable prediction performance. Meanwhile, the original XPS data were separated into other training and test patterns to examine the performance of prediction-based diagnosis. In this case, the PCA-XPS model corresponding to 100% data variance yielded better prediction and diagnosis than pure XPS model.