On the choice of tuning parameters for use with Robust GCV, Modified GCV and the Discrepancy Principle in the inversion of ARXPS data

Abstract

Composition depth profiles were extracted from simulated ARXPS data using regularization, with the regularization parameter determined by three different methods (Robust GCV, Modified GCV, and the Discrepancy Principle) that require tuning parameters. For each method, the optimal tuning parameter was determined for two input profile shapes, three Tikhonov regulators (0th, 1st, and 2nd order), and data noise ranging from 1% to 9%. Although universally applicable optimal tuning parameters were not identified, it was found that certain values could consistently produce acceptable results for the input profiles used in this study.