Deconvolution of SIMS depth profiles: Towards simple and faster techniques

Abstract The continued reduction in the scaling of semiconductor components has pushed the technique of SIMS to achieve unprecedented degrees of high depth resolution. For this purpose, various deconvolution methods have been developed during the last decade, among which is the iterative algorithm with hard constraints and Miller regularization, the heart of this contribution. In this paper, a comparative study in both cases of considering and neglecting the regularization term characterizing this algorithm is provided and analyzed. Both algorithms are tested on several theoretical structures and implemented for the deconvolution of a real structure. It was shown that the algorithm without the regularization term provides regular solutions close to the results given by the basic algorithm using this term, with a reduction in the number of iterations. Besides, gains in depth resolution and in maximum peak intensity without the regularization term are better than those achieved using it. These features may favor the first approach, and attribute it more considerations in SIMS. However, the proposed approach in cases of Gaussian and thick-layer profiles has the semiconvergence property. In this work, the behaviour of the deconvolved profiles as a function of the uncertainties in the depth resolution function (DRF) fit is also addressed.