Improvement of Depth Resolution in Secondary Ion Mass Spectrometry Analysis Using Multiresolution Deconvolution

Abstract

Multiresolution deconvolution (MD), based on Tikhonov–Miller regularization and wavelet transformation, was developed and applied to improve the depth resolution of secondary ion mass spectrometry (SIMS) profiles. Both local application of the regularization parameter and shrinking the wavelet coefficients of blurred and estimated solutions at each resolution level in MD provide to smoothed results without the risk of generating artifacts related to noise content in the profile. This led to a significant improvement in the depth resolution. The SIMS profiles were obtained by analysis of delta layers of boron in a silicon matrix using a Cameca-Ims6f instrument. The results obtained by using the MD algorithm are compared to those obtained by monoresolution deconvolution which is Tikhonov–Miller regularization with a model of solution (TMMS). Finally, the advantages and limitations of the MD algorithm are presented and discussed.

Signal Processing of Secondary Ion Mass Spectrometry Profiles. New Algorithm for Enhancement of Depth Resolution

“…The Tikhono-Miller regularization is achieved through a compromise between choosing a solution that both leads to a reconstructed signal close to the measured data and conform to some prior knowledge of the original signal [2,[4][5][6]. This means that the solution x is considered to be close to the data if the reconstruction signal Hx is close to the measured one y, i.e.…”

“…Indeed, D is usually designed to smooth the estimated signal, and then a gradient or a discrete Laplacien is conventionally chosen. Its spectrum is a high-pass filter [2,9], this results in the minimisation of the quadratic functional proposed by Tikhonov:…”

“…In the last few years, improvement of depth resolution in secondary ion mass spectrometry (SIMS) analysis is a critical issue for depth profiling of silicon semiconductor films [1][2][3][4].…”

“…The origin of these oscillations is related to the presence of strong local concentrations of high frequencies in the signal which belong to noise. For this reason, it is important to eliminate noise components from the signal [2]. Denoising with the sole purpose of extracting desired information from measured data has proven to be a crucial preliminarily steps in any analytical method.…”

Multi-Scale Deconvolution of Mass Spectrometry Signals

“…The different steps in the multiresolution deconvolution algorithm are as follows (Boulakroune, 2009). …”