Parts | Days | Selection | Search | Downloads | Help

MS: Massenspektrometrie

MS III: HV III

MS III.1: Invited Talk

Monday, March 22, 2004, 14:00–14:30, HS 112

Bayesian Decomposition of Quadrupole Mass Spectra — •Thomas Schwarz-Selinger, Hyungdong Kang, and Volker Dose — Centre for Interdisciplinary Plasma Science, Max-Planck-Institut für Plasmaphysik, Boltzmannstr. 2, D-85748 Garching

Identification of components in a gas mixture from quadrupole mass spectra is a challenging task due to fragmentation of the molecules by electron impact in the ionizer. The pedestrian subtraction method and/or least-square evaluation are often used for the decomposition of multi-component mass spectra. These however suffer from the excessive error build-up and provide only poor and sometimes unphysical results like negative concentrations.

We present an advanced analysis technique of mass spectra using Bayesian probability theory. It handles the data and all other information consistently and is able to identify components in unknown mixtures via model comparison. It provides estimates not only for the species concentrations but also for their cracking pattern. In addition the confidence intervals for the values of interest are derived. Therefore this method does not only allow to identify reliably radicals in an unknown mixture but also allows to determine their cracking pattern. The basics of the new method are presented together with illustrating examples from low temperature methane plasmas.