Abstract An important step of the factor analysis method is the determination of the number of significant factors. Common techniques as Malinowski indicator function provide perhaps uncertain results in spectroscopic cases depending on signal to noise ratio, the overlap of spectra, sample misalignment and concentration of the factors. A new method for the determination of the number of components in Auger depth profiling is established. The new method of correlation of abstract basic spectra (CABS) takes into consideration the influence of experi- mental error especially noise. Simulations were carried out, to prove the CABS method. The results are compared with those obtained by established methods. Auger depth profiles of SiC and Si 3 N 4 -layers on Si(111) were analyzed by means of the CABS method and Target Factor Analy- sis. Aspects of data pretreatment as filtering are discussed. Introduction Factor analysis has become a common technique to ana- lyze Auger depth profiles. If more than one chemical state of an element is supposed within a layer system or if the chemical bonds are changing via sample depth, factor analysis is the suitable method for quantification of Auger depth profiles. The factor analysis produces three impor- tant results. The first result is the number of significant chemical states the sample consists of, as a second we get the spectra of the chemical components and as a third, factor analysis provides their concentrations. This paper deals with the estimation of the number of chemical components in factor analysis of Auger depth profiles. This important step of factor analysis is com- monly connected with empirical criteria as the Mali- nowski indicator function [1]. But in the spectroscopic case this empirical criterion provides ambiguous results depending on experimental error as noise and sample mis- alignment [2]. This fact leads to uncertainties in the esti- mation of the number of components. By using factor analysis programs the user has to estimate the number of components by comparing different empirical criteria, but finally the decision is a manual operation. To improve this unsatisfactory state in electron spectroscopy some graphi- cal methods of estimation of sample components were es- tablished [3, 4]. This paper puts forward a new method of the estimation of sample components, which oppresses the influence of experimental noise. The influence of sample misalignment is not discussed here, but it can be omitted by measuring the same spectrum under the same experimental conditions several times [2]. Experimental Measurements were carried out with Auger spectrometer ASC 2000 (Riber). The Auger depth profiles were acquired at 3 keV primary electron energy, 1 μA current, 0 ... 22.5° incidence angle. The properties of the cylindrical mirror analyzer were 0.3% energy resolution, modulation 0.95V eff at frequency of 14 kHz. An Ar + sputter beam of an energy of 1 keV, 60° ... 80° incidence angle and 10 ... 32 μA/cm 2 current density was used. The spectra (Si LVV peak) were recorded in the kinetic energy range of 70 to100 eV with 0.5 eV energy steps. The noise was recorded in the energy range of 150 to 180 eV with 0.5 eV energy steps, too. A sample of Si 3 N 4 on Si(111) was prepared with a layer thick- ness of 95 nm. Furthermore a 30 nm SiC layer on Si (111) was de- posited by solid source molecular beam epitaxy at 680 °C growth temperature. Ralf Pieterwas · Gernot Ecke · Rastislav Kosiba · Hans Rößler A new method of the determination of significant factors with factor analysis in AES Fresenius J Anal Chem (2000) 368 : 326–334 © Springer-Verlag 2000 Received: 20 January 2000 / Revised: 5 April 2000 / Accepted: 16 April 2000 ORIGINAL PAPER R. Pieterwas · G. Ecke () · R. Kosiba · H. Rößler TU Ilmenau, Institut für Festkörperelektronik, Postfach 100565, D-98684 Ilmenau, Germany R. Kosiba STU Bratislava, Faculty of Electrical Engineering and Information Technology, Department of Microelectronics, Ilkova 3, 81219 Bratislava, Slovakia R. Pieterwas IGH Automation GmbH, Bahnhofstr. 35, 40467 Langenfeld, Germany