Reconstruction Of The Edge Z eff Profile From Bremsstrahlung Data Via Extensions Of Independent Component Analysis On TEXTOR G. Verdoolaege , G. Telesca * , G. Van Oost and G. Van Den Berge Department of Applied Physics, Ghent University, Rozier 44, 9000 Gent, Belgium 1. Introduction The understanding of the behaviour of impurities is a critical issue in tokamak physics. Firstly, these impurities increase the fuel dilution and, at reactor parameters, they are responsible for a considerable power loss from the plasma core by bremsstrahlung. Secondly, they largely determine the edge radiation profile and finally, they can affect the growth rate of plasma instabilities. At the tokamak TEXTOR (Forschungszentrum Jülich, Germany), we run a diagnostic for determining the effective ion charge Z eff from bremsstrahlung measurements in the visible [1]. To obtain a radial profile for Z eff from our line-integrated data, an Abel-inversion is used. 2. Edge radiation, T e and n e uncertainties In our measuring wavelength region, 99 percent of the collected light coming from the central plasma consists of bremsstrahlung, but there is an incertitude at the plasma edge on the different contributions (e.g. recombination and black-body radiation and molecular bands) to the measured continuum signal. Also, at the edge only estimated values are routinely used for the electron temperature T e and the electron density n e , which are moreover determined at different toroidal locations. These various uncertainties render the interpretation of a Z eff profile problematic outside the centre. In fact, so far none of the available methods for the determination of Z eff has provided a full profile, which is at present a real challenge. Therefore, in addition to the efforts already being done to obtain a knowledge of the different radiation processes (and, consequently, of Z eff ) in the plasma edge from the physical point of view, we are also trying to extract quantitative information on the edge Z eff directly from the measured signals using methods from statistical data analysis. 3. Independent Component Analysis and extraction of Z eff information Recent advances in statistical data analysis have made it possible to decompose mixtures of signals into their individual components (sources), solely based on the statistics of the * Currently at LPP ERM/KMS, Brussels, Belgium 29th EPS Conference on Plasma Phys. and Contr. Fusion Montreux, 17-21 June 2002 ECA Vol. 26B, P-2.060 (2002)